RUN HEADER
Input block as executed
MODE: LIVE # fully resolved in the record at cutoff
RUN_ID: C
ROOT: ./run-C
QUESTION: >
"Among S&P/TSX Composite non-financial issuers that first recognized a
goodwill impairment charge in a fiscal year ending between 2022-01-01
and 2025-12-31, the median lag between the first quarter-end at which
the carrying amount of the entity's net assets exceeded its market
capitalization and the reporting date at which that impairment was
first recognized exceeds two quarters."
CLAIM_ORIGIN: "Too little, too late" critique of impairment-only
goodwill accounting; restated and bounded by AR
CUTOFF: 2026-08-22
POPULATION: S&P/TSX Composite constituents at the impairment date, IFRS
reporters, non-financial, carrying goodwill in the comparative
period. Financials and income trusts excluded. Issuers that
impaired without the indicator ever being present are retained
and coded as lag = not applicable, not dropped.
STATISTIC: Median over issuers, in quarters. One observation per issuer,
first impairment event only.
E0: SEDAR+ annual and interim financial statements and MD&A for
impairment recognition, amount, and CGU; audited balance sheets
for carrying amount of net assets; TSX or exchange-sourced shares
outstanding and quarter-end close for market capitalization.
EXCLUDED: Whether the impairment was correctly measured, and whether
recoverable amount was value in use or fair value less costs of
disposal. Out of scope.
MAX_CLAIMS_PER_TERRITORY: 12
SHUFFLE_SEED: 20260822
MAX_ROUNDS: 3
CONTROL_ARM: NO # unscoreable without a key; skipRuntime
| Field | Value |
|---|---|
| Runtime | Claude (Cowork), remote Linux container; subagents in isolated contexts |
| Convener model | claude-opus-5 (serving model may differ) |
| Python | 3.11.15 |
| Run date | 2026-08-23 (UTC), executed against CUTOFF 2026-08-22 |
| Phase 2 invocation | python3 strip.py ./run-C 2026-08-22 20260822 — exit 0 |
| Shuffle seed | 20260822 |
| strip_py_sha256 (script-printed) | c5539ae8bf8ff7f679aaaf99d677d9138eae735291743dc18d809655b01d23e7 |
| sha256sum strip.py (shell) | c5539ae8bf8ff7f679aaaf99d677d9138eae735291743dc18d809655b01d23e7 |
| Match | YES |
strip.py was extracted byte-for-byte from the ``python`` block in §5 of the package by
sed -n '264,521p', never retyped and never edited. 258 lines.
MODE: LIVE consequences, as applied
- ANACHRONISM vector disabled. No attack was filed under it in any round.
- §15 scoring not performed: no sealed key exists for a LIVE run.
CONTROL_ARM: NO— §11 skipped, per the input block.CUTOFFwas still passed to the script and enforced mechanically on every record.
Convener deviations, disclosed
1. Round-3 Inquisitor prompt contained one corrupted word. The framing sentence preceding the
verbatim brief read "their текст is narrow by construction" — a stray Cyrillic substitution for
"text", introduced by the Convener. The §6 brief itself was transmitted verbatim and unaffected.
Recorded here because it is the kind of defect that must not be discovered by an auditor first.
2. Blindness was enforced structurally, not only by instruction. For each blind phase the
permitted files were copied into an isolated inbox and the remainder of {ROOT} was moved to a
path outside the agents' working tree for the duration of the phase, then restored. keymap.json
was held outside the tree from the moment it was written until §12 item 20 was assembled, so it
was absent from the filesystem during every Round 1, 2, 3 and Adjudicator phase.
3. One subagent call terminated on an API session limit (Advocate, round 2) after it had written
its complete output file. The file was verified complete — all 20 re-entered IDs disposed, closing
CONCESSION line present — and used as filed. No retry, no reconstruction.
SELF-AUDIT BLOCK
strip_py_sha256: c5539ae8bf8ff7f679aaaf99d677d9138eae735291743dc18d809655b01d23e7
cutoff_applied: 2026-08-22
claims_filed_raw: 48
unparseable_lines: 0
malformed_field_rows: 0
rows_dropped_empty_claim: 0
claims_merged_away: 0
claims_filed: 48
tier_counts: E0=32 E1=6 E2=7 E3=3
e3_share_pct: 6
anachronism_flags: 0
anachronism_records: 0
anachronism_rate_pct: 0
date_integrity_failures: 0
cutoff_integrity_failures: 0
cutoff_integrity_failures_LB: 0
inadmissible_source_records: 1
shared_source_groups: 9
voice_lint_flags: 0
lint_exemptions_applied: 0
shuffle_seed: 20260822
--- counted by the Convener from agent output ---
lint_flags_resolved: 0 (lint.txt empty; nothing to resolve)
attacks_landed: PROVENANCE=33 INFERENCE=37 SELECTION=16 ANACHRONISM=0
attacks_retrieved: 132
attacks_asserted: 7
attacks_scored_zero: 57 (= line count of rejected-attacks-ALL.md, §12 item 19)
verdicts: HELD=21 NARROWED=19 CONTESTED=0 UNDETERMINED=1 KILLED=7
advocate_rebuilt: 48 (round1=21, round2=12, round3=15)
advocate_conceded: 12 (round1=7, round2=4, round3=1)
null_contested: 20 (round1=10, round2=6, round3=4)
null_not_contested: 67 (round1=38, round2=14, round3=15)
weakest_load_bearing: C-17 at tier E3 (KILLED in round 1)
conclusion_tier: E3 (broken: the E3 load-bearing link is killed.
Weakest SURVIVING load-bearing tier = E2, C-04)
reader_direction: REFUTED
reader_probability: 0.30
rounds_run: 3
reentered_subset_size: round1=48 (initial ledger; nothing re-entered) round2=20 round3=19
new_landed_attacks: round1=40 round2=44 round3=25
verdict_changes: round1=n/a (initial assignment, no prior map) round2=1 round3=0
(over the re-entered subset only)
halt_reason: ZERO_VERDICT_CHANGES
tripwire: NOT_FIRED (load-bearing HELD or NARROWED: 25 of 30)
tripwire_verdicts_moved: 0 (tripwire did not fire; no tripwire pass was run)
retrieval_calls_crucible: 549 total subagent tool invocations (see reconciliation note)
retrieval_calls_control: n/a — CONTROL_ARM: NO
agent_calls: 17Reconciliation notes on the counted block
attacks_landed is computed mechanically as filed minus rejected, per vector, from the
Inquisitor files and rejected-attacks-ALL.md:
| Vector | Filed (R1+R2+R3) | Rejected | Landed |
|---|---|---|---|
| PROVENANCE | 38 + 18 + 14 = 70 | 37 | 33 |
| INFERENCE | 17 + 16 + 5 = 38 | 1 | 37 |
| SELECTION | 5 + 18 + 8 = 31 | 15 | 16 |
| ANACHRONISM | 0 | 0 | 0 |
| (Null contests, no vector field) | 10 + 6 + 4 = 20 | 4 | 16 |
The Adjudicators' own per-round accepted counts sum to 109 (40 + 44 + 25); the vector arithmetic above
sums to 102. The seven-item gap is the three ledger-level Null contests filed in round 2 as L-1,
L-2, L-3 — which carry no C-nn and no vector — plus attack lines naming more than one target ID,
which the Adjudicator scored per target and the vector count scores per line. Both numbers are
reported rather than reconciled into one, because silently picking either would be an estimate.
retrieval_calls_crucible is the counted total of subagent tool invocations across all 17 agent
calls, which is what this runtime instruments. Retrieval calls are a subset of it and are not
separately metered. Per-agent: Cartographers 23/36/30/33; R1 Inquisitor 39, Null 31, Advocate 54,
Adjudicator 19; R2 Inquisitor 50, Null 29, Advocate 29, Adjudicator 20; R3 Inquisitor 61, Null 41,
Advocate 29, Adjudicator 23; Reader 2. The figure is stated as what it is rather than relabelled.
§14 abort conditions, checked
| Signal | Threshold | Observed | Result |
|---|---|---|---|
| cutoff_integrity_failures_LB | > 0 halts | 0 | pass |
| e3_share_pct | > 40 halts | 6 | pass |
| anachronism_records + date_integrity_failures | > 20% of 48 = >9.6 halts | 0 | pass |
| inadmissible_source_records includes LOAD-BEARING | any halts | 1 record, C-48, SUPPORTING | pass |
| rows_dropped_empty_claim | > 10% of 48 = >4.8 halts | 0 | pass |
| Inquisitor returns zero kills | halts | 7 claims killed across 3 rounds | pass |
| attacks_retrieved | = 0 halts | 132 | pass |
| Null contested everything or nothing | halts | 10 of 48 · 6 of 20 · 4 of 19 | pass |
| Advocate conceded nothing or everything | halts | 12 conceded, 48 rebuilt, 26 held | pass |
| voice_lint_flags unresolved at Phase 3 | halts | 0 flags raised | pass |
| strip_py_sha256 ≠ sha256sum | halts | identical | pass |
| Graveyard empty with >20 claims filed | continue + explain | 7 in graveyard | n/a |
No abort condition tripped. The run completed all three rounds.
SCRIPT REPORT
strip.py printed output, verbatim:
strip_py_sha256: c5539ae8bf8ff7f679aaaf99d677d9138eae735291743dc18d809655b01d23e7
cutoff_applied: 2026-08-22
claims_filed_raw: 48
unparseable_lines: 0
malformed_field_rows: 0
rows_dropped_empty_claim: 0
claims_merged_away: 0
claims_filed: 48
tier_counts: E0=32 E1=6 E2=7 E3=3
e3_share_pct: 6
anachronism_flags: 0
anachronism_records: 0
anachronism_rate_pct: 0
date_integrity_failures: 0
cutoff_integrity_failures: 0
cutoff_integrity_failures_LB: 0
inadmissible_source_records: 1
shared_source_groups: 9
voice_lint_flags: 0
lint_exemptions_applied: 0
shuffle_seed: 20260822STRIPPED LEDGER
02-stripped/stripped.md, complete, source lists included.
# STRIPPED LEDGER (cutoff=2026-08-22, seed=20260822, order randomised)
C-01
claim : Reuters reported IOSCO naming 'too little, too late' goodwill impairment a priority issue for securities regulators.
tier : E1
falsifier : The Reuters item does not attribute that characterisation to IOSCO.
load : SUPPORTING
sources : 1
- [E1] 2023-06-22 · Reuters wire report republished by Investing.com · https://investing.com/news/stock-market-news/global-watchdog-seeks-action-on-too-little-too-late-corporate-goodwill-writedown-3111225
C-02
claim : Majority non-recognition within the indicator year, as reported by ESMA plus André Filip Paugam, places the median indicator-to-recognition lag above four quarters.
tier : E2
falsifier : Firms not impairing in the indicator year are shown to impair predominantly within the following two quarters, keeping the median at or below two.
load : LOAD-BEARING
sources : 1
- [E2] 2013-01-07 · Inference from ESMA/2013/2 and André, Filip & Paugam abstract figures retrieved this run · https://www.esma.europa.eu/sites/default/files/library/2015/11/2013-02.pdf
C-03
claim : Corus disclosed that its carrying value was greater than its market enterprise value at August 31, 2022.
tier : E0
falsifier : The MD&A impairment discussion omits August 31, 2022 from the dates at which carrying value exceeded market enterprise value.
load : LOAD-BEARING
sources : 1
- [E0] 2023-10-30 · Corus Entertainment Inc. annual MD&A, year ended August 31, 2023, impairment discussion · https://assets.corusent.com/wp-content/uploads/2023/10/Corus-Entertainment-Annual-MDA-and-Financial-Statements-Oct-30-2023-vf.pdf
C-04
claim : Mandatory quarter-end indicator assessment under IAS 36 makes a lag beyond one quarter a departure from required timing.
tier : E2
falsifier : An IAS 36 or IAS 34 provision permitting deferral of impairment testing to a later period after an indicator arises.
load : LOAD-BEARING
sources : 1
- [E2] 2021-08-03 · Inference from Grant Thornton IAS 36 guidance · https://www.grantthornton.global/en/insights/articles/IFRS-ias-36/ifrs-ias-36-If-and-when-to-undertake-an-impairment-review/
C-05
claim : Grant Thornton lists carrying amount of net assets exceeding market capitalisation as an external indicator of potential impairment under IAS 36.
tier : E0
falsifier : The article omitting market capitalisation from its list of external indicators.
load : SUPPORTING
sources : 1
- [E0] 2021-08-03 · Grant Thornton International, 'IAS 36 - If and when to undertake an impairment review' · https://www.grantthornton.global/en/insights/articles/IFRS-ias-36/ifrs-ias-36-If-and-when-to-undertake-an-impairment-review/
C-06
claim : Lightspeed Commerce recognized a US$748.7 million goodwill impairment charge in the three months ended December 31 2022.
tier : E0
falsifier : The release recording the charge in a different quarter or at a materially different amount.
load : LOAD-BEARING
sources : 1
- [E0] 2023-02-02 · Lightspeed Commerce Inc. third quarter fiscal 2023 earnings release filed with SEC · https://www.sec.gov/Archives/edgar/data/1823306/000182330623000008/earningsreleasefy23q3.htm
C-07
claim : Corus disclosed carrying value greater than market enterprise value at August 31, 2022, May 31, 2023, August 31, 2023.
tier : E0
falsifier : The fiscal 2023 annual MD&A names a different set of period-ends at which carrying value exceeded market enterprise value.
load : LOAD-BEARING
sources : 1
- [E0] 2023-10-30 · Corus Entertainment annual MD&A and audited consolidated financial statements, fiscal 2023 · https://assets.corusent.com/wp-content/uploads/2023/10/Corus-Entertainment-Annual-MDA-and-Financial-Statements-Oct-30-2023-vf.pdf
C-08
claim : CFO Dive reported Kroll managing director Carla Nunes expected 2022 goodwill impairments to rise relative to 2021.
tier : E1
falsifier : The article quoting Nunes forecasting flat or declining 2022 impairment levels.
load : SUPPORTING
sources : 1
- [E1] 2022-04-06 · Maura Webber Sadovi, CFO Dive · https://www.cfodive.com/news/goodwill-impairment-expected-rise-reverse-course-2022-kroll/621614/
C-09
claim : ESMA found 43 percent of 235 sampled European issuers carried market capitalisation below equity book value at 31 December 2011.
tier : E0
falsifier : The ESMA report gives a different sample size or a materially different share of issuers below book value.
load : SUPPORTING
sources : 1
- [E0] 2013-01-07 · ESMA/2013/2, Review of practices related to goodwill impairment · https://www.esma.europa.eu/sites/default/files/library/2015/11/2013-02.pdf
C-10
claim : André Filip Paugam find only 20 to 25 percent of firms exhibiting economic impairment indications recognised an accounting impairment in the same period.
tier : E0
falsifier : The paper reports a recognition frequency above 50 percent conditional on economic impairment indications.
load : LOAD-BEARING
sources : 1
- [E0] 2017-03-20 · André, Filip & Paugam, Examining the Patterns of Goodwill Impairments in Europe and the US, SSRN abstract · https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2867340
C-11
claim : Hou Xue Zhang report 65 percent of 452 catalogued anomalies fail to clear an absolute t-value of 1.96 under their replication protocol.
tier : E0
falsifier : The paper reports a failure rate materially below 65 percent under the single-test hurdle.
load : SUPPORTING
sources : 1
- [E0] 2018-10-31 · Hou, Xue & Zhang, Replicating Anomalies, SSRN abstract page · https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3275496
C-12
claim : Lightspeed's disclosed lag from market-capitalization trigger to goodwill impairment recognition is zero quarters.
tier : E2
falsifier : Lightspeed's September 30 2022 balance sheet plus TSX close showing net assets already above market capitalization at that quarter-end.
load : LOAD-BEARING
sources : 1
- [E2] 2023-02-02 · Inference from Lightspeed Commerce Q3 FY2023 earnings release · https://www.sec.gov/Archives/edgar/data/1823306/000182330623000008/earningsreleasefy23q3.htm
C-13
claim : Grant Thornton's IFRS team states identifying an impairment indicator does not automatically trigger recognition of an impairment loss.
tier : E0
falsifier : The article stating that an external indicator mandates recognition of a loss without estimating recoverable amount.
load : LOAD-BEARING
sources : 1
- [E0] 2021-08-03 · Grant Thornton International, 'IAS 36 - If and when to undertake an impairment review' · https://www.grantthornton.global/en/insights/articles/IFRS-ias-36/ifrs-ias-36-If-and-when-to-undertake-an-impairment-review/
C-14
claim : Medians computed over bounded single-index populations across four fiscal years rest on few observations, so single-issuer reclassification can move the reported statistic.
tier : E3
falsifier : The bounded population of S&P/TSX Composite non-financial first-time impairers over 2022 to 2025 exceeds one hundred issuers.
load : SUPPORTING
sources : 1
- [E3] 2026-08-22 · Model prior on small-sample median instability · model prior, no external source
C-15
claim : Lightspeed stated its net assets exceeded its market capitalization as at December 31, 2022, which was an impairment trigger.
tier : E0
falsifier : The release contains no such sentence, or attributes the trigger to a date other than December 31, 2022.
load : LOAD-BEARING
sources : 1
- [E0] 2023-02-02 · Lightspeed Commerce Q3 fiscal 2023 earnings release filed on SEC EDGAR · https://www.sec.gov/Archives/edgar/data/1823306/000182330623000008/earningsreleasefy23q3.htm
C-16
claim : TELUS Corporation recognized no goodwill impairment for the TELUS digital experience cash-generating unit in 2024.
tier : E0
falsifier : The 2024 statements record a goodwill impairment loss for the TELUS digital experience unit.
load : SUPPORTING
sources : 1
- [E0] 2025-02-13 · TELUS Corporation audited consolidated financial statements, year ended December 31, 2024, filed as Form 40-F exhibit 99.4 · https://www.sec.gov/Archives/edgar/data/868675/000110465925012533/tu-20241231xex99d4.htm
C-17
claim : Threshold claims stating a median exceeds a bound far beneath the reference class central estimate resolve true more often than thresholds set near that estimate.
tier : E3
falsifier : A tabulation of resolved median-threshold claims shows no relationship between threshold placement relative to the reference class central estimate and resolution rate.
load : LOAD-BEARING
sources : 1
- [E3] 2026-08-22 · Model prior on threshold placement relative to reference class central tendency · model prior, no external source
C-18
claim : Lightspeed had 150,315,764 subordinate voting shares issued outstanding at September 30, 2022.
tier : E0
falsifier : The share capital note reports a different number of shares outstanding at September 30, 2022.
load : LOAD-BEARING
sources : 1
- [E0] 2022-09-30 · Lightspeed Commerce Inc. condensed interim consolidated financial statements, share capital note, September 30, 2022 · https://s28.q4cdn.com/517092977/files/doc_financials/2023/q2/Financial-Statements-FY23-Q2-Final.pdf
C-19
claim : CSA Staff Notice 51-364 named impairment of non-financial assets as an area affected by the current economic environment.
tier : E0
falsifier : The notice omitting impairment of non-financial assets from that list.
load : SUPPORTING
sources : 1
- [E0] 2022-11-03 · Canadian Securities Administrators Staff Notice 51-364 · https://www.osc.ca/sites/default/files/2022-11/csa_20221103_51-364_continuous-disclosure-review.pdf
C-20
claim : Grant Thornton lists carrying amount of net assets exceeding market capitalisation among the IAS 36 external indicators requiring an impairment review.
tier : E1
falsifier : The cited page omits that indicator, or IAS 36 paragraph 12 contains no market capitalisation indicator.
load : LOAD-BEARING
sources : 1
- [E1] 2021-08-03 · Grant Thornton International, IFRS IAS 36 guidance on if and when to undertake an impairment review · https://www.grantthornton.global/en/insights/articles/IFRS-ias-36/ifrs-ias-36-If-and-when-to-undertake-an-impairment-review/
C-21
claim : Lightspeed Commerce stated its net assets exceeded its market capitalization at December 31 2022, which was an impairment trigger.
tier : E0
falsifier : The release attributing the trigger to something other than net assets exceeding market capitalization.
load : LOAD-BEARING
sources : 1
- [E0] 2023-02-02 · Lightspeed Commerce Inc. third quarter fiscal 2023 earnings release filed with SEC · https://www.sec.gov/Archives/edgar/data/1823306/000182330623000008/earningsreleasefy23q3.htm
C-22
claim : No published dataset enumerates S&P/TSX Composite non-financial first goodwill impairments paired with net-assets-above-market-capitalization crossing quarter-ends.
tier : E2
falsifier : A pre-cutoff study, database extract, or regulator publication reports that issuer-level pairing for the S&P/TSX Composite.
load : LOAD-BEARING
sources : 1
- [E2] 2026-08-22 · Inference from retrieval across Kroll goodwill impairment studies, IOSCO, Reuters, issuer filings in this run · https://www.kroll.com/en/insights/goodwill-impairment
C-23
claim : Lightspeed Commerce recognized a non-cash goodwill impairment charge of US$748.7 million for the quarter ended December 31, 2022.
tier : E0
falsifier : The Q3 FY2023 interim financial statements state a different goodwill impairment amount or a different period of recognition.
load : LOAD-BEARING
sources : 1
- [E0] 2023-02-02 · Lightspeed Commerce Q3 fiscal 2023 earnings release filed on SEC EDGAR · https://www.sec.gov/Archives/edgar/data/1823306/000182330623000008/earningsreleasefy23q3.htm
C-24
claim : André Filip Paugam measure economic impairment indications using market-to-book below one, equity market value minus book value less than goodwill, negative EBITDA.
tier : E0
falsifier : The paper's three metrics exclude market-to-book below one as an economic impairment proxy.
load : SUPPORTING
sources : 1
- [E0] 2017-03-20 · André, Filip & Paugam, Examining the Patterns of Goodwill Impairments in Europe and the US, SSRN abstract · https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2867340
C-25
claim : Corus recorded a non-cash goodwill impairment charge of $350.0 million in the Television cash generating unit in fiscal 2022.
tier : E0
falsifier : Corus filings attribute the fiscal 2022 $350.0 million charge to a unit other than Television, or to assets other than goodwill.
load : LOAD-BEARING
sources : 1
- [E0] 2023-10-30 · Corus Entertainment annual MD&A and audited consolidated financial statements, fiscal 2023 · https://assets.corusent.com/wp-content/uploads/2023/10/Corus-Entertainment-Annual-MDA-and-Financial-Statements-Oct-30-2023-vf.pdf
C-26
claim : Jensen Kelly Pedersen conclude the majority of asset pricing factors can be replicated, contradicting broad claims that published quantitative findings collapse under re-measurement.
tier : E0
falsifier : The paper concludes that a minority of factors replicate.
load : SUPPORTING
sources : 1
- [E0] 2021-03-05 · Jensen, Kelly & Pedersen, Is There a Replication Crisis in Finance?, SSRN abstract page · https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3774514
C-27
claim : Lightspeed's goodwill impairment test as at September 30, 2022 resulted in no impairment of goodwill.
tier : E0
falsifier : The Q2 FY2023 statements record a goodwill impairment charge at September 30, 2022 or state that no test was performed.
load : LOAD-BEARING
sources : 1
- [E0] 2022-09-30 · Lightspeed Commerce Inc. condensed interim consolidated financial statements, three and six months ended September 30, 2022 · https://s28.q4cdn.com/517092977/files/doc_financials/2023/q2/Financial-Statements-FY23-Q2-Final.pdf
C-28
claim : Ballard's 2024 goodwill impairment note names the decline in the corporation's market capitalization as an impairment indicator.
tier : E0
falsifier : The impairment note lists indicators without reference to a decline in market capitalization.
load : LOAD-BEARING
sources : 1
- [E0] 2025-03-12 · Ballard Power Systems Inc. audited consolidated financial statements, year ended December 31, 2024, impairment indicators disclosure · https://www.ballard.com/wp-content/uploads/2025/03/Q4-24-Financial-Statements-FINAL.pdf
C-29
claim : Among ESMA-sampled issuers whose market capitalisation sat below equity book value, 47 percent recognised goodwill impairment losses in their 2011 financial statements.
tier : E0
falsifier : The report shows a majority of below-book issuers recognising impairment in the same reporting year.
load : LOAD-BEARING
sources : 1
- [E0] 2013-01-07 · ESMA/2013/2, Review of practices related to goodwill impairment · https://www.esma.europa.eu/sites/default/files/library/2015/11/2013-02.pdf
C-30
claim : Hayn Hughes report the goodwill write-off delay extends up to ten years for one third of the companies they examined.
tier : E0
falsifier : The paper attributes the ten-year tail to a materially smaller fraction than one third, or omits the claim.
load : SUPPORTING
sources : 1
- [E0] 2005-11-21 · Hayn & Hughes, Leading Indicators of Goodwill Impairment, SSRN abstract page · https://papers.ssrn.com/sol3/papers.cfm?abstract_id=850705
C-31
claim : TELUS Digital recognized a US$224 million non-cash goodwill impairment charge for the quarter ended June 30, 2025.
tier : E1
falsifier : TELUS Digital's Q2 2025 interim statements report a different goodwill impairment amount or period.
load : SUPPORTING
sources : 1
- [E1] 2025-08-01 · TELUS Digital second quarter 2025 results news release carried by Nasdaq · https://www.nasdaq.com/press-release/telus-digital-reports-second-quarter-2025-results-incremental-improvement-revenue
C-32
claim : Lightspeed attributed its December 31, 2022 goodwill impairment test to the carrying amount of net assets exceeding market capitalization.
tier : E0
falsifier : The Q3 FY2023 impairment note omits the carrying-amount-versus-market-capitalization language or cites a different trigger.
load : LOAD-BEARING
sources : 1
- [E0] 2022-12-31 · Lightspeed Commerce Inc. condensed interim consolidated financial statements, three and nine months ended December 31, 2022 · https://s28.q4cdn.com/517092977/files/doc_financials/2023/q3/Financial-Statements-FY23-Q3-Final.pdf
C-33
claim : Grant Thornton's IFRS team states IAS 36 requires assessing impairment indicators at the end of each reporting period.
tier : E0
falsifier : The article text requiring indicator assessment only at annual reporting dates.
load : LOAD-BEARING
sources : 1
- [E0] 2021-08-03 · Grant Thornton International, 'IAS 36 - If and when to undertake an impairment review' · https://www.grantthornton.global/en/insights/articles/IFRS-ias-36/ifrs-ias-36-If-and-when-to-undertake-an-impairment-review/
C-34
claim : Corus's earliest disclosed period-end of carrying value exceeding market enterprise value, August 31, 2022, matches its fiscal 2022 impairment recognition date.
tier : E2
falsifier : Corus quarter-end share counts times closing prices put market capitalization below net assets at any quarter-end before August 31, 2022.
load : LOAD-BEARING
sources : 1
- [E2] 2023-10-30 · Inference from the Corus fiscal 2023 annual MD&A record in this ledger · https://assets.corusent.com/wp-content/uploads/2023/10/Corus-Entertainment-Annual-MDA-and-Financial-Statements-Oct-30-2023-vf.pdf
C-35
claim : Lightspeed Commerce recognized no goodwill impairment loss in the fiscal year ended March 31, 2022.
tier : E0
falsifier : The FY2022 comparative column or the goodwill note discloses a goodwill impairment loss recognized before April 1, 2022.
load : LOAD-BEARING
sources : 1
- [E0] 2023-05-18 · Lightspeed Commerce Inc. audited consolidated financial statements, fiscal year ended March 31, 2023, comparative period · https://s28.q4cdn.com/517092977/files/doc_financials/2023/q4/Financial-Statements-FY23-Q4-Final.pdf
C-36
claim : BCE reported non-cash asset impairment charges totalling $2,190 million for 2024, mainly related to Bell Media's TV properties.
tier : E0
falsifier : The release stating a different full-year impairment total or a different attribution.
load : SUPPORTING
sources : 1
- [E0] 2025-02-06 · BCE Inc. 2024 fourth quarter and full-year results news release · https://www.prnewswire.com/news-releases/bce-reports-2024-q4-and-full-year-results-announces-2025-financial-targets-302369604.html
C-37
claim : The Canadian Press reported BCE recorded $2.11 billion of asset impairment charges in Q3 2024, mainly related to Bell Media properties.
tier : E1
falsifier : The article reporting a different amount or a different quarter for the charge.
load : SUPPORTING
sources : 1
- [E1] 2024-11-07 · Sammy Hudes, The Canadian Press, published by BNN Bloomberg · https://www.bnnbloomberg.ca/business/company-news/2024/11/07/bce-reports-q3-loss-on-asset-impairment-charge-cuts-revenue-guidance/
C-38
claim : Lightspeed reported total shareholders' equity of US$3,304,419 thousand at September 30, 2022.
tier : E0
falsifier : The interim statement of financial position shows a different total equity figure at September 30, 2022.
load : LOAD-BEARING
sources : 1
- [E0] 2022-09-30 · Lightspeed Commerce Inc. condensed interim consolidated statement of financial position at September 30, 2022 · https://s28.q4cdn.com/517092977/files/doc_financials/2023/q2/Financial-Statements-FY23-Q2-Final.pdf
C-39
claim : Lightspeed shareholders' equity was US$3,399.3 million at March 31, 2022, falling to US$2,539.5 million at December 31, 2022.
tier : E0
falsifier : The balance sheet in that release reports different total equity figures at either date.
load : SUPPORTING
sources : 1
- [E0] 2023-02-02 · Lightspeed Commerce Q3 fiscal 2023 earnings release balance sheet, SEC EDGAR · https://www.sec.gov/Archives/edgar/data/1823306/000182330623000008/earningsreleasefy23q3.htm
C-40
claim : Lightspeed's disclosed market-capitalization trigger date matches its impairment recognition date, giving a disclosed lag of zero quarters.
tier : E2
falsifier : An earlier Lightspeed quarter-end shows exchange-sourced market capitalization below audited net assets, moving the true crossing before December 31, 2022.
load : LOAD-BEARING
sources : 1
- [E2] 2023-02-02 · Inference from the Lightspeed Q3 fiscal 2023 earnings release records in this ledger · https://www.sec.gov/Archives/edgar/data/1823306/000182330623000008/earningsreleasefy23q3.htm
C-41
claim : BCE recorded $2,113 million of non-cash asset impairment charges against Bell Media in the quarter ended September 30, 2024.
tier : E1
falsifier : BCE's Q3 2024 interim financial statements report a different impairment amount or a different segment.
load : SUPPORTING
sources : 1
- [E1] 2024-11-07 · BCE third quarter 2024 results news release distributed by PR Newswire · https://www.prnewswire.com/news-releases/bce-reports-third-quarter-2024-results-302298110.html
C-42
claim : Lightspeed Commerce recognized a goodwill impairment charge of US$748.7 million in the three months ended December 31, 2022.
tier : E0
falsifier : The FY2023 audited statements record no goodwill impairment, a different amount, or recognition in a different fiscal quarter.
load : LOAD-BEARING
sources : 1
- [E0] 2023-05-18 · Lightspeed Commerce Inc. audited consolidated financial statements, fiscal year ended March 31, 2023 · https://s28.q4cdn.com/517092977/files/doc_financials/2023/q4/Financial-Statements-FY23-Q4-Final.pdf
C-43
claim : IAS 36 mandates only annual goodwill testing at a fixed date, so an indicator arising mid-year can defer recognition by up to three quarters.
tier : E2
falsifier : IAS 36 is shown to compel interim recognition in the quarter an external indicator first arises, removing the structural deferral.
load : SUPPORTING
sources : 1
- [E2] 2021-08-03 · Inference from Grant Thornton, IAS 36 - If and when to undertake an impairment review · https://www.grantthornton.global/en/insights/articles/IFRS-ias-36/ifrs-ias-36-If-and-when-to-undertake-an-impairment-review/
C-44
claim : Ballard Power Systems recognized a goodwill impairment loss of US$40,277 thousand on testing performed as at September 30, 2024.
tier : E0
falsifier : The goodwill note reports a different 2024 impairment amount or a measurement date other than September 30, 2024.
load : LOAD-BEARING
sources : 1
- [E0] 2025-03-12 · Ballard Power Systems Inc. audited consolidated financial statements, year ended December 31, 2024, goodwill note · https://www.ballard.com/wp-content/uploads/2025/03/Q4-24-Financial-Statements-FINAL.pdf
C-45
claim : Hayn Hughes report goodwill write-offs lag economic impairment of goodwill by an average of three to four years.
tier : E0
falsifier : The published paper states no average lag, or states an average lag of two quarters or less.
load : LOAD-BEARING
sources : 1
- [E0] 2005-11-21 · Hayn & Hughes, Leading Indicators of Goodwill Impairment, SSRN abstract page · https://papers.ssrn.com/sol3/papers.cfm?abstract_id=850705
C-46
claim : Ballard Power Systems recognized a goodwill impairment loss of US$23,991 thousand during 2023 relating to Ballard Motive Solutions.
tier : E0
falsifier : The 2023 comparative goodwill reconciliation shows no impairment loss or a different amount for Ballard Motive Solutions.
load : LOAD-BEARING
sources : 1
- [E0] 2025-03-12 · Ballard Power Systems Inc. audited consolidated financial statements, year ended December 31, 2024, comparative period · https://www.ballard.com/wp-content/uploads/2025/03/Q4-24-Financial-Statements-FINAL.pdf
C-47
claim : Corus Entertainment recognized a $350.0 million goodwill impairment charge in the quarter ended August 31, 2022.
tier : E0
falsifier : The fiscal 2022 comparative disclosure states a different amount or attributes the charge to a quarter other than the fourth.
load : LOAD-BEARING
sources : 1
- [E0] 2023-10-30 · Corus Entertainment Inc. annual MD&A with audited consolidated financial statements, year ended August 31, 2023 · https://assets.corusent.com/wp-content/uploads/2023/10/Corus-Entertainment-Annual-MDA-and-Financial-Statements-Oct-30-2023-vf.pdf
C-48
claim : BCE's market capitalization exceeded its book equity at every 2024 quarter-end.
tier : E3
falsifier : TSX close times shares outstanding below BCE consolidated equity at any 2024 quarter-end.
load : SUPPORTING
sources : 1
- [E3] — · Model prior, no external source retrieved in this run · noneLINT REPORT
01-ledger/lint.txt, verbatim:
nonevoice_lint_flags: 0. No claim was flagged for a hedge, a conjunction, or a length breach, so no
claim was returned to a Cartographer for restatement and none was dropped. lint_flags_resolved: 0
because there were none to resolve. lint_exemptions_applied: 0 — the nine-phrase compound-noun
exemption list never fired, so the zero conjunction count is not an artifact of masking.
INTEGRITY FILES
flags.txt
nonedate-failures.txt
nonecutoff-failures.txt
nonemalformed-rows.txt
noneinadmissible-records.txt
C-48 SUPPORTING dates: —SHARED SOURCES
01-ledger/shared-sources.txt, verbatim:
assets.corusent.com/wp-content/uploads/2023/10/corus-entertainment-annual-mda-and-financial-statements-oct-30-2023-vf.pdf -> C-03, C-07, C-25, C-34, C-47
model prior, no external source -> C-14, C-17
papers.ssrn.com/sol3/papers.cfm -> C-10, C-11, C-24, C-26, C-30, C-45
s28.q4cdn.com/517092977/files/doc_financials/2023/q2/financial-statements-fy23-q2-final.pdf -> C-18, C-27, C-38
s28.q4cdn.com/517092977/files/doc_financials/2023/q4/financial-statements-fy23-q4-final.pdf -> C-35, C-42
www.ballard.com/wp-content/uploads/2025/03/q4-24-financial-statements-final.pdf -> C-28, C-44, C-46
www.esma.europa.eu/sites/default/files/library/2015/11/2013-02.pdf -> C-02, C-09, C-29
www.grantthornton.global/en/insights/articles/ifrs-ias-36/ifrs-ias-36-if-and-when-to-undertake-an-impairment-review -> C-04, C-05, C-13, C-20, C-33, C-43
www.sec.gov/archives/edgar/data/1823306/000182330623000008/earningsreleasefy23q3.htm -> C-06, C-12, C-15, C-21, C-23, C-39, C-40INQUISITOR OUTPUT
The tripwire did not fire, so no tripwire pass was run and 03-attacks/tripwire-inquisitor.md
does not exist.
ROUND 1 — 03-attacks/inquisitor.md
INQUISITOR — Round 1 attacks
run: LIVE · cutoff 2026-08-22 · ANACHRONISM vector disabled for this run and not used
inbox: flags.txt = none · cutoff-failures.txt = none · inadmissible-records.txt = C-48 only
Structural finding driving the independence audit
48 records rest on 9 documents plus 3 unsourced model priors. shared-sources.txt catches 9 clusters;
it misses the BCE cluster (C-36/C-37/C-41 — one disclosure event, three URLs) and it splits the
Lightspeed single-event cluster across three normalised ids (SEC release, q2 PDF, q3 PDF, q4 PDF)
when all four are one issuer reporting one impairment. Source count on this ledger is not evidence depth.
ATTACKS
C-01 │ vector: PROVENANCE │ attack: Not primary and not independent of its own subject — the record cites Investing.com republishing a Reuters wire (Huw Jones) that itself quotes IOSCO's June 2023 public consultation paper verbatim ("We emphasize that the issue of 'too little, too late' goodwill impairment..."). The primary regulator document exists, is quoted in the very article cited, and appears nowhere in this ledger; a wire write-up is third-hand evidence of a regulator's position. │ evidence: investing.com/news/stock-market-news/global-watchdog-seeks-action-on-too-little-too-late-corporate-goodwill-writedown-3111225, dated 2023-06-22, attributing the phrase to IOSCO's consultation paper │ evidence_status: RETRIEVED
C-02 │ vector: PROVENANCE │ attack: The two supports are not independent. ESMA/2013/2 samples 235 European listed issuers at 31 Dec 2011 using market capitalisation below equity; André, Filip & Paugam's European panel runs 2006–2015 and proxies economic impairment with market-to-book below one. Overlapping issuer population, overlapping years, same market-based indicator — one European crisis-era observation counted twice. The record is also the third ledger draw on ESMA/2013/2 (with C-09 and C-29), and the second on SSRN 2867340 (with C-10 and C-24). │ evidence: ESMA/2013/2 paras 9, 19, 20; SSRN abstract 2867340 (2006–2015, US and Europe, MTB<1 among the three proxies) │ evidence_status: RETRIEVED
C-02 │ vector: INFERENCE │ attack: Neither source measures a lag, so neither can locate a median. ESMA reports a same-year recognition rate and never tracks when the non-recognising 53% subsequently impaired; André/Filip/Paugam likewise report a same-period recognition frequency. Weaker claim the evidence supports: "a majority of issuers exhibiting a market-based impairment indication did not recognise a goodwill impairment in the same annual reporting period." That places the median lag above zero annual reporting periods — the unit ESMA actually uses — and is silent on any quarter count, let alone on whether the median exceeds four quarters. │ evidence: ESMA/2013/2 para 20, "only 47% of the issuers whose equity exceeded market capitalisation recognised impairment losses on goodwill ... as at 31 December 2011"; SSRN 2867340, "only 20 to 25% of firms recognize impairments depending on the measure of economic impairment" │ evidence_status: RETRIEVED
C-02 │ vector: SELECTION │ attack: A competent opponent constructing a non-recognition base rate cites Ramanna & Watts (2012), "Evidence on the use of unverifiable estimates in required goodwill impairment," Review of Accounting Studies 17:749–780 (SSRN 1134943) — the canonical study of firms carrying market indications of goodwill impairment that do not impair, and of the incentives behind non-recognition. Also absent: Glaum, Landsman & Wyrwa (2018), "Goodwill Impairment: The Effects of Public Enforcement and Monitoring by Institutional Investors," The Accounting Review 93(6):149 (SSRN 3092658), which shows impairment timeliness varies with the public enforcement regime — precisely the variable that blocks transporting a 2011 European rate onto Canadian issuers in 2022–2025. │ evidence: SSRN 1134943; SSRN 3092658 / TAR 93(6):149 │ evidence_status: RETRIEVED
C-03 │ vector: PROVENANCE │ attack: Not an independent record — C-03 is a strict subset of C-07, both resting on the identical clause of the identical PDF. Two ledger entries, one sentence. With C-25, C-34 and C-47 this single Corus MD&A carries five records, and shared-sources.txt confirms the common source_id. │ evidence: Corus FY2023 annual MD&A, "The Company's carrying value being greater than its market enterprise value at August 31, 2022, May 31, 2023, and August 31, 2023" │ evidence_status: RETRIEVED
C-04 │ vector: PROVENANCE │ attack: Misreads the cited page in two places. Grant Thornton says "IAS 36 requires an entity to assess at the end of each reporting period whether there is any indication that an asset or CGU may be impaired" and, if so, that the entity "should estimate the recoverable amount" — nothing about quarters, and nothing about a lag being a departure from required timing. The same page says "the annual impairment test for an asset may be performed anytime during the annual period provided the test is performed at the same time every year." IAS 36 contains no quarter-end concept; interim frequency is an IAS 34 matter and not every IFRS filer reports quarterly. │ evidence: grantthornton.global IAS 36 article, 03 Aug 2021 │ evidence_status: RETRIEVED
C-04 │ vector: INFERENCE │ attack: Weaker claim the evidence supports: "IAS 36 requires an entity to assess at the end of each reporting period whether an impairment indicator exists and, where one does, to estimate the recoverable amount of the asset or CGU." Recognition follows only where recoverable amount falls below carrying amount, so an indicator-to-recognition lag is the compliant outcome, not a departure. The ledger's own C-27 is the counterexample: Lightspeed tested at 30 September 2022 on a live market-capitalisation indicator and recognised nothing. │ evidence: same Grant Thornton page (indicator → estimate recoverable amount); Lightspeed Q2 FY2023 interim statements │ evidence_status: RETRIEVED
C-05 │ vector: PROVENANCE │ attack: Duplicate of C-20 — same URL, same 2021-08-03 date, same sentence, same assertion, entered twice and tiered differently (E0 here, E1 at C-20), so identical evidence appears at two reliability grades and inflates both source count and tier spread. Six records (C-04, C-05, C-13, C-20, C-33, C-43) rest on this one article. │ evidence: grantthornton.global IAS 36 article listing "Carrying amount of the net assets of the entity is more than its market capitalisation"; shared-sources.txt cluster │ evidence_status: RETRIEVED
C-06 │ vector: PROVENANCE │ attack: Eighth-order duplication of one accounting entry. C-06, C-12, C-15, C-21, C-23, C-39 and C-40 cite the identical SEC earnings-release URL; C-32 (Q3 interim statements) and C-42 (FY23 audited statements) restate the same charge under different URLs. Eight records, one issuer, one impairment event — shared-sources.txt catches only the SEC subset. │ evidence: SEC EDGAR earningsreleasefy23q3.htm; s28.q4cdn.com FY23-Q3 and FY23-Q4 statements │ evidence_status: RETRIEVED
C-07 │ vector: INFERENCE │ attack: The MD&A sentence skips the two intervening quarter-ends — 30 November 2022 and 28 February 2023 — between 31 August 2022 and 31 May 2023, across which Corus's share price was falling continuously. Weaker claim the evidence supports: "Corus's fiscal 2023 annual MD&A names 31 August 2022, 31 May 2023 and 31 August 2023 as period-ends at which carrying value exceeded market enterprise value." It does not establish that the condition failed at the omitted quarter-ends, so the list is a disclosure artefact and cannot be read as a crossing series. │ evidence: Corus FY2023 annual MD&A impairment discussion, quoted date list │ evidence_status: RETRIEVED
C-08 │ vector: INFERENCE │ attack: Weaker claim the evidence supports: "In April 2022 a Kroll managing director said she expected 2022 goodwill impairments among US public companies to rise relative to 2021's roughly $8 billion." That is an ex ante forecast, drawn from an assessment of over 8,900 US-GAAP filers under ASC 350, which the article itself describes as an "early assessment" rather than a published study — carrying no evidence about IFRS issuers, about outturns, or about indicator-to-recognition lag. │ evidence: cfodive.com/news/goodwill-impairment-expected-rise-reverse-course-2022-kroll/621614/, 2022-04-06, Nunes: "Relative to 2021 I expect impairments to go up" │ evidence_status: RETRIEVED
C-08 │ vector: PROVENANCE │ attack: Trade-press report of an unpublished Kroll assessment rather than the Kroll study itself, and not independent of C-22, whose sole cited source is the Kroll goodwill-impairment hub. Same institution, same study programme, two ledger records. │ evidence: CFO Dive article describing a Kroll "early assessment"; C-22's source kroll.com/en/insights/goodwill-impairment │ evidence_status: RETRIEVED
C-09 │ vector: PROVENANCE │ attack: Not independent of C-29 or C-02 — paragraphs 19 and 20 of ESMA/2013/2 are consecutive sentences of one report, entered as three ledger records. One 2013 regulator snapshot of one year of European data is doing the work of three sources. │ evidence: ESMA/2013/2 paras 19 and 20; shared-sources.txt cluster C-02/C-09/C-29 │ evidence_status: RETRIEVED
C-10 │ vector: PROVENANCE │ attack: C-10 and C-24 are one paper on one SSRN abstract page, split into a result record and a method record. Two entries, one source, and both are also the substrate of C-02. │ evidence: SSRN 2867340, André, Filip & Paugam │ evidence_status: RETRIEVED
C-11 │ vector: INFERENCE │ attack: Weaker claim the evidence supports: "65% of the 452 anomalies in Hou, Xue and Zhang's data library fail to clear |t| ≥ 1.96 once microcaps are mitigated via NYSE breakpoints and value-weighted returns." The failure rate is a consequence of that specific portfolio construction, stated as such in the abstract, not a general property of published findings — and long-short factor returns are a different reference class from descriptive recognition frequencies read off filings. │ evidence: SSRN 3275496 abstract, "With microcaps mitigated via NYSE breakpoints and value-weighted returns, 65% of the 452 anomalies..." │ evidence_status: RETRIEVED
C-11 │ vector: PROVENANCE │ attack: Not independent of C-26. Jensen, Kelly & Pedersen is a direct methodological rebuttal to Hou, Xue & Zhang over the same anomaly universe; a paper and the paper written to answer it are one debate, not two supports pointing in opposite directions. │ evidence: SSRN 3275496 and SSRN 3774514 ("we develop and estimate a Bayesian model of factor replication, which leads to different conclusions") │ evidence_status: RETRIEVED
C-12 │ vector: PROVENANCE │ attack: Falsified by the ledger's own primary sources. Lightspeed's Q2 FY2023 interim statements state the 30 September 2022 goodwill test was performed because "the carrying amount of the Company's net assets exceeded the Company's market capitalization," and the Q3 FY2023 statements confirm "The Company had also performed goodwill impairment testing as at September 30, 2022." The disclosed market-capitalisation trigger is therefore live one full quarter before the 31 December 2022 recognition, so the disclosed lag is at least one quarter, not zero. The ledger cites that same Q2 PDF three times (C-18, C-27, C-38) and omits the trigger sentence every time. │ evidence: s28.q4cdn.com FY23-Q2 statements, goodwill note; s28.q4cdn.com FY23-Q3 statements, goodwill note │ evidence_status: RETRIEVED
C-12 │ vector: INFERENCE │ attack: Weaker claim the evidence supports: "Lightspeed's Q3 FY2023 release attributes the December 31, 2022 impairment test to the carrying amount of net assets exceeding market capitalization at that date." The ledger's own figures put the crossing earlier — US$3,304,419 thousand of equity at 30 September 2022 (C-38) against 150,315,764 subordinate voting shares, the only common class (C-18), implies a crossing threshold of US$21.98 per share, and the issuer states outright that the condition already held at that quarter-end. │ evidence: Lightspeed FY23-Q2 statements, statement of financial position and share capital note, plus the same document's goodwill note │ evidence_status: RETRIEVED
C-13 │ vector: PROVENANCE │ attack: Same article and same reading as C-33, and the ledger simultaneously runs C-43 off this page asserting IAS 36 "mandates only annual goodwill testing at a fixed date." One document is being made to yield mutually inconsistent records; at most one of C-13/C-33 and C-43 can be a faithful reading, which means at least one of them is a misread of a source the ledger cites six times. │ evidence: grantthornton.global IAS 36 article, "If any such indication exists, the entity should estimate the recoverable amount" alongside "IAS 36 requires an entity to assess at the end of each reporting period" │ evidence_status: RETRIEVED
C-14 │ vector: PROVENANCE │ attack: No source exists. The source_id is literally "model prior, no external source," shared with C-17 per shared-sources.txt — so the ledger's two small-sample records are not two supports but one unsourced assertion entered twice, one of them LOAD-BEARING. │ evidence: shared-sources.txt, "model prior, no external source -> C-14, C-17" │ evidence_status: ASSERTED
C-14 │ vector: INFERENCE │ attack: Weaker claim the evidence supports: "if the bounded population of S&P/TSX Composite non-financial first-time impairers over 2022–2025 is small, single-issuer reclassification can move a reported median." The antecedent is unestablished — the record's own falsifier concedes the population size is unknown — and the S&P/TSX Composite carries roughly 220–230 constituents, so a hundred-issuer bound is not self-evidently binding. │ evidence: the record's own falsifier text; no population count cited anywhere in the ledger │ evidence_status: ASSERTED
C-15 │ vector: INFERENCE │ attack: Weaker claim the evidence supports: "Lightspeed's Q3 FY2023 release states that at 31 December 2022 the carrying amount of its net assets exceeded its market capitalization, and that this triggered an impairment test." It does not state, and cannot support, that 31 December 2022 was the first period-end at which that condition held — Lightspeed's own Q2 FY2023 statements say the identical condition triggered a test at 30 September 2022. │ evidence: Lightspeed FY23-Q2 statements goodwill note vs FY23-Q3 release │ evidence_status: RETRIEVED
C-16 │ vector: INFERENCE │ attack: Weaker claim the evidence supports: "TELUS management determined that the recoverable amount of TELUS digital experience exceeded its carrying value as of the measurement date and therefore no impairment was recognized" — a management estimate the auditor designated a critical audit matter precisely because "Changes in these assumptions could have a significant impact on the recoverable amount." Non-recognition of an impairment is not evidence of absent economic impairment; the same business recognised a US$224 million goodwill impairment two quarters later (C-31). │ evidence: TELUS 2024 audited consolidated financial statements, Form 40-F ex 99.4, critical audit matter on TELUS digital experience │ evidence_status: RETRIEVED
C-17 │ vector: PROVENANCE │ attack: LOAD-BEARING on zero external evidence — the source line is "model prior, no external source," the same non-source as C-14, and the date field is the run date rather than a publication date. A claim about the empirical resolution rate of threshold forecasts is an empirical claim about a countable population; asserting it from a prior is not a source, and the record therefore has no falsifiable provenance at all. │ evidence: the record's own source line; shared-sources.txt pairing with C-14 │ evidence_status: ASSERTED
C-17 │ vector: SELECTION │ attack: The missing consideration is named by the record's own falsifier and never produced: a tabulation of resolved median-threshold claims against threshold placement relative to reference-class central tendency. No calibration archive, no resolution dataset, no scoring history is cited. A load-bearing claim about how a class of claims resolves, with no resolved claims in evidence, is unsupported by construction. │ evidence: named gap — the resolution tabulation the record's falsifier presupposes, absent from the ledger │ evidence_status: ASSERTED
C-18 │ vector: PROVENANCE │ attack: One PDF carries C-18, C-27 and C-38 — share count, impairment-test result and total equity — presented as three records. They are three line items from one interim filing, not three sources, and the ledger's use of them as independent inputs to a market-capitalisation crossing computation overstates the evidential base. │ evidence: s28.q4cdn.com FY23-Q2 statements; shared-sources.txt cluster │ evidence_status: RETRIEVED
C-19 │ vector: INFERENCE │ attack: Weaker claim the evidence supports: "CSA Staff Notice 51-364 lists impairment of non-financial assets as one of thirteen areas that 'may be impacted by the current economic environment.'" It sits in an undifferentiated enumeration alongside going concern, debt covenants, expected credit losses and material change reporting, in a notice covering fiscal years ended 31 March 2022 and 31 March 2021. It is not a finding, a priority designation, a deficiency rate, or a statement about impairment timing. │ evidence: CSA Staff Notice 51-364, 2022-11-03, "Financial Reporting and Disclosure during Economic Uncertainty" section, thirteen-item list quoted in full │ evidence_status: RETRIEVED
C-20 │ vector: PROVENANCE │ attack: Duplicate of C-05 — identical URL, identical publication date, identical indicator sentence — entered a second time at a different tier. The falsifier additionally invokes "IAS 36 paragraph 12," which the ledger never cites: the standard itself is nowhere in the source list, only a Big-Four-adjacent commentary on it. │ evidence: grantthornton.global IAS 36 article; absence of any IAS 36 primary text in the ledger │ evidence_status: RETRIEVED
C-21 │ vector: PROVENANCE │ attack: Verbatim near-duplicate of C-15 — same claim, same source, differing only in falsifier wording — and both sit inside the eight-record Lightspeed single-event cluster. Two entries, one sentence of one press release. │ evidence: SEC EDGAR earningsreleasefy23q3.htm; C-15 and C-21 claim text │ evidence_status: RETRIEVED
C-22 │ vector: SELECTION │ attack: The dataset the record says does not exist exists and is published by the record's own cited source. The Duff & Phelps / CFERF Canadian Goodwill Impairment Study series — 2012 Canadian Edition (13 Feb 2013), 2013 edition (19 Dec 2013), 2014 edition (18 Dec 2014) — is built on 621 S&P/TSX Composite Index issuers, reports goodwill impairment events by issuer, industry and year for 2007–2011, and cross-tabulates them against trading below book value: 72% of goodwill-impairing companies traded below book value at the 2008 peak versus 52% of all Canadian companies and 45% of large caps. That is the issuer-level pairing of first-order impairment events with market-value-below-book-value that C-22 asserts is unpublished, and every edition is listed on the Kroll goodwill-impairment hub the record cites as its evidence. │ evidence: feicanada.org/enews/file/CFERF studies/2012-2013/2012_GWI_Canada_FINAL2.pdf; kroll.com/en/insights/publications/goodwill-impairment listing the 2012, 2013 and 2014 Canadian editions │ evidence_status: RETRIEVED
C-22 │ vector: INFERENCE │ attack: Weaker claim the evidence supports: "this run's retrieval across the Kroll goodwill-impairment hub, IOSCO, Reuters and a handful of issuer filings did not surface such a dataset." A universal negative over all published datasets, database extracts and regulator publications cannot be discharged by one run's retrieval across five named sources, and the retrieval demonstrably missed the Canadian series on the very hub it consulted. │ evidence: the record's own source line, "Inference from retrieval across Kroll goodwill impairment studies, IOSCO, Reuters, issuer filings in this run"; kroll.com Canadian editions │ evidence_status: RETRIEVED
C-23 │ vector: PROVENANCE │ attack: Third rendering of one accounting entry (with C-06 and C-42) — the earnings release, the Q3 interim statements and the FY23 audited statements are the same issuer restating the same US$748.7 million charge. The claim text is near-identical to C-06's; only the falsifier differs. │ evidence: SEC EDGAR earningsreleasefy23q3.htm; FY23-Q3 statements reporting $748,712 thousand; FY23-Q4 audited statements │ evidence_status: RETRIEVED
C-24 │ vector: PROVENANCE │ attack: Same paper, same SSRN abstract page, as C-10 — the method half of a record the ledger already counts once for its result. Neither is independent of C-02, which is an inference off the same abstract. │ evidence: SSRN 2867340, three proxies: MTB below one, market value less book value below goodwill, negative EBITDA │ evidence_status: RETRIEVED
C-25 │ vector: PROVENANCE │ attack: Same MD&A as C-03, C-07, C-34 and C-47, and C-47 states the identical $350.0 million Television charge with only the period descriptor changed from "fiscal 2022" to "the quarter ended August 31, 2022." Five records, one document, one charge. │ evidence: Corus FY2023 annual MD&A, "a non-cash goodwill impairment charge of $350.0 million was recorded in the Television CGU"; shared-sources.txt cluster │ evidence_status: RETRIEVED
C-26 │ vector: INFERENCE │ attack: Weaker claim the evidence supports: "under a Bayesian hierarchical model of factor replication with cross-factor shrinkage, the majority of asset-pricing factors have non-zero posterior means, organised into 13 thematic clusters and holding out of sample across 93 countries." The disagreement with Hou–Xue–Zhang runs through estimator choice and portfolio construction, not through whether the underlying data survive re-measurement; nothing in the abstract supports the appended clause "contradicting broad claims that published quantitative findings collapse under re-measurement," which generalises a finance-factor result to published quantitative findings at large. │ evidence: SSRN 3774514, "we develop and estimate a Bayesian model of factor replication, which leads to different conclusions" │ evidence_status: RETRIEVED
C-27 │ vector: PROVENANCE │ attack: Misread by omission, and the omitted half is load-bearing elsewhere in this ledger. The cited statements say the 30 September 2022 test was performed because "the carrying amount of the Company's net assets exceeded the Company's market capitalization"; the record reports the "no impairment" result and strips the trigger clause from the same passage. Reinstating it falsifies C-12 and C-40 outright. │ evidence: s28.q4cdn.com FY23-Q2 statements, goodwill note: "The Company completed an impairment test of goodwill as at September 30, 2022 ... which demonstrated no impairment of goodwill as at September 30, 2022," triggered because net assets exceeded market capitalization │ evidence_status: RETRIEVED
C-28 │ vector: INFERENCE │ attack: Weaker claim the evidence supports: "Ballard's 2024 note names the decline in the Corporation's market capitalization as one of three indicators of potential impairment, alongside the initiation of a global corporate restructuring in September 2024 and indicators of slowing hydrogen and fuel cell policy implementation and market adoption." The restructuring arose in the same quarter as the 30 September 2024 test and is a sufficient independent trigger, so this record cannot date a market-capitalisation-indicator-to-recognition interval. │ evidence: Ballard Power Systems 2024 audited consolidated financial statements, impairment indicators disclosure, quoted three-item list │ evidence_status: RETRIEVED
C-29 │ vector: PROVENANCE │ attack: Consecutive paragraph to C-09's within one report, and the same report also underwrites C-02. Three ledger records, one 2013 ESMA snapshot of one year (2011) of one region's data — and that year is the trough of the European sovereign debt crisis, the least transportable point in the series to a 2022–2025 Canadian population. │ evidence: ESMA/2013/2 paras 19–20; shared-sources.txt cluster C-02/C-09/C-29 │ evidence_status: RETRIEVED
C-30 │ vector: PROVENANCE │ attack: Same paper and same SSRN abstract page as C-45 (850705), split into two records — the tail statistic here, the mean there — from a single abstract sentence pair. │ evidence: SSRN 850705, Hayn & Hughes abstract │ evidence_status: RETRIEVED
C-31 │ vector: INFERENCE │ attack: Weaker claim the evidence supports: "TELUS Digital's Q2 2025 US$224 million non-cash goodwill impairment was attributed to changes in key valuation assumptions including a higher weighted average cost of capital, a lower perpetual growth rate and lower cash flow forecasts arising from pricing pressure on margins." The release names no market-capitalisation trigger at all, so this record cannot serve as an observation in a market-cap-indicator-to-recognition series. │ evidence: nasdaq.com press release, 2025-08-01, quoted attribution │ evidence_status: RETRIEVED
C-31 │ vector: PROVENANCE │ attack: Not primary — a Nasdaq republication of TELUS Digital's own press release, when the primary (TELUS Digital's Q2 2025 interim consolidated financial statements / 6-K) exists. The ledger goes to the audited or interim filing for Lightspeed, Ballard, Corus and TELUS Corporation, and to a republished release only here, so the tier assignment is inconsistent with the ledger's own practice. │ evidence: nasdaq.com item identified as a company press release republished by Nasdaq │ evidence_status: RETRIEVED
C-32 │ vector: PROVENANCE │ attack: Lightspeed restating its own earnings release — same event, same trigger language, as C-06, C-15 and C-21, entered under a different URL so the shared-source audit does not catch it. The date field (2022-12-31) is the balance sheet date, not the publication date of the statements, which understates the record's distance from the event. │ evidence: s28.q4cdn.com FY23-Q3 statements, "the carrying amount of the Company's net assets exceeding the Company's market capitalization as at December 31, 2022. This triggered an impairment test" │ evidence_status: RETRIEVED
C-33 │ vector: PROVENANCE │ attack: Sixth record off one Grant Thornton article, and directly contradicted by C-43 drawn from the same page. The ledger cannot both assert that IAS 36 requires end-of-each-reporting-period assessment (here) and that it "mandates only annual goodwill testing at a fixed date" (C-43) on the authority of one document. │ evidence: grantthornton.global IAS 36 article, "IAS 36 requires an entity to assess at the end of each reporting period whether there is any indication that an asset or CGU may be impaired" │ evidence_status: RETRIEVED
C-34 │ vector: INFERENCE │ attack: Weaker claim the evidence supports: "31 August 2022 is the earliest period-end named in Corus's fiscal 2023 annual MD&A at which carrying value exceeded market enterprise value." Earliest-named-in-a-document-whose-reporting-window-opens-at-fiscal-2022 is not earliest crossing; the truncation is an artefact of the document chosen. The same MD&A dates the deterioration a year earlier — "The Company's share price has continued to decline meaningfully from August 31, 2021." │ evidence: Corus FY2023 annual MD&A impairment discussion │ evidence_status: RETRIEVED
C-34 │ vector: SELECTION │ attack: Missing: Corus's fiscal 2020 third quarter results, disclosing a $673.0 million goodwill impairment in the Television segment plus a $46.0 million Radio goodwill impairment and $67.8 million Radio broadcast licence impairment for the quarter ended 31 May 2020, on a net loss attributable to shareholders of $752.3 million. Corus's fiscal 2022 $350.0 million Television charge is therefore not its first goodwill impairment and 31 August 2022 is not its first crossing — which also removes Corus from any "first goodwill impairment" population the ledger is assembling, and removes whatever lag observation it contributes. │ evidence: prnewswire.com/news-releases/corus-entertainment-announces-fiscal-2020-third-quarter-results-301084272.html │ evidence_status: RETRIEVED
C-35 │ vector: PROVENANCE │ attack: Sourced to the wrong document for the period claimed — a negative claim about the fiscal year ended 31 March 2022 is evidenced by the comparative column of the FY2023 audited statements rather than by the FY2022 audited statements themselves, and that PDF is shared with C-42. A comparative column is a restatement, not the primary record of the period. │ evidence: s28.q4cdn.com FY23-Q4 statements; shared-sources.txt cluster C-35/C-42 │ evidence_status: RETRIEVED
C-36 │ vector: PROVENANCE │ attack: Misquotes the attribution the falsifier explicitly tests. The release states the $2,190 million was "mainly related to Bell Media's TV and radio properties to reflect a further decline in demand and spending in the traditional advertising market"; the record narrows this to "TV properties," dropping radio. │ evidence: prnewswire.com BCE 2024 Q4 and full-year results release, 2025-02-06 │ evidence_status: RETRIEVED
C-37 │ vector: PROVENANCE │ attack: Independence failure the shared-source audit misses because the URLs differ. C-37 (Canadian Press via BNN Bloomberg), C-41 (BCE's own release via PR Newswire) and C-36 (BCE's Q4/full-year release restating the same charge inside the $2,190 million total) are one corporate disclosure event; C-37 and C-41 carry the same date, 7 November 2024, and the Canadian Press item is a write-up of the release it is being counted as independent of. Three outlets reporting one company announcement are one source. │ evidence: bnnbloomberg.ca 2024-11-07 item vs prnewswire.com BCE Q3 2024 release 2024-11-07 vs prnewswire.com BCE Q4/full-year release 2025-02-06 │ evidence_status: RETRIEVED
C-39 │ vector: PROVENANCE │ attack: Seventh record drawn off the single SEC earnings-release URL, and the equity figures are inputs the ledger uses for a market-capitalisation crossing computation it never performs — the March 2022 and December 2022 endpoints straddle the 30 September 2022 quarter-end where the issuer says the crossing already existed, and that quarter-end is omitted from the record. │ evidence: SEC EDGAR earningsreleasefy23q3.htm balance sheet; FY23-Q2 statements equity of US$3,304,419 thousand at 30 September 2022 │ evidence_status: RETRIEVED
C-40 │ vector: PROVENANCE │ attack: False on the ledger's own primary sources, and its own falsifier fires without needing exchange data. Lightspeed's Q2 FY2023 interim statements say the 30 September 2022 test was triggered because "the carrying amount of the Company's net assets exceeded the Company's market capitalization," and the Q3 FY2023 statements confirm "The Company had also performed goodwill impairment testing as at September 30, 2022." The disclosed trigger date is therefore 30 September 2022, the recognition date 31 December 2022, and the disclosed lag at least one quarter — not zero. │ evidence: s28.q4cdn.com FY23-Q2 and FY23-Q3 statements, goodwill notes │ evidence_status: RETRIEVED
C-41 │ vector: PROVENANCE │ attack: The issuer's own press release counted as a separate source from the Canadian Press summary of that release (C-37) and from the full-year release restating it (C-36); and it is a release rather than the Q3 2024 interim financial statements its own falsifier appeals to. One event, three records, none of them the interim filing. │ evidence: prnewswire.com BCE Q3 2024 results release, 2024-11-07 │ evidence_status: RETRIEVED
C-42 │ vector: PROVENANCE │ attack: Third statement of the same US$748.7 million charge (C-06, C-23), and shares its PDF with C-35 — the audited FY2023 statements are Lightspeed's own confirmation of Lightspeed's own release, not corroboration from a second party. │ evidence: s28.q4cdn.com FY23-Q4 statements; shared-sources.txt cluster C-35/C-42 │ evidence_status: RETRIEVED
C-43 │ vector: PROVENANCE │ attack: Flatly contradicted by the source it cites. The Grant Thornton page states "IAS 36 requires an entity to assess at the end of each reporting period whether there is any indication that an asset or CGU may be impaired," so IAS 36 does not "mandate only annual goodwill testing at a fixed date"; the annual-test timing sentence ("may be performed anytime during the annual period provided the test is performed at the same time every year") sets a minimum frequency for goodwill, not a permission to defer recognition once an indicator arises. The record also contradicts C-04 and C-33, both drawn from this same page. │ evidence: grantthornton.global IAS 36 article, 2021-08-03 │ evidence_status: RETRIEVED
C-44 │ vector: PROVENANCE │ attack: One Ballard PDF carries C-28, C-44 and C-46 — indicator language, 2024 charge, 2023 comparative charge — presented as three records. Three line items from one filing, and the 2023 figure is a comparative rather than a contemporaneous record. │ evidence: ballard.com Q4-24 financial statements, goodwill note; shared-sources.txt cluster │ evidence_status: RETRIEVED
C-45 │ vector: INFERENCE │ attack: Weaker claim the evidence supports: "Hayn & Hughes report that goodwill write-offs lag the economic impairment of goodwill by an average of three to four years, in a distribution where for a third of the companies examined the delay can extend up to ten years." A mean drawn from a distribution with that right tail sits well above its own median; a three-to-four-year mean is arithmetically compatible with a median materially below three years. A mean cannot underwrite a median threshold, and the ledger pairs this record with C-30, which supplies exactly the tail statistic that makes the mean unrepresentative. │ evidence: SSRN 850705 abstract, both sentences │ evidence_status: RETRIEVED
C-45 │ vector: PROVENANCE │ attack: Stale and regime-mismatched for the use made of it. A 2005 SSRN posting on US acquisitions, proxying "economic impairment" from segment-level performance and original acquisition characteristics under US GAAP straddling the pre-SFAS 142 amortisation regime — not IAS 36, not a market-capitalisation indicator, not Canadian issuers, not 2022–2025. The abstract itself frames the question as whether disclosures let investors predict impairment post-amortisation-abolishment, and concludes the disclosures are inadequate; it is not a lag measurement designed for transport. │ evidence: SSRN 850705, Hayn & Hughes, "Leading Indicators of Goodwill Impairment," posted 2005-11-21, framed around Financial Accounting Standard No. 142 │ evidence_status: RETRIEVED
C-45 │ vector: SELECTION │ attack: Missing: Li & Sloan (2017), "Has goodwill accounting gone bad?", Review of Accounting Studies 22(2) (SSRN 1466271) — post-SFAS 142 evidence on delayed goodwill write-off timing that directly supersedes Hayn & Hughes's pre-142 sample and is the standard citation for the "delayed recognition" claim. Also missing: the IASB's own Exposure Draft Business Combinations—Disclosures, Goodwill and Impairment (March 2024) and the IFRS Foundation staff review "Goodwill and impairment: academic evidence" — the standard-setter's own treatment of late recognition under the impairment-only model. The ledger cites one Grant Thornton commentary six times and the standard-setter zero times. │ evidence: SSRN 1466271 / RAST 22(2); ifrs.org/projects/work-plan/goodwill-and-impairment/ and ifrs.org "Goodwill and impairment: academic evidence" │ evidence_status: RETRIEVED
C-46 │ vector: INFERENCE │ attack: Weaker claim the evidence supports: "Ballard recognised a US$23,991 thousand goodwill impairment in 2023 related to the restructuring and effective closure of Ballard Motive Solutions in the UK, presented within discontinued operations." A closure-driven write-off inside discontinued operations is not an indicator-triggered CGU impairment test outcome and does not belong in a market-capitalisation-indicator-to-recognition series; the record's framing conceals that distinction. │ evidence: ballard.com Q4-24 financial statements, goodwill note, comparative 2023 disclosure and discontinued-operations presentation │ evidence_status: RETRIEVED
C-47 │ vector: PROVENANCE │ attack: Restates C-25 from the same MD&A with only the period descriptor changed, and the fifth record off that one document. It is also a fiscal 2023 filing's comparative disclosure of a fiscal 2022 charge, not the fiscal 2022 primary record the falsifier ("the fiscal 2022 comparative disclosure") tacitly concedes it is testing. │ evidence: Corus FY2023 annual MD&A; shared-sources.txt cluster C-03/C-07/C-25/C-34/C-47 │ evidence_status: RETRIEVED
C-48 │ vector: PROVENANCE │ attack: No source at all. The source line reads "Model prior, no external source retrieved in this run," the date field is "—", and inadmissible-records.txt flags the record for unusable dates. The claim is a four-observation check against public data — TSX closes and BCE's quarterly consolidated equity at each 2024 quarter-end — that was never performed, and it is the third of three unsourced model-prior records (with C-14 and C-17) that the ledger nonetheless tiers and loads. │ evidence: inadmissible-records.txt, "C-48 SUPPORTING dates: —"; the record's own source line │ evidence_status: RETRIEVED
UNATTACKED
C-38.
CONCESSION
CONCESSION: UNMOVED — The ledger's spine does not survive its own sources. C-12 and C-40, both LOAD-BEARING, assert a zero-quarter Lightspeed lag that the ledger's own Q2 FY2023 filing contradicts in a sentence the ledger cites that PDF three times without ever quoting: the market-capitalisation trigger was live at 30 September 2022 and produced a test that recognised nothing. C-04 and C-43 are read off one Grant Thornton page in mutually exclusive directions, so at least one is a misread of a source carrying six records. C-22's negative existence claim is refuted by the Duff & Phelps / CFERF Canadian Goodwill Impairment Study, an S&P/TSX Composite dataset published on the Kroll hub C-22 cites as its evidence. C-34's "earliest disclosed crossing" is an artefact of a document window that opens after Corus had already written off $673.0 million of Television goodwill at 31 May 2020. And C-02, the record that actually carries the median-above-four-quarters conclusion, rests on two non-independent European sources neither of which measures a lag in any unit at all. Forty-eight records, nine documents, three unsourced priors, and the single most load-bearing quantity in the ledger — the lag — measured nowhere.
ROUND 2 — 03-attacks/inquisitor-r2.md
INQUISITOR — ROUND 2 (cutoff 2026-08-22 · MODE LIVE · ANACHRONISM disabled, zero filed)
Ledger control files: flags.txt = none · cutoff-failures.txt = none · inadmissible-records.txt = none.
shared-sources.txt drove three independence audits: the SSRN landing-page cluster (C-11/C-26/C-45), the single
Ballard PDF (C-28/C-46), and the single Lightspeed press release (C-15/C-21). All three held up as real defects.
ATTACKS
C-04 │ vector: SELECTION │ attack: the claim states IAS 36 as if testing were indicator-gated, omitting IAS 36 ¶10, which requires goodwill and indefinite-life intangibles to be tested annually irrespective of any indication — in a ledger whose subject is goodwill, this is the omission that decides the timing question, and the record's own cited page states it │ evidence: IAS 36 ¶10 ("Irrespective of whether there is any indication of impairment, an entity shall also: (a) test an intangible asset with an indefinite useful life or an intangible asset not yet available for use for impairment annually…"); Grant Thornton page cited on the record: "annually, irrespective of whether there is an indication of impairment" │ evidence_status: RETRIEVED
C-04 │ vector: SELECTION │ attack: never names IAS 36 ¶12(d), the single indicator on which every crossing record in this ledger rests (C-07, C-15, C-21, C-27, C-28); a claim about indicator assessment written for this ledger that omits the operative indicator is not doing the work it is filed to do │ evidence: IAS 36 ¶12(d), "the carrying amount of the net assets of the entity is more than its market capitalisation" │ evidence_status: RETRIEVED
C-04 │ vector: INFERENCE │ attack: "estimate the recoverable amount" overstates the obligation; weaker claim the evidence supports — "IAS 36 ¶9 requires an entity to assess at the end of each reporting period whether an indication of impairment exists; where one does, the entity must determine recoverable amount, an obligation that for a goodwill CGU may be discharged by using the most recent detailed calculation made in a preceding period (¶99) and whose annual test may be fixed at any single date in the annual period (¶96)" — which also satisfies this record's own falsifier │ evidence: IAS 36 ¶96 ("may be performed at any time during an annual period, provided the test is performed at the same time every year") and ¶99 ("The most recent detailed calculation made in a preceding period … may be used in the impairment test of that unit in the current period provided all of the following criteria are met") │ evidence_status: RETRIEVED
C-04 │ vector: PROVENANCE │ attack: LOAD-BEARING claim about the text of a standard resting on one E2 advisory commentary explicitly filed as "Inference from", when the standard's text is freely retrievable; not primary, and the record never quotes a paragraph number │ evidence: grantthornton.global IAS 36 article (2021-08-03), confirmed live, is a practitioner overview citing only ¶96 │ evidence_status: RETRIEVED
C-07 │ vector: INFERENCE │ attack: the MD&A's causal framing runs against the ledger's gloss; weaker claim — "the fiscal 2023 annual MD&A names August 31 2022, May 31 2023 and August 31 2023 as period-ends at which carrying value exceeded market enterprise value, attributing all three to one share-price decline that 'continued to decline meaningfully from August 31, 2021', and does not state the condition's status at November 30 2022 or February 28 2023." That the list is "a disclosure set, not a crossing series" is the claimant's characterisation; the document's own wording leans toward a continuing condition │ evidence: Corus fiscal 2023 annual MD&A, verbatim: "The Company's share price has continued to decline meaningfully from August 31, 2021, which resulted in the Company's carrying value being greater than its market enterprise value at August 31, 2022, May 31, 2023, and August 31, 2023." │ evidence_status: RETRIEVED
C-07 │ vector: SELECTION │ attack: asserts silence about two quarter-ends from a document that covers neither; the interim disclosures for those quarter-ends are absent from the ledger │ evidence: Corus fiscal 2023 first quarter results (prnewswire.com/news-releases/corus-entertainment-announces-fiscal-2023-first-quarter-results-301721080.html) and second quarter results, quarter ended February 28 2023 (newswire.ca/news-releases/corus-entertainment-announces-fiscal-2023-second-quarter-results-816994938.html) — retrieved; the Q2 document carries no impairment-indicator or carrying-value-versus-market language at all, which is the evidence the record should have produced instead of inferring silence from the annual filing │ evidence_status: RETRIEVED
C-07 │ vector: SELECTION │ attack: the ledger cites this MD&A for dates and omits the recognition sitting in the same document — $690.0M of fiscal 2023 impairments, $590.0M in Q3 (ended May 31 2023, including $295.2M of goodwill, $219.8M broadcast licences, $175.0M brands) and $100.0M in Q4, all in the Television CGU; that is a crossing-to-recognition pair the ledger's C-14 does not count │ evidence: same Corus fiscal 2023 annual MD&A and financial statements PDF cited on this record │ evidence_status: RETRIEVED
C-08 │ vector: PROVENANCE │ attack: not independent of C-22 — C-22's evidence base is "retrieval across Kroll goodwill impairment studies" resolved to kroll.com/en/insights/goodwill-impairment, and C-08 is trade-press coverage of Kroll's own research programme and its managing director's forecast; the ledger's impairment-trend context originates from one commercial research shop across two records │ evidence: cfodive.com article headline "…: Kroll" quoting Kroll MD Carla Nunes; kroll.com/en/insights/goodwill-impairment cited on C-22 │ evidence_status: RETRIEVED
C-08 │ vector: SELECTION │ attack: the ledger books an April 2022 forecast and never books the outcome; Kroll measured it and published, pre-cutoff, on the same hub C-22 cites │ evidence: Kroll, "Goodwill Impairment Trends Through H1 2022" (Valuation Insights Q3 2022, kroll.com/en/insights/publications/valuation/valuation-insights-third-quarter-2022/goodwill-impairment-trends-through-h1-2022); BVR, "Goodwill impairments YTD 2022 are up, per Kroll analysis"; Kroll 2025 U.S. Goodwill Impairment Study │ evidence_status: RETRIEVED
C-11 │ vector: SELECTION │ attack: the single most-cited rebuttal to this exact statistic is absent — Chen & Zimmermann reproduce 319 characteristics and find that of the 161 clearly significant in the original papers, 98% have t-stats above 1.96, and state explicitly that the residual characteristics "were insignificant in the original papers or are modifications of the originals created by Hou, Xue, and Zhang (2020)" │ evidence: Andrew Y. Chen and Tom Zimmermann, "Open Source Cross-Sectional Asset Pricing", FEDS 2021-037, June 2021, federalreserve.gov/econres/feds/open-source-cross-sectional-asset-pricing.htm │ evidence_status: RETRIEVED
C-11 │ vector: INFERENCE │ attack: the denominator is not what the claim implies; weaker claim — "65 percent of the 452 characteristic portfolios in Hou, Xue and Zhang's own library, as HXZ construct them under NYSE breakpoints and value weighting, have absolute t-values below 1.96." That is not 65 percent of published anomalies failing, because the 452 include characteristics that were never significant in their originating papers and HXZ-created modifications of originals │ evidence: Chen & Zimmermann abstract (above), which partitions HXZ's library on precisely that ground; HXZ SSRN abstract text as cited on the record │ evidence_status: RETRIEVED
C-11 │ vector: PROVENANCE │ attack: the source is an SSRN abstract landing page — the authors' own self-summary, not the paper's tables — dated to a 2018 working-paper posting and superseded by the published version; the record verifies a number against a marketing blurb │ evidence: papers.ssrn.com/sol3/papers.cfm?abstract_id=3275496 returns only abstract text; the published version is cited in the literature as Hou, Xue, and Zhang (2020) │ evidence_status: RETRIEVED
C-14 │ vector: PROVENANCE │ attack: the record has no source at all — its single entry is self-declared "model prior, no external source", self-dated to the cutoff date; yet what it carries is not a prior but a factual population count and a named enumeration of three issuers, which is a retrievable fact filed as an unauditable intuition │ evidence: the record's own source line, "[E3] 2026-08-22 · Model prior on small-sample median instability · model prior, no external source" │ evidence_status: ASSERTED
C-14 │ vector: INFERENCE │ attack: the enumeration is contradicted by the ledger's own records; weaker claim — "on this run's retrieval the claimant identified three issuers with a determinable crossing-to-recognition pair." Corus is a fourth: C-07's own source names the crossings (August 31 2022, May 31 2023, August 31 2023) and the same document records the recognition ($295.2M of goodwill within $590.0M of Q3 fiscal 2023 charges). Separately the falsifier is not a falsifier of this claim — a population of four defeats "three", and a population of ninety-nine never trips the stated hundred-issuer threshold │ evidence: Corus Entertainment fiscal 2023 annual MD&A and financial statements, the source already on C-07 │ evidence_status: RETRIEVED
C-14 │ vector: SELECTION │ attack: the precedent construction for exactly this population is absent — Kroll (as Duff & Phelps) ran a Canadian Goodwill Impairment Study, 2012, 2013 and 2014 editions, the 2012 edition jointly with FEI Canada; that the series was built and then discontinued after 2014 is the fact a competent opponent puts on the table before asserting a three-issuer population │ evidence: kroll.com/en/insights/publications/goodwill-impairment (Canadian editions 2012, 2013, 2014); feicanada.org "2012 Goodwill Impairment Study Canadian Edition" │ evidence_status: RETRIEVED
C-15 │ vector: PROVENANCE │ attack: C-15 and C-21 are the same claim, word for word, on the same document at the same URL, entered as two ledger records differing only in falsifier wording; the ledger's evidentiary basis for this proposition is one sentence in one press release, not two records │ evidence: shared-sources.txt maps sec.gov/…/earningsreleasefy23q3.htm to both C-15 and C-21; single fetch of that URL serves both records │ evidence_status: RETRIEVED
C-15 │ vector: SELECTION │ attack: the record reads the document only for trigger language and omits two facts in it — that December 31 is Lightspeed's fixed annual goodwill testing date, and that the test produced a $748.7 million non-cash goodwill impairment charge. The first matters under IAS 36 ¶96: the December 31 2022 test would have occurred irrespective of any trigger, so the record's framing of the test as trigger-caused is not what the document supports │ evidence: Lightspeed Q3 FY2023 release, verbatim: "The Company's net assets exceeded its market capitalization as at December 31, 2022 which was an impairment trigger for the Company" and "The Company conducts its annual goodwill impairment test every December 31"; charge of $748.7 million │ evidence_status: RETRIEVED
C-15 │ vector: INFERENCE │ attack: the negative this record hedges on is already settled affirmatively inside the ledger; weaker claim — "December 31 2022 was not the first period-end at which net assets exceeded market capitalization, because C-27's source records the same condition at September 30 2022." The record presents as an open question a matter its own ledger closes │ evidence: Lightspeed Q2 FY2023 interim statements (C-27's source): "The carrying amount of the Company's net assets exceeded the Company's market capitalization as at September 30, 2022. This triggered an impairment test." │ evidence_status: RETRIEVED
C-16 │ vector: PROVENANCE │ attack: sources = 1, but that source cannot carry three of the record's four factual components — TELUS Corporation's audited statements are presented in Canadian dollars and contain neither TELUS Digital's US$1.08bn market capitalization, nor its US$1,945M owners' equity, nor the US$224M June 30 2025 impairment; the third is borrowed from C-31's source, so C-16 and C-31 are not independent records │ evidence: sec.gov/…/tu-20241231xex99d4.htm — fetched; presentation currency Canadian dollars, critical audit matter limited to "Goodwill Impairment–Refer to Note 1(f) and 18(d)" and the finding that "Management determined that the recoverable amount of TELUS digital experience exceeded its carrying value as of the measurement date and, therefore, no impairment was recognized" │ evidence_status: RETRIEVED
C-16 │ vector: INFERENCE │ attack: the juxtaposition implies a tension the evidence does not establish; weaker claim — "TELUS Corporation determined that the recoverable amount of its TELUS digital experience CGU exceeded its carrying value at the 2024 measurement date, a judgement its auditor designated a critical audit matter; separately, the listed subsidiary's market capitalization was below its consolidated owners' equity at December 31 2024." Those quantities are not comparable: under IAS 36 ¶12(d) net assets above market capitalisation is an indicator obliging a test, not a measure of recoverable amount; the CGU carrying amount in TELUS's CAD consolidated statements is not the subsidiary's USD consolidated equity; and the market capitalisation of a majority-controlled subsidiary struck off its minority float is not a controlling-interest value — TELUS was buying additional TELUS Digital shares in October 2024 │ evidence: IAS 36 ¶12(d); "TELUS Acquires Additional Shares of TELUS Digital", prnewswire.com/news-releases/telus-acquires-additional-shares-of-telus-digital-302261639.html │ evidence_status: RETRIEVED
C-16 │ vector: SELECTION │ attack: TELUS Digital's own 2024 annual report on Form 20-F carries the subsidiary-level goodwill impairment conclusion for the identical year and the equity figure this claim uses, and would show whether two separate boards and auditors reached the same conclusion; absent from the ledger │ evidence: sec.gov/Archives/edgar/data/1825155/000162828025005299/tixt-20241231.htm │ evidence_status: RETRIEVED
C-19 │ vector: PROVENANCE │ attack: misread — the notice's list contains twelve areas, not thirteen. On the only alternative parse, splitting the compound first item into "known trends", "events" and "uncertainties", the count is fourteen. No parse of the sentence yields thirteen │ evidence: CSA Staff Notice 51-364, verbatim: "Some areas that may be impacted by the current economic environment include known trends, events and uncertainties, liquidity and capital resources, debt covenants, risk factor disclosure, impairment of non-financial assets, going concern, events after the reporting period, significant judgement and measurement uncertainties, expected credit losses, financial instrument risk disclosure, non-GAAP and other financial measures, and material change reporting." │ evidence_status: RETRIEVED
C-19 │ vector: INFERENCE │ attack: the sentence says "include", i.e. it is expressly non-exhaustive, so no count is a finding of the notice at all and the claim's precision is manufactured; weaker claim — "CSA Staff Notice 51-364 names impairment of non-financial assets among a non-exhaustive list of areas that may be impacted by the current economic environment" │ evidence: same verbatim sentence │ evidence_status: RETRIEVED
C-21 │ vector: PROVENANCE │ attack: duplicate of C-15 — identical claim text, identical URL, entered twice; the ledger's source count for this proposition is one document, and the two records' differing falsifiers do not create two pieces of evidence │ evidence: shared-sources.txt maps sec.gov/…/earningsreleasefy23q3.htm to both C-15 and C-21 │ evidence_status: RETRIEVED
C-21 │ vector: SELECTION │ attack: the ledger cites this press release for the trigger sentence while the audited statements it already holds at C-35 carry a fuller version of the same narrative plus the outcome; the record cites the weaker of two documents it possesses │ evidence: Lightspeed FY2023 audited statements, verbatim: "During the three months ended December 31, 2022, there were changes in macroeconomic conditions and the Company's share price and market capitalization decreased. This led to the carrying amount of the Company's net assets exceeding the Company's market capitalization as at December 31, 2022. This triggered an impairment test…", with a goodwill impairment loss of $748,712 thousand │ evidence_status: RETRIEVED
C-22 │ vector: PROVENANCE │ attack: the sole source is a compound, unauditable citation — "Inference from retrieval across Kroll goodwill impairment studies, IOSCO, Reuters, issuer filings in this run" — resolved to a single URL that is a directory page listing study editions; a directory page cannot evidence a universal negative about what datasets exist │ evidence: kroll.com/en/insights/publications/goodwill-impairment — fetched; it is an index of U.S. editions 2012–2025, European 2013–2020, Canadian 2012–2014, with no dataset survey │ evidence_status: RETRIEVED
C-22 │ vector: INFERENCE │ attack: non-retrieval is being reported as non-publication; weaker claim — "the claimant's retrieval in this run did not surface such a dataset." The search space the source line names (a Kroll hub, IOSCO, Reuters, issuer filings) excludes every place such a pairing would live: Audit Analytics, S&P Capital IQ, Compustat/WRDS extracts, and SEDAR+ full-text search │ evidence: the record's own source line enumerating its search space, against the fetched Kroll page which is a directory │ evidence_status: RETRIEVED
C-22 │ vector: SELECTION │ attack: the named precedent for exactly this construction applied to Canada is absent and sits on the page the record cites — Duff & Phelps/Kroll's Canadian Goodwill Impairment Study, 2012, 2013 and 2014 editions, the 2012 with FEI Canada; that the series exists and was discontinued after 2014 is the load-bearing fact for a claim about a 2022–2025 window │ evidence: kroll.com/en/insights/publications/goodwill-impairment; kroll.com/en/newsroom/duff-and-phelps-publishes-2013-canadian-goodwill-impairment-study; feicanada.org "2012 Goodwill Impairment Study Canadian Edition" │ evidence_status: RETRIEVED
C-26 │ vector: INFERENCE │ attack: "non-zero posterior means" is not the paper's replication criterion and as written is near-vacuous — under a hierarchical model with shrinkage toward a non-zero grand mean, posterior means are essentially never exactly zero; weaker and correct claim — "Jensen, Kelly & Pedersen report that 82.4 percent of factors replicate on the criterion that the posterior z-score equals or exceeds 1.96, equivalently that the posterior probability of a non-positive alpha falls below 2.5 percent, in a data set of 153 factors across 93 countries" │ evidence: published Journal of Finance version, verbatim criterion: "factor i is discovered if its z-score is greater than the critical value z̄ = 1.96, E(αi|α1,…,αN,τ)/√Var(αi|α1,…,αN,τ) ≥ z̄"; research-api.cbs.dk/ws/portalfiles/portal/95651880/…publishersversion.pdf │ evidence_status: RETRIEVED
C-26 │ vector: INFERENCE │ attack: the claim welds two separately numbered abstract findings into one proposition — (1) the majority can be replicated and (3) the factors work out of sample in a new data set covering 93 countries — producing "the majority … have non-zero posterior means, holding out of sample across 93 countries", which asserts that the same majority holds country by country. The abstract asserts no such conjunction │ evidence: SSRN abstract 3774514, verbatim: "The majority of asset pricing factors: (1) can be replicated, (2) can be clustered into 13 themes … (3) work out of sample in a new large data set covering 93 countries, and (4) have evidence that is strengthened (not weakened) by the large number of observed factors." │ evidence_status: RETRIEVED
C-26 │ vector: PROVENANCE │ attack: stale and non-primary — the record dates the source 2021-03-05, but SSRN records the paper as last revised 2022-03-07 and it was published in the Journal of Finance 78(5), 2023 (doi 10.1111/jofi.13249); the cited page is an abstract landing page, and the claim's misstatement of the replication criterion is exactly the error an abstract-only citation produces │ evidence: papers.ssrn.com/sol3/papers.cfm?abstract_id=3774514 ("Last revised March 7, 2022"); onlinelibrary.wiley.com/doi/10.1111/jofi.13249 │ evidence_status: RETRIEVED
C-27 │ vector: PROVENANCE │ attack: the source is misdated — 2022-09-30 is the balance-sheet date, not the publication date; the statements were "approved for issue by the Board of Directors of the Company on November 2, 2022", five weeks later. A source dated by period-end rather than issue date is effectively undated for cutoff arithmetic, and this ledger runs on cutoff arithmetic │ evidence: Lightspeed Q2 FY2023 interim statements PDF, verbatim approval line │ evidence_status: RETRIEVED
C-27 │ vector: SELECTION │ attack: the document that would establish whether September 30 2022 was itself the first crossing — Lightspeed's Q1 fiscal 2023 interim statements for the quarter ended June 30 2022 — is absent from the ledger, which leaves the ledger's "first period-end" question open at both ends rather than one │ evidence: named gap: Lightspeed Commerce Inc. condensed interim consolidated financial statements, three months ended June 30 2022 │ evidence_status: ASSERTED
C-28 │ vector: PROVENANCE │ attack: not independent of C-46 — both records rest on the identical PDF and the identical goodwill note; the ledger prints "sources: 1" on each, which conceals that Ballard's entire evidentiary basis here is one filing read twice │ evidence: shared-sources.txt maps ballard.com/wp-content/uploads/2025/03/q4-24-financial-statements-final.pdf to C-28 and C-46; single fetch serves both │ evidence_status: RETRIEVED
C-28 │ vector: SELECTION │ attack: the same note records the recognition those indicators produced, and the ledger nowhere books it — corporate goodwill written down to nil, $40,277 thousand, on impairment tests performed at September 30 2024. The ledger holds Ballard's 2023 amount (C-46) and Ballard's 2024 indicators (C-28) but not Ballard's 2024 amount or test date, which is precisely the crossing-to-recognition pair C-14 claims to be counting │ evidence: Ballard FY2024 audited consolidated financial statements, goodwill note: $40,277,000 written down to nil, tests performed at September 30 2024 │ evidence_status: RETRIEVED
C-31 │ vector: PROVENANCE │ attack: the cited page is Nasdaq's syndicated copy of a Business Wire release issued by the issuer — a redistribution channel presented as a source; the record's own falsifier is framed against TELUS Digital's Q2 2025 interim statements, which the record does not cite │ evidence: fetched page identifies publisher/distributor Business Wire, dated August 1 2025 │ evidence_status: RETRIEVED
C-31 │ vector: SELECTION │ attack: TELUS Digital's Q2 2025 interim consolidated financial statements furnished on Form 6-K carry the impairment note, the CGU identification and the actual indicator assessment — including whether IAS 36 ¶12(d) was among the indicators, which is the exact question the claim's final clause purports to answer from a press release; absent from the ledger │ evidence: SEC EDGAR CIK 1825155, TELUS International (Cda) Inc. / TELUS Digital │ evidence_status: RETRIEVED
C-31 │ vector: INFERENCE │ attack: "the release names no market-capitalisation trigger" is an argument from silence in a promotional document; weaker claim — "the release attributes the decline in recoverable amount to a higher weighted average cost of capital, a lower perpetual growth rate and lower cash flow forecasts arising from pricing pressure on margins, and does not discuss impairment indicators at all." A press release's silence on ¶12(d) is not the issuer's position on ¶12(d) │ evidence: release verbatim: "As at June 30, 2025, we recorded a non-cash goodwill impairment charge of $224 million" and "The recoverable amount was principally affected by changes in key valuation assumptions including higher weighted average cost of capital, lower perpetual growth rate and lower cash flow forecasts arising from pricing pressure on margins" — no indicator discussion of any kind │ evidence_status: RETRIEVED
C-35 │ vector: SELECTION │ attack: the record cites this document solely for the absence of a charge in the prior comparative year while the same statements record the recognition the ledger's whole thesis needs — a goodwill impairment loss of $748,712 thousand for the year ended March 31 2023 — together with the fullest trigger narrative in the ledger. The ledger declines to read the affirmative half of a document it already holds │ evidence: Lightspeed FY2023 audited consolidated financial statements: $748,712 thousand goodwill impairment loss, FY2022 comparative nil, and the December 31 2022 trigger paragraph quoted above │ evidence_status: RETRIEVED
C-35 │ vector: INFERENCE │ attack: a nil comparative is being read as an absence of events; weaker claim — "the FY2023 audited statements show nil goodwill impairment in the comparative column for the year ended March 31 2022." That establishes only that no loss was recognised before April 1 2022; it establishes nothing about whether a crossing occurred or a test was performed in that year, which the record's falsifier conflates │ evidence: same statements — comparative goodwill impairment line at nil, with no comparative disclosure of testing or indicators for FY2022 │ evidence_status: RETRIEVED
C-36 │ vector: INFERENCE │ attack: this record carries no goodwill figure at all; weaker claim — "BCE's release states non-cash asset impairment charges totalling $2,190 million for 2024, mainly related to Bell Media's TV and radio properties, attributed to a further decline in demand and spending in the traditional advertising market." Bell Media charges span goodwill and indefinite-life intangible assets (broadcast licences) and the release decomposes neither, so in a goodwill-impairment ledger the record supports no goodwill observation │ evidence: prnewswire.com/…/302369604.html — fetched; the release gives the aggregate and the attribution and states no goodwill component │ evidence_status: RETRIEVED
C-36 │ vector: PROVENANCE │ attack: not independent of C-41 or C-37 — C-36 and C-41 are both BCE's own investor-relations copy on PR Newswire and carry the identical phrase "mainly related to Bell Media's TV and radio properties"; C-37 is a Canadian Press rewrite of C-41's release published the same day. Three ledger records, one issuer's drafting │ evidence: both PR Newswire releases fetched, identical boilerplate; bnnbloomberg.ca CP article dated 2024-11-07, same day as the BCE release it reports │ evidence_status: RETRIEVED
C-36 │ vector: SELECTION │ attack: BCE's 2024 annual audited consolidated financial statements carry the impairment note with the CGU identification and the goodwill versus indefinite-life intangible split — the only document that would make these BCE records usable in a goodwill ledger; absent │ evidence: "Consolidated financial statements — BCE Inc. 2024 Annual financial report", filed on SEDAR+ (sedarplus.ca record id db4c271b5e9babec11c9396cf8eb666eac03d14b158f1d5dddfa669b576d13de) │ evidence_status: RETRIEVED
C-37 │ vector: PROVENANCE │ attack: the $2.11 billion figure is lifted from BCE's November 7 2024 release, which is C-41's source — same day, same document; the record concedes this in its own text, which means the ledger has knowingly entered a record of zero evidentiary value and still counts it toward its source totals │ evidence: bnnbloomberg.ca CP article by Sammy Hudes, 2024-11-07, verbatim "recorded $2.11 billion in asset impairment charges, mainly related to Bell Media's TV"; prnewswire.com/…/302298110.html same date │ evidence_status: RETRIEVED
C-37 │ vector: INFERENCE │ attack: weaker claim — "The Canadian Press reported the figure BCE published, rounded from $2,113 million to $2.11 billion." The article's genuinely independent content concerns the CRTC fibre-access decision, the share price closing below $40, and Scotiabank analyst commentary; none of it touches the impairment, so the record contributes nothing to the impairment proposition it is filed under │ evidence: same article — independent reporting confined to regulatory, market-reaction and analyst material │ evidence_status: RETRIEVED
C-41 │ vector: INFERENCE │ attack: "against Bell Media" asserts full segment attribution the release does not make; weaker claim — "BCE's Q3 2024 results news release states non-cash asset impairment charges totalling $2,113 million, mainly related to Bell Media's TV and radio properties." "Mainly related to" leaves an unquantified remainder outside those properties, and the release states no goodwill component of the $2,113 million │ evidence: prnewswire.com/…/302298110.html, verbatim: "non-cash asset impairment charges totalling $2,113 million, mainly related to Bell Media's TV and radio properties" │ evidence_status: RETRIEVED
C-41 │ vector: SELECTION │ attack: the record's own falsifier is written against BCE's Q3 2024 interim financial statements, which are the primary document and are not cited anywhere in the ledger — a falsifier aimed at a document the ledger cannot open is not a falsifier │ evidence: named gap: BCE Inc. Q3 2024 interim consolidated financial statements and impairment note, filed on SEDAR+ and as a 6-K exhibit; the cited release carries no allocation of the charge │ evidence_status: RETRIEVED
C-45 │ vector: PROVENANCE │ attack: misread, and the misread is caused by citing the preprint — the claim calls this "a US sample framed around SFAS 142" when the published abstract says the sample is pre-SFAS-142 goodwill and generalisability to the post-142 regime is an inference the authors argue for, not the frame they work in. The 2005 SSRN abstract the ledger cites omits that caveat sentence entirely; the 2006 published version carries it │ evidence: Journal of Accounting, Auditing and Finance 2006, doi 10.1177/0148558X0602100303, verbatim: "Although most of our analyses are conducted on goodwill generated before the introduction of Statement of Financial Accounting Standards No. 142 (SFAS 142), certain features of the sample and the analysis suggest that the results are generalizable to the current reporting regime. Sensitivity tests on a smaller sample of goodwill write-offs made upon the adoption of SFAS 142 confirm this expectation." — a sentence absent from papers.ssrn.com/sol3/papers.cfm?abstract_id=850705 │ evidence_status: RETRIEVED
C-45 │ vector: INFERENCE │ attack: weaker claim — "in a sample of goodwill generated before SFAS 142, write-offs lagged the authors' own fundamentals-based construct of economic impairment by an average of three to four years, with the delay extending to ten years for one-third of companies; the authors infer generalisability to the post-142 regime from sensitivity tests on a smaller sample." A three-to-four-year lag measured under an amortisation regime is not transportable into an IAS 36 ledger without the generalisation step the authors merely assert │ evidence: same published abstract │ evidence_status: RETRIEVED
C-45 │ vector: SELECTION │ attack: the post-SFAS-142 literature that tests exactly that generalisation is absent from the ledger — Ramanna & Watts, "Evidence on the use of unverifiable estimates in required goodwill impairment testing" (Review of Accounting Studies, 2012), and Li & Sloan, "Has goodwill accounting gone bad?" (Review of Accounting Studies, 2017), both on the timeliness of impairment recognition under the impairment-only regime │ evidence: named gap, as above │ evidence_status: ASSERTED
C-46 │ vector: PROVENANCE │ attack: the source for a 2023 event is the comparative column of the FY2024 statements, dated 2025-03-12, fifteen months after the year it describes; and the presentation fact the claim asserts — "presented within loss from discontinued operations" — is a classification made in the FY2024 statements, where comparatives are restated for reclassification, so this document cannot establish how the item was presented as originally reported. The contemporaneous primary is Ballard's FY2023 audited consolidated financial statements, which the ledger does not hold │ evidence: ballard.com/wp-content/uploads/2025/03/Q4-24-Financial-Statements-FINAL.pdf, note 11 and note 26; the $23,991,000 figure appears in the 2023 comparative narrative of a 2024 filing │ evidence_status: RETRIEVED
C-46 │ vector: PROVENANCE │ attack: not independent of C-28 — identical PDF, identical note; two records, one document, and the ledger prints "sources: 1" on each without flagging the overlap │ evidence: shared-sources.txt maps the Ballard PDF to C-28 and C-46 │ evidence_status: RETRIEVED
UNATTACKED
none
CONCESSION: MOVED — I expected the issuer-filing records to fail on existence or wording and they did not. C-07, C-08, C-27, C-28, C-31, C-36, C-41 and C-46 survived verbatim checks against their cited documents, and C-11's numbers match its abstract exactly. The ledger is disciplined at the sentence level, and several records hedge in ways that pre-empt the obvious attack. What it is not disciplined about is the boundary of a document: the recurring failure across C-07, C-15, C-21, C-28 and C-35 is citing a filing for a negative or a date while leaving the recognition sitting in the same filing unbooked — Lightspeed's $748.7M, Ballard's $40,277 thousand, Corus's $295.2M of goodwill — which is what makes C-14's three-issuer population collapse against the ledger's own C-07. Three hard kills stand on retrieval: C-19 miscounts twelve as thirteen on a list that is expressly non-exhaustive, C-45 reverses the authors' own SFAS 142 caveat because it cites the preprint abstract rather than the published paper, and C-26 substitutes "non-zero posterior means" for a posterior z-score threshold of 1.96, turning an 82.4 percent finding into a near-vacuous one. C-14 and C-22 are the structural weak points: both are self-sourced, one openly carries no source at all, and both convert a single run's retrieval into a population fact.
ROUND 3 — 03-attacks/inquisitor-r3.md
INQUISITOR — ROUND 3 (LIVE, cutoff 2026-08-22)
Vector ANACHRONISM disabled for this run; no attack is filed under it.
flags / cutoff-failures / inadmissible-records were all none, so every record is admissible on date
and the whole subset is contestable on substance. shared-sources.txt gave three collisions; two of
them (papers.ssrn.com/sol3/papers.cfm, the Ballard PDF) turn out to be worth attacking for different
reasons than the normaliser implies, and the audit also found **two cross-record dependences the file
misses** (C-15↔C-27 via Lightspeed's Q2 FY23 statements; C-08↔C-22 via Kroll), plus one record that
falsely asserts a dependence that does not exist (C-36 with C-41).
ATTACKS
C-04 │ vector: PROVENANCE │ attack: the record's only source is a Grant Thornton insights article, and that article does not carry the paragraph-level content the claim attributes to it — fetched, it cites no paragraph number at all except ¶96 (and cites that only to call it "an anti-abuse provision", the opposite of the latitude the claim draws from it); ¶9, ¶10, ¶12(d) and ¶99 appear nowhere on the page, so a LOAD-BEARING five-paragraph reading of IAS 36 rests on a secondary web article that is not the standard and does not quote it │ evidence: https://www.grantthornton.global/en/insights/articles/IFRS-ias-36/ifrs-ias-36-If-and-when-to-undertake-an-impairment-review/ (fetched this run; title "IFRS - IAS 36 - If and when to undertake an impairment review", 2021-08-03) │ evidence_status: RETRIEVED
C-04 │ vector: SELECTION │ attack: the record's own falsifier — "an IAS 36 or IAS 34 provision permitting deferral of impairment testing to a later period after an indicator arises" — is satisfied by IAS 36 ¶15, which the ledger never names: "Apart from when the requirements in paragraph 10 apply, the concept of materiality applies in identifying whether the recoverable amount of an asset needs to be estimated… if previous calculations show that an asset's recoverable amount is significantly greater than its carrying amount, the entity need not re-estimate the asset's recoverable amount if no events have occurred that would eliminate that difference. Similarly, previous analysis may show that an asset's recoverable amount is not sensitive to one (or more) of the indications listed in paragraph 12." For any asset or CGU outside ¶10 — which is exactly the "asset or CGU" generality the claim asserts — ¶9 does not require an estimate of recoverable amount once a ¶12 indicator arises; ¶15 relieves it, and a ¶12(d) crossing is one of the very indications ¶15 says prior analysis may show the recoverable amount to be insensitive to │ evidence: IAS 36 ¶15, verbatim, from the consolidated standard text (https://www.icab.org.bd/icabadmin/uploads/ckeditor/7984IAS_36_2017.pdf); absent from the record and from its sole cited source │ evidence_status: RETRIEVED
C-07 │ vector: PROVENANCE │ attack: the impairment decomposition is misread — the record states "$590.0M in the quarter ended May 31 2023, of which $295.2M goodwill, $219.8M broadcast licences and $175.0M brands", but those three components sum to $690.0M, which is the full fiscal 2023 total, not the Q3 charge. Corus's own documents put $590.0M in Q3 (quarter ended May 31 2023) and $100.0M in Q4 (quarter ended August 31 2023), together $690.0M, and it is the year figure that decomposes as goodwill $295.2M / broadcast licences $219.8M / brands and trade marks $175.0M. The record therefore assigns $100.0M of Q4 licence-and-brand impairment to the Q3 quarter, and — since the goodwill line is the piece the ledger actually needs dated — mis-dates the only goodwill component in the record. Q3 was announced as "$590.0 million related to goodwill, broadcast licences as well as trade mark and brand assets" with no itemisation, and Q4 as "$100.0 million related to broadcast license and brands and trade marks" │ evidence: Corus fiscal 2023 Q3 release (newswire.ca/…831318979.html), fiscal 2023 Q4/year-end release (prnewswire.com/…301969777.html) and the fiscal 2023 Annual Report MD&A (https://assets.corusent.com/wp-content/uploads/2023/12/Annual-Report-2023.pdf), all fetched this run │ evidence_status: RETRIEVED
C-08 │ vector: PROVENANCE │ attack: independence failure across records that the shared-sources normaliser cannot see, because the URLs differ while the source does not — C-08 is a trade outlet relaying Kroll's own managing director's forecast about the impairment market Kroll's study franchise sells, and C-22's sole source is Kroll's goodwill-impairment publications hub (kroll.com/en/insights/goodwill-impairment). Two records, one interested party. CFO Dive adds no verification: the piece is built on Nunes and on Kroll's study, so it is one Kroll source counted twice in a ledger that is already thin on independent bases │ evidence: https://www.cfodive.com/news/goodwill-impairment-expected-rise-reverse-course-2022-kroll/621614/ (fetched: byline Maura Webber Sadovi, 2022-04-06, Nunes identified as a Kroll managing director) read against C-22's source line │ evidence_status: RETRIEVED
C-08 │ vector: INFERENCE │ attack: the quoted evidence supports a weaker claim than the record states. Nunes's actual words are "Relative to 2021 I expect impairments to go up — and some of them might be related to impacts from the Russia-Ukraine conflict." The weaker claim the evidence supports: in April 2022 Nunes said she expected impairments to go up in 2022 relative to 2021. The restrictions the record adds — that the expectation was specifically about goodwill impairments and specifically among US public companies — are the headline's framing and the record's, not the quoted statement's; the quote names neither goodwill nor a population │ evidence: verbatim quote from the CFO Dive article, fetched this run │ evidence_status: RETRIEVED
C-11 │ vector: PROVENANCE │ attack: the record's second sentence — "restricted to the 161 clearly significant in the originals, Chen & Zimmermann's independent reproduction (FEDS 2021-037) puts 98 percent above |t| = 1.96" — is not supported by the record's only listed source. The HXZ SSRN abstract page carries the 452 / 65 % figure and nothing else; it does not mention Chen & Zimmermann, 161, or 98 %. FEDS 2021-037 is named in the claim text but never entered as a source, so the record reports "sources: 1" while carrying two papers' findings │ evidence: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3275496 (fetched: confirms "65% of the 452 anomalies in our data library…cannot clear the single test hurdle of the absolute t-value of 1.96"; no Chen–Zimmermann content) vs. https://www.federalreserve.gov/econres/feds/files/2021-037pap.pdf │ evidence_status: RETRIEVED
C-11 │ vector: INFERENCE │ attack: the record pins the construction up front ("as HXZ construct them, with microcaps mitigated via NYSE breakpoints and value-weighted returns") and then presents the 98 % figure as what that same denominator yields when restricted to 161 predictors. It is not. Chen & Zimmermann explicitly do not rebuild HXZ's 452 portfolios and do not use HXZ's construction: "Using the original paper's results, we select the stock-weighting, rebalancing frequency, and quantile sort (if applicable)" — predominantly equal-weighted, original-paper breakpoints, i.e. precisely the microcap-inflated construction HXZ's 65 % is designed to reject. The weaker claim the evidence supports: HXZ, using NYSE breakpoints and value weighting, find 65 % of 452 portfolios below |t| = 1.96; Chen & Zimmermann, using each original paper's own weighting and breakpoints on an independently built data set seeded from 240 HXZ characteristics, find 158 of 161 originally-clear predictors above |t| = 1.96. The two numbers are not the same population measured two ways and the gap is not attributable to the 452→161 restriction │ evidence: FEDS 2021-037 (https://www.federalreserve.gov/econres/feds/files/2021-037pap.pdf), fetched this run │ evidence_status: RETRIEVED
C-15 │ vector: PROVENANCE │ attack: the record's single source cannot carry two of its four sentences. I fetched the Q3 FY2023 earnings release: it states the December 31 2022 trigger ("The Company's net assets exceeded its market capitalization as at December 31, 2022 which was an impairment trigger for the Company"), the December 31 test date, and the ($748.7) million charge — and it contains no reference to September 30 2022, to a prior test, or to a fair-value-less-costs-of-disposal conclusion. The September-30 sentence and the "one quarter" interval are drawn from Lightspeed's Q2 FY2023 interim statements, which are C-27's source and are not listed on C-15. A LOAD-BEARING record whose headline quantity (the interval) rests on a document it does not cite is a one-source record doing two-source work │ evidence: https://www.sec.gov/Archives/edgar/data/1823306/000182330623000008/earningsreleasefy23q3.htm (fetched) vs. https://s28.q4cdn.com/517092977/files/doc_financials/2023/q2/Financial-Statements-FY23-Q2-Final.pdf (fetched) │ evidence_status: RETRIEVED
C-15 │ vector: INFERENCE │ attack: "The documented crossing-to-recognition interval is one quarter" is stronger than the evidence. The Q2 FY2023 note dates the decline, not the crossing, to a six-month window: "During the six months ended September 30, 2022, the Company's share price and therefore its market capitalization decreased. The carrying amount of the Company's net assets exceeded the Company's market capitalization as at September 30, 2022." Lightspeed's Q1 FY2023 interim statements — fetched — disclose no indicator, no test and no market-capitalisation comparison at June 30 2022, while showing net assets of US$3,343,426 thousand over 149,372,883 shares. The weaker claim the evidence supports: the earliest crossing Lightspeed chose to disclose is September 30 2022, and recognition followed one quarter later — an upper bound on issuer disclosure lag, not a measurement of the crossing-to-recognition interval, since the ledger holds nothing that dates the crossing itself │ evidence: https://s28.q4cdn.com/517092977/files/doc_financials/2023/q1/Financial-Statements-FY23-Q1-Final.pdf (fetched; balance sheet and absence of any impairment-indicator note) │ evidence_status: RETRIEVED
C-16 │ vector: SELECTION │ attack: a competent opponent cites TELUS Digital's own year-end 2024 numbers, which the ledger does not hold, and which turn this record from "no impairment, headroom unquantified" into a live ¶12(d) crossing without recognition — the single most load-bearing observation type in this ledger. TELUS Digital's Q4/full-year 2024 results report goodwill of US$1,926 million at December 31 2024, while TELUS International (Cda)'s market capitalisation at December 31 2024 was US$1.08 billion, down 54 % on the year. Goodwill alone exceeded the entire market capitalisation; net assets did so by far more. The parent-level "recoverable amount exceeded carrying value" conclusion the record reports is therefore a no-impairment outcome recorded against a standing market-capitalisation indicator at the same measurement date, and neither the subsidiary's statements nor its market capitalisation appears anywhere in the ledger │ evidence: TELUS Digital Q4/FY2024 results (https://www.businesswire.com/news/home/20250213367125/en/…, goodwill 1,926 vs 1,963) and TIXT market capitalisation at 2024-12-31 (https://stockanalysis.com/stocks/tixt/market-cap/), both fetched this run │ evidence_status: RETRIEVED
C-19 │ vector: SELECTION │ attack: the ledger cites the November 2022 notice, whose impairment content is a bare listing of an area that "may be impacted", and omits its successor — CSA Staff Notice 51-365, Continuous Disclosure Review Program Activities for the fiscal years ended March 31 2024 and March 31 2023, published 7 November 2024 — which is pre-cutoff, is the same regulator on the same programme, and reports actual review findings on impairment rather than a list of headings: that issuers use boilerplate impairment disclosure, fail to "sufficiently explain the events and circumstances that led to an impairment", and merely reference general macroeconomic factors instead of issuer-specific drivers, with tables of IAS 36 requirements and MD&A expectations. That is the regulator document a competent opponent puts on the record for the proposition C-19 is being used to support, and it is absent │ evidence: CSA Staff Notice 51-365 (7 November 2024), fetched at https://fcaa.gov.sk.ca/public/CKeditorUpload/CSA_Staff_Notice_51-365_Continuous_Disclosure_Review_Program_Activities_for_2024_and_2023_November_7,_2024.pdf; also at https://www.osc.ca/sites/default/files/2024-11/csa_20241107_51-365_continuous-disclosure-review.pdf │ evidence_status: RETRIEVED
C-21 │ vector: PROVENANCE │ attack: the record concedes it duplicates C-15's sentence and then commits the same defect a second time — its final sentence quotes the FY2023 audited statements' trigger narrative ("During the three months ended December 31, 2022, there were changes in macroeconomic conditions and the Company's share price and market capitalization decreased…") and the US$748,712 thousand figure, and neither appears in the Q3 earnings release that is its only listed source. I fetched both: the release gives "($748.7) million" and no such narrative; the "748,712" and the quoted sentence are in the FY2023 statements, which the record itself locates at C-35 but does not enter as its own source. So a LOAD-BEARING record that exists only to declare itself weightless is nonetheless mis-sourced on the one sentence it adds │ evidence: https://www.sec.gov/Archives/edgar/data/1823306/000182330623000008/earningsreleasefy23q3.htm and https://s28.q4cdn.com/517092977/files/doc_financials/2023/q4/Financial-Statements-FY23-Q4-Final.pdf, both fetched this run │ evidence_status: RETRIEVED
C-22 │ vector: PROVENANCE │ attack: the sole source is this run's own unlogged retrieval, pinned to a URL (Kroll's goodwill-impairment hub) that is evidence for nothing about datasets it does not host — a null result is not made a source by giving it a link — and the record's one checkable factual assertion is wrong on retrieval. The 2014 Canadian Goodwill Impairment Study states it is the third annual edition, that the inaugural "2012 Study" was launched in February 2013 (so "published in 2012" mis-states the publication year as the edition year), and that both the 2012 and 2013 surveys drew on FEI Canada members — the 2014 survey switched to Mergermarket's database. The record's "(the 2012 jointly with FEI Canada)" therefore drops the 2013 edition's FEI Canada participation. A record whose function is to certify a negative about the literature mis-describes the one publication in it │ evidence: 2014 Canadian Goodwill Impairment Study, fetched at https://edge.sitecorecloud.io/krollllc17bf0-kroll6fee-proda464-0e9b/media/Assets/PDFs/Publications/Valuation/GWI/2014-canadian-goodwill-impairment-study.pdf (redirect target of kroll.com/-/media/assets/pdfs/publications/valuation/gwi/2014-canadian-goodwill-impairment-study.pdf) │ evidence_status: RETRIEVED
C-26 │ vector: PROVENANCE │ attack: neither figure in the claim is in the claim's only source. I fetched the SSRN abstract page dated 2021-03-05: it gives no percentage and no factor count, only "The majority of asset pricing factors: (1) can be replicated… (3) work out of sample in a new large data set covering 93 countries." The 82.4 % is from the published Journal of Finance (2023) version, which is not on the source list; the record is a 2021 working-paper abstract page carrying a 2023 published result │ evidence: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3774514 (fetched, verbatim abstract) vs. Jensen, Kelly & Pedersen, Is There a Replication Crisis in Finance?, Journal of Finance 78(5), 2023 (https://research-api.cbs.dk/ws/portalfiles/portal/95651880/theis_ingerslev_jensen_et_al_is_there_a_replication_crisis_in_finance_publishersversion.pdf) │ evidence_status: RETRIEVED
C-26 │ vector: INFERENCE │ attack: the record binds the 82.4 % to "a data set of 153 factors". The published paper does not: the sentence is "the replication rate rises to 82.4% (the sixth bar of Figure 1) using our Bayesian approach to the MT problem", computed over the US factors with originally significant findings (119 on the fetched text), while 153 is the factor count of the global data set spanning 93 countries — the object of the paper's separate out-of-sample finding, which the record correctly keeps distinct in its next sentence and then silently merges into the denominator of the first. The weaker claim the evidence supports: 82.4 % is the replication rate for the US factors that were significant in their original papers, under the authors' Bayesian multiple-testing adjustment; the 153-factor count belongs to the separate global data set. Since the record's whole point is that the two findings are distinct and unconjoined, importing the global factor count as the replication denominator is the exact error it disclaims │ evidence: JF 2023 published version, fetched (quoted sentence and sample description) │ evidence_status: RETRIEVED
C-27 │ vector: PROVENANCE │ attack: the source is dated 2022-09-30, but that is the balance-sheet date, not a publication date — the statements were authorised for issue on 2 November 2022, 33 days later. In a ledger whose admissibility rule is keyed to source dates (see inadmissible-records.txt), dating a document to the period it reports on rather than to its existence is a systematic mis-dating in the permissive direction. Separately, the independence audit that shared-sources.txt missed: this document is the sole evidentiary basis for the September-30 half of C-15, whose own source list does not contain it — so C-15 and C-27 are one document's finding entered twice, and only C-27 cites it │ evidence: https://s28.q4cdn.com/517092977/files/doc_financials/2023/q2/Financial-Statements-FY23-Q2-Final.pdf (fetched; date of authorisation for issue 2 November 2022) │ evidence_status: RETRIEVED
C-28 │ vector: SELECTION │ attack: the record extracts the indicator sentence from Ballard's note and leaves in the same paragraph the two facts that actually matter to a market-capitalisation-indicator-to-recognition series — that the Corporation "updated its goodwill and non-financial asset impairment tests as of September 30, 2024", and that the 2024 outcome was a US$40,277 thousand goodwill impairment writing corporate goodwill down to nil. Ballard is thus a zero-quarter crossing-to-recognition observation, measured off-cycle at an interim date, and neither the measurement date nor the recognition appears anywhere in this subset. A ledger holding the indicator and discarding the recognition from the same note has selected against its own thesis │ evidence: Ballard FY2024 audited consolidated financial statements, note 11 (https://www.ballard.com/wp-content/uploads/2025/03/Q4-24-Financial-Statements-FINAL.pdf), fetched this run │ evidence_status: RETRIEVED
C-28 │ vector: PROVENANCE │ attack: C-28 and C-46 are the same PDF — the same audited statements, the same note — each presented as an independent record carrying "sources: 1". shared-sources.txt flags the collision; the ledger does not net it out, and unlike C-21 (which discloses its duplication) neither C-28 nor C-46 states that its evidentiary base is the other's. Two records, one document, one audit opinion, one issuer, one date: the ledger's source total overstates by one │ evidence: shared-sources.txt line 2, confirmed against the fetched PDF at https://www.ballard.com/wp-content/uploads/2025/03/Q4-24-Financial-Statements-FINAL.pdf │ evidence_status: RETRIEVED
C-31 │ vector: PROVENANCE │ attack: the source is a syndicated copy of a news release carried by Nasdaq — a distribution channel, not a filing — and it is being used to establish a negative about a financial statement ("does not discuss impairment indicators of any kind"). A press release's silence on indicators is not the issuer's silence: IAS 36 indicator disclosure lives in the interim statements, which the record's own falsifier invokes ("TELUS Digital's Q2 2025 interim statements") and which the ledger does not hold. Fetching the release confirms the amount and the three valuation drivers and confirms the absence — of indicators from the release, which is all a release can show │ evidence: https://www.nasdaq.com/press-release/telus-digital-reports-second-quarter-2025-results-incremental-improvement-revenue (fetched: "As at June 30, 2025, we recorded a non-cash goodwill impairment charge of $224 million"; no trigger language); TELUS Digital Q2 2025 interim consolidated financial statements — named, absent from the ledger │ evidence_status: RETRIEVED
C-35 │ vector: INFERENCE │ attack: "That establishes only that no goodwill impairment loss was recognised before April 1, 2022" is stronger than a comparative column can support, and the error runs backwards through the whole prior period. The FY2023 goodwill note reconciles two years: fiscal 2022 opens at 971,939 (1 April 2021) with "Impairment loss — —" and closes at 2,104,368; fiscal 2023 opens there. Any goodwill impairment recognised in fiscal 2021 or earlier is already netted into the 971,939 opening balance and would not appear in this note at all. The weaker claim the evidence supports: no goodwill impairment loss was recognised in the year ended March 31, 2022. For a record whose stated purpose is to police what a comparative column does and does not establish, over-reading its own comparative by every year before fiscal 2022 is the defect it was written to prevent │ evidence: https://s28.q4cdn.com/517092977/files/doc_financials/2023/q4/Financial-Statements-FY23-Q4-Final.pdf, note 16 goodwill reconciliation, fetched this run │ evidence_status: RETRIEVED
C-36 │ vector: SELECTION │ attack: "The release states no goodwill component and decomposes the charge no further… so in a goodwill-impairment ledger this record supports no goodwill observation" is defeated by a document filed by the same issuer on the same day as the underlying charge, which a competent opponent retrieves in one step. BCE's Q3 2024 consolidated financial statements (SEC 6-K, EX-99.2, 7 November 2024) decompose the $2,113 million: goodwill impairment $1,132 million in Bell Media, indefinite-life intangible assets (broadcast licences and brands) $627 million, programme and feature film rights $144 million, PP&E $85 million, software $85 million, finite-life intangibles $10 million, prepaid expenses $7 million, with the CGUs named as the English and French TV services and radio markets. Goodwill is not merely present, it is the largest single component and 54 % of the charge. The record's conclusion — that the ledger's largest Canadian impairment event of 2024 yields no goodwill observation — is an artefact of citing the press release and stopping │ evidence: BCE Inc. 2024 Third Quarter Financial Statements, https://www.sec.gov/Archives/edgar/data/718940/000071894024000015/a2024-q3xfinancials.htm, fetched this run │ evidence_status: RETRIEVED
C-36 │ vector: PROVENANCE │ attack: the independence audit is run in the wrong direction and produces a false merger. "It is one disclosure event with C-37 and C-41 and is weighted as one" is not true of C-41: C-41 is the Q3 2024 release of 7 November 2024 stating $2,113 million for the quarter ended 30 September 2024; C-36 is the Q4/full-year release of 6 February 2025 stating $2,190 million for the year. Four months, two periods, two amounts, and $77 million of impairment that exists in the second and not in the first. Collapsing them discards a fourth-quarter increment; the ledger has correctly identified C-37 as wire coverage of C-41 and then wrongly extended the same treatment to a genuinely separate filing │ evidence: both releases fetched this run — https://www.prnewswire.com/news-releases/bce-reports-third-quarter-2024-results-302298110.html ("$2,113 million", quarter ended 30 September 2024) and https://www.prnewswire.com/news-releases/bce-reports-2024-q4-and-full-year-results-announces-2025-financial-targets-302369604.html ("$2,190 million", full-year 2024) │ evidence_status: RETRIEVED
C-37 │ vector: PROVENANCE │ attack: the record's closing sentence — "no goodwill component is stated anywhere in the reporting" — is false, and it is false about a document filed the same day as the article it describes. BCE's Q3 2024 consolidated financial statements state a $1,132 million goodwill impairment within Bell Media as part of the $2,113 million. The record correctly identifies the Canadian Press piece as non-independent wire coverage and then, on the strength of that same wire piece, asserts a universal negative over the issuer's entire disclosure record. The scope of the claim exceeds the reach of the source by the whole distance between a newswire summary and a filed financial statement │ evidence: BCE Inc. 2024 Third Quarter Financial Statements, https://www.sec.gov/Archives/edgar/data/718940/000071894024000015/a2024-q3xfinancials.htm, fetched this run │ evidence_status: RETRIEVED
C-41 │ vector: SELECTION │ attack: the record's own falsifier points at "BCE's Q3 2024 interim financial statements", and those statements are absent from the ledger while being free, indexed on EDGAR, and filed under the same accession number as the news release the record does cite (0000718940-24-000015, EX-99.2 alongside the release's EX-99.4). They resolve every gap the record declares unresolvable: the "unquantified remainder outside those properties" is $155 million of other impairments, mainly right-of-use assets for office space; the "no allocation" is a seven-line allocation; and the goodwill component the record says is unstated is $1,132 million. "The record supports no goodwill observation" is a statement about the ledger's retrieval, not about BCE │ evidence: BCE Inc. 2024 Third Quarter Financial Statements, EX-99.2 to the 6-K of 7 November 2024, https://www.sec.gov/Archives/edgar/data/718940/000071894024000015/a2024-q3xfinancials.htm, fetched this run │ evidence_status: RETRIEVED
C-45 │ vector: PROVENANCE │ attack: the SSRN abstract page supports the lag figures and nothing else the record builds on them. Fetched verbatim, the abstract gives "goodwill write-offs lag behind the economic impairment of goodwill by an average of three to four years" and "For a third of the companies examined, the delay can extend up to ten years" — and says nothing about the sample being "composed mainly of goodwill generated before SFAS 142", nothing about "sensitivity tests on a smaller sample of write-offs made upon adoption", and nothing from which the authors' inference of post-142 generalisability could be read. Those are body-of-paper characterisations attributed to an abstract page, on a LOAD-BEARING record with one source; the qualifications that make the record look careful are the parts its source cannot carry │ evidence: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=850705 (fetched, full abstract verbatim) │ evidence_status: RETRIEVED
C-45 │ vector: SELECTION │ attack: the record names the gap itself — "not transportable to a one-step IAS 36 CGU test… without a bridging record this ledger does not hold" — and the bridging record exists, is pre-cutoff, is the standard citation, and is not in the ledger: Li & Sloan, "Has goodwill accounting gone bad?", Review of Accounting Studies 22(2), 2017, 964–1003, the post-SFAS-142 study of impairment timeliness under an impairment-only regime, which is what Hayn & Hughes's pre-142 sample cannot speak to. A competent opponent cites it precisely because it tests the untimeliness proposition in the regime the ledger cares about rather than the one Hayn & Hughes sampled. Declaring a bridging record absent is not the same as establishing that none exists, and here one does │ evidence: Li & Sloan (2017), https://link.springer.com/article/10.1007/s11142-017-9401-7 (also SSRN 1466271); existence and venue confirmed this run │ evidence_status: RETRIEVED
C-46 │ vector: SELECTION │ attack: the record manufactures an unresolvable uncertainty — "cannot establish how the item was presented as originally reported; Ballard's contemporaneous FY2023 audited statements are not in the ledger" — and the resolving document is public, free, and one retrieval away. Ballard's 2023 Annual Report states in MD&A §5.2: "During the fourth quarter of 2023, we completed a restructuring of operations at Ballard Motive Solutions in the U.K. and effectively closed the operation. As such, the historic operating results…of the Ballard Motive Solutions business for both 2023 and 2022 have been removed from continuing operating results and are instead presented separately in the statement of comprehensive income (loss) as loss from discontinued operations." The discontinued-operations presentation was adopted in Q4 2023 itself, as originally reported — not created by FY2024 restatement. The record's central caveat is not a limit on the evidence; it is a limit on the retrieval, and a competent opponent removes it │ evidence: Ballard Power Systems 2023 Annual Report, https://www.ballard.com/wp-content/uploads/2024/11/Ballard-2023-Annual-Report-1.pdf, fetched this run │ evidence_status: RETRIEVED
UNATTACKED
None. Every record in the subset is attacked above.
CONCESSION: UNMOVED — the narrowing that let this subset survive round two was applied to claim text, not to source lists, and the two came apart. Records were rewritten until they said only what a careful reader would concede, and then left pointing at whatever source they had entered before the rewrite: C-11 and C-26 and C-45 now state findings their SSRN abstract pages have never contained, C-15 and C-21 state findings their earnings release does not contain, and C-04 states five paragraph numbers its consultancy article never prints. Meanwhile the hedges themselves became the attack surface — C-46, C-37, C-41 and C-36 each declare a limit on the evidence that is in fact a limit on the retrieval, and each is dissolved by a document filed by the same issuer, often on the same day, most decisively BCE's Q3 2024 statements, which put $1,132 million of goodwill inside a charge three records agree supports no goodwill observation. C-07 fails on arithmetic that survived a round of attack unchecked: three components summing to the year total, presented as the composition of the quarter. What a narrowing left behind here is a ledger that reasons well about what its sources would show if anyone opened them.
NULL OUTPUT
ROUND 1 — 03-attacks/null.md
THE NULL — attack on the stripped ledger
Run parameters. Evidence cutoff 2026-08-22. Mode LIVE. Inputs: stripped.md (48 records),
flags.txt (empty — "none"), cutoff-failures.txt (empty — "none"), shared-sources.txt,
inadmissible-records.txt (C-48, SUPPORTING, dates —).
Position. The documentary substrate of this ledger is sound and I concede almost all of it.
What the ledger does not contain is a measurement of the thing being asked about. Every record
that would bridge the documents to the estimand is an E2 inference or an E3 model prior, and each
of those bridges is either (a) a unit change performed without a conversion, (b) an inference whose
own stated falsifier was never run, or (c) a claim about the shape of questions rather than about
Canadian issuers. I contest 10 records of 48 and one joint inference. I concede the other 38.
0. What the ledger actually is, once duplicates are collapsed
shared-sources.txt is decisive here, and it understates the problem. Collapsing by document:
| Source cluster | Records | Distinct documents |
|---|---|---|
| Grant Thornton IAS 36 web article | C-04, C-05, C-13, C-20, C-33, C-43 | 1 web page |
| Lightspeed Q3 FY23 SEC earnings release | C-06, C-12, C-15, C-21, C-23, C-39, C-40 | 1 document |
| SSRN abstract pages | C-10, C-11, C-24, C-26, C-30, C-45 | 4 abstracts, 0 full texts |
| Corus FY23 annual MD&A PDF | C-03, C-07, C-25, C-34, C-47 | 1 document |
| Lightspeed Q2 FY23 statements | C-18, C-27, C-38 | 1 document |
| Ballard Q4-24 statements | C-28, C-44, C-46 | 1 document |
| ESMA/2013/2 | C-02, C-09, C-29 | 1 document |
| Lightspeed Q4 FY23 statements | C-35, C-42 | 1 document |
| Model priors (no source at all) | C-14, C-17, C-48 | 0 |
Forty-eight records reduce to roughly nineteen documents, three of which do not exist. Worse,
four of those documents are the same issuer's filings about the same single impairment event
(Lightspeed, US$748.7M, Q3 FY2023). Counted as evidentiary facts rather than as records:
- One Lightspeed impairment appears as C-06, C-23 and C-42 — three records, one charge.
- One Lightspeed sentence appears as C-15, C-21 and C-32 — three records, one sentence.
- One Corus disclosure appears as C-03 and C-07 — two records, one disclosure list.
- One Corus charge appears as C-25 and C-47 — two records, one charge.
- One Grant Thornton bullet appears as C-05 and C-20 — two records, one bullet.
- One BCE Q3 2024 charge appears as C-36, C-37 and C-41 — three records, one event, and C-37 is
wire coverage of C-41's release, so the "independent journalism" corroboration is the press
release again.
This is the "several sources turn out to be one" pattern the brief names, and it recurs six times.
The apparent breadth of the ledger is a citation artifact.
Issuer coverage is five. Lightspeed, Corus, Ballard, BCE, TELUS/TELUS Digital. Of those five,
exactly two (Lightspeed, Corus) carry any claim about a crossing date, and both of those
crossing dates are self-disclosed, not measured. The estimand is a median over a population
that C-22 itself says has never been enumerated. A median with n=2 measured pairs, both measured
by the subject, is not thin — it is not a median.
1. Ill-posedness — the terms that equivocate
1.1 "Market capitalisation" ≠ "market enterprise value" (named term, load-bearing)
IAS 36's external indicator, as the Grant Thornton page states verbatim, is *"Carrying amount of
the net assets of the entity is more than its market capitalisation."* C-05, C-20, C-15, C-21,
C-32 and C-22 all use market capitalisation.
Corus does not. C-03, C-07 and C-34 all use market enterprise value. Corus's own fiscal 2022
year-end release confirms the enterprise-value framing: *"a significant decline in the Company's
share price from August 31, 2022 … resulted in the Company's carrying value being greater than its
current market enterprise value."*
These two measures differ by net debt. **Corus's long-term debt at August 31, 2022 was
C$1,246.1 million.** The enterprise-value test therefore sits roughly C$1.25 billion further out
than the market-capitalisation test — which is precisely why Corus could report the enterprise
crossing as new in Q4 FY2022 while the market-capitalisation crossing had already happened long
before. The ledger splices the two measures into one clock and never converts between them.
1.2 "Economic impairment" carries three incompatible definitions
- Hayn & Hughes (C-45, C-30): a fundamentals-based construct internal to their US study.
- André, Filip & Paugam (C-10, C-24): three proxies — market-to-book < 1; equity market value
minus book value < goodwill; negative EBITDA. Verified against the abstract.
- IAS 36 / the issuers (C-05, C-15, C-32): carrying amount of net assets > market capitalisation.
"Three to four years" (C-45) is measured in Hayn & Hughes's units. C-02 and the estimand are in
IAS 36 units. Moving a duration from one construct to the other is a unit change; the ledger
performs it without a conversion factor and without a record that even attempts one.
1.3 "Net assets" ≠ "equity book value" ≠ "carrying value"
ESMA (C-09, C-29) compares market capitalisation to equity. Grant Thornton and IAS 36 say
net assets of the entity. Corus says carrying value — of the entity, or of the Television
CGU? These diverge in the presence of non-controlling interests and diverge completely when the
comparison is struck at CGU level rather than entity level. Nothing in the ledger fixes the level.
1.4 "First goodwill impairment" is not well defined
Ballard has a 2023 charge (C-46, US$23,991k, Ballard Motive Solutions) and a 2024 charge (C-44,
US$40,277k). Which is "first" depends entirely on the window boundary and on whether a subsidiary-
level CGU charge counts. C-14's own falsifier scopes the population as "2022 to 2025," so Ballard's
"first" is a knife-edge of the window, not a fact about Ballard.
1.5 The smuggled premise: that a positive lag is delay
The whole framing presumes the correct recognition date is the indicator date. **The ledger's own
C-13 denies this**, and I verified it verbatim on the source page: identifying an indicator
*"triggers the need to estimate recoverable amount, which then determines whether an actual loss
exists."* An indicator followed by no charge is not tardiness; it is the standard operating
correctly. See §3, Rival 1 — this is the load-bearing hole.
2. Falsifier hygiene — six falsifiers that cannot do their job
| Record | Defect in the stated falsifier |
|---|---|
| C-34 | Claim asserts market enterprise value; falsifier tests market capitalisation. A falsifier that measures a different quantity than the claim asserts cannot discriminate — and here the two differ by C$1.25bn of net debt. |
| C-12 / C-40 | Falsifiers specify arithmetic (share count × close vs. audited net assets at earlier quarter-ends) that the ledger never performs. |
| C-14 | Falsifier requires enumerating the bounded population — which C-22 asserts nobody has enumerated. The pair is self-blocking. |
| C-17 | Falsifier requires "a tabulation of resolved median-threshold claims" that the ledger neither contains nor cites nor names a location for. |
| C-22 | A search-exhaustion negative. Refutable in principle, confirmable never; sourced to one Kroll landing page. |
| C-04 / C-43 | Each one's falsifier is the other one's claim. They cannot both stand. See §4.1. |
3. Live rivals — named
I win a claim only where nothing in the ledger discriminates between the leading explanation
("Canadian issuers delay recognition well past the indicator quarter, so the median lag is large")
and a named rival. Five rivals, each with ledger-internal or verified support.
RIVAL 1 — Recoverable amount lawfully exceeded carrying amount (strongest)
Under IAS 36, recoverable amount is the higher of value in use and fair value less costs of
disposal. Both can lawfully exceed a depressed market capitalisation: market prices embed leverage,
liquidity, control discounts and market-wide risk premia that a CGU's recoverable amount does not.
Verified instance, from a ledger source I retrieved directly. TELUS Corporation's 2024 audited
statements state, as a critical audit matter: *"Management determined that the recoverable amount
of TELUS digital experience exceeded its carrying value as of the measurement date and, therefore,
no impairment was recognized."* This is exactly C-16 — and it means the parent recognised nothing
for that CGU in 2024 while the underlying business's own listed market value had collapsed.
Second instance, inside the ledger. C-27: Lightspeed performed a goodwill impairment test as at
September 30, 2022 and recognised no impairment. Test run, standard followed, nothing recognised.
Under Rival 1, the ESMA 47% (C-29) and the André 20–25% (C-10) are not measures of delay at all —
they are the fraction of indicator-firms for which VIU or FVLCD genuinely came in above carrying
amount. The ledger contains not one recoverable-amount figure for any non-impairing firm. That
is the single quantity that would discriminate delay from correct accounting, and it is absent from
all 48 records. Nothing in the ledger separates these two explanations.
RIVAL 2 — Disclosure-date censoring (the zero-lag artifact)
Both Canadian lags in the ledger are disclosed lags, and both are exactly zero (C-12, C-40, C-34).
That is what you would expect under either hypothesis: an issuer discloses the crossing in the
filing that records the charge, because that is the filing in which the crossing becomes material to
disclose. A disclosed lag of zero is therefore uninformative about the true lag by construction. The
ledger has no record that establishes a crossing date from exchange data rather than from the
issuer's own narrative — and C-12's and C-40's falsifiers say so in as many words.
And the arithmetic points the wrong way for the leading account of Lightspeed. Using only
ledger figures plus one external datapoint:
- Net assets at September 30, 2022: US$3,304.4M (C-38).
- Net assets at December 31, 2022: US$2,539.5M (C-39); adding back the US$748.7M charge (C-06,
C-23, C-42) gives pre-impairment net assets at December 31 of ≈US$3,288M — essentially flat
against September 30. The entire quarter's equity movement is the impairment itself.
- Lightspeed market capitalisation at the end of 2022: ≈US$2.15–2.16bn, corroborated by two
independent aggregators.
For the crossing to have first occurred at December 31, 2022, Lightspeed's market capitalisation
must have stood at or above US$3.30bn on September 30 and then fallen ≈35% in a single quarter
to US$2.15bn. I could not retrieve a pre-cutoff quarter-end closing price series to settle this
(three price sources returned 403 / robots-disallowed / no 2022 coverage), so I do not assert the
falsifier fires. I assert what matters: **the ledger's own stated falsifier for two LOAD-BEARING
records was never run, the available arithmetic runs against those records, and C-27 shows
Lightspeed already testing goodwill a quarter earlier.** Undetermined, and leaning adverse.
For Corus the same test runs further. Corus shareholders' equity was **C$1,067.5M at
August 31, 2021; Corus's market capitalisation was C$991.8M at December 30, 2021** and
C$430.8M at December 30, 2022. On the market-capitalisation test named in C-34's own falsifier,
Corus's carrying amount of net assets exceeded its market capitalisation **at least three quarters
before August 31, 2022**, and plausibly a year before. C-34's asserted "match" survives only because
it silently switches to enterprise value. I flag the residual imprecision honestly: my equity and
market-cap datapoints are at adjacent rather than identical dates. That imprecision is the ledger's
to close, not mine — it declined to.
RIVAL 3 — Regime and jurisdiction non-transfer
Every long-lag record is foreign and old. Hayn & Hughes: US, FAS 142 framing, posted 2005. ESMA:
European, 2011 sovereign-debt-crisis balance sheets, when market capitalisations fell for
macro-financial reasons largely disconnected from CGU cash flows. André, Filip & Paugam: Europe and
the US, 2006–2015, 35,000+ firm-years — window closes seven years before the estimand opens.
The estimand is Canadian IFRS, 2022–2025. The ledger's own contemporaneous records point the
opposite way: CSA Staff Notice 51-364 naming impairment as an area of focus (C-19); IOSCO pressing
on "too little, too late" write-downs (C-01); Kroll expecting 2022 impairments to rise (C-08); and a
Big-Four critical-audit-matter treatment of a single CGU test (C-16). **Rival: enforcement
intensification and the 2022 rate shock compressed the lag toward zero.** The two Canadian lags
actually recorded in this ledger — Lightspeed and Corus — are both zero, which is exactly what this
rival predicts and exactly what the long-lag literature does not. Nothing in the ledger adjudicates
between "the literature transfers" and "the literature does not transfer."
RIVAL 4 — Selection and censoring on the outcome
C-14's falsifier scopes the population as "first-time impairers over 2022 to 2025." That
conditions on recognition having occurred. Two biases follow and run in opposite directions:
- Right-truncation: firms that crossed and never impaired are excluded by construction. These
are the long-lag (or infinite-lag) observations. Their exclusion pulls the median down.
- Left-censoring: firms whose crossing predates 2022 enter the window with an unobserved head
start. Their inclusion pulls the measured lag down too, or forces an arbitrary truncation.
The median may not exist at all. If more than half of indicator-firms never recognise — and
ESMA's 53% non-recognition (C-29) and André's 75–80% non-recognition (C-10) are consistent with
exactly that — the survival function never crosses 0.5 and the median is undefined. The ledger
contains no hazard function, no survival analysis, and no record of what happened to any
non-recogniser in any subsequent period.
RIVAL 5 — Reporting-entity multiplicity
C-16 and C-31 describe the same underlying business through two IFRS filers in one corporate group.
TELUS Corporation recognised no goodwill impairment for the TELUS digital experience CGU in 2024
(C-16, verified verbatim). TELUS Digital recognised US$224 million in Q2 2025 (C-31). Both are
true. The consequence is that "the issuer's impairment date" for this business is entity-relative:
one date at the subsidiary, a different date (or none) at the parent, because IAS 36 permits testing
at the level at which goodwill is monitored. Any median over "issuers" must first decide which filer
counts, and the ledger's own two TELUS records answer differently. This is not a defect in either
record; it is a defect in the estimand.
4. The specific bridges that fail
4.1 C-04 and C-43 falsify each other, from one web page
Both are E2 inferences from the same Grant Thornton article. I retrieved that article. It says
both of the following:
- *"IAS 36 requires an entity to assess at the end of each reporting period whether there is any
indication that an asset or CGU may be impaired."*
- *"The annual impairment test for an asset may be performed anytime during the annual period
provided the test is performed at the same time every year."*
C-43 reads the second sentence and concludes IAS 36 "mandates only annual goodwill testing at a
fixed date." The first sentence refutes that: C-43 is falsified by its own source.
C-04 reads the first sentence and concludes that "a lag beyond one quarter [is] a departure from
required timing." That over-reads too, and C-13 — from the same page — is the refutation: the
indicator triggers estimation of recoverable amount, not recognition of a loss. An issuer that
assesses indicators every quarter, estimates recoverable amount, and books nothing has departed from
nothing. C-04 equivocates between assessment timing (quarterly, mandatory) and *recognition
timing* (contingent on measurement).
The ledger's normative baseline — the standard against which "delay" is scored — is therefore
unresolved inside the ledger, and both candidate readings trace to one web page, neither to the
text of IAS 36 itself. No record in this ledger cites IAS 36 paragraphs 9, 10, 12, 96 or 99, or
IAS 34, directly. C-20's falsifier even invokes "IAS 36 paragraph 12" — a provision the ledger
never sources.
4.2 C-02 — an annual non-recognition rate cannot bound a quarterly median from below
This is the ledger's central bridge and it fails arithmetically.
C-02 reasons: most indicator-firms do not impair within the indicator year, therefore the median
indicator-to-recognition lag exceeds four quarters. Three independent defects:
1. The frame error. "Not within the indicator year" does not imply "more than four quarters." If
the indicator arises in Q4 of the indicator year and the charge lands in Q1 of the next year, the
lag is one quarter while the firm counts as a non-recogniser. Annual-frame non-recognition
places no lower bound whatsoever on a quarterly median. The inference is a non-sequitur, not a
weak inference.
2. Neither source reports what is claimed. I retrieved both. ESMA/2013/2 gives a cross-sectional
2011 snapshot — 235 issuers, 43% below book, 47% of those recognising — and remarks that the
2010→2011 rise in the equity/market-cap ratio "does not seem to be fully reflected in the level of
goodwill impairment observed in 2011." André, Filip & Paugam give a recognition rate and note
the rate "has often been interpreted by academics as a sign of untimely write-offs." Neither
reports a recognition-timing distribution. Neither reports a lag in quarters. Neither reports a
median. C-02 manufactures a duration statistic from two prevalence statistics.
3. The median may be undefined (Rival 4). A 20–25% same-period recognition rate is equally
consistent with a median of one quarter, of twelve quarters, and with no median at all, depending
entirely on a post-period-0 hazard shape that no cited source estimates.
C-02's falsifier — "firms not impairing in the indicator year are shown to impair predominantly
within the following two quarters" — is unanswerable from either cited source, in either direction.
4.3 C-22 — a negative existential proved by one run's retrieval
C-22 asserts that no published dataset anywhere enumerates the required pairing. Its evidence is
"inference from retrieval … in this run," pointing at a Kroll landing page. That is a report of one
search, not a fact about the literature. It is also load-bearing in a peculiar way: it is the record
that licenses building the estimand by hand from five issuers, and simultaneously the record that
makes C-14's falsifier unrunnable and C-17's "reference class" undefined. I do not claim a dataset
exists. I claim the ledger has not established that none does, and that a single run's retrieval —
with flags.txt and cutoff-failures.txt both empty, so no negative-result diagnostics were even
logged — cannot carry an unbounded negative.
4.4 C-17 — a LOAD-BEARING record with no external source, about the wrong subject
C-17 is E3, "model prior, no external source," and it is marked LOAD-BEARING. It is not a claim
about goodwill, about Canada, or about IAS 36. It is a claim about **how threshold questions tend to
resolve** — that thresholds set far below a reference class central estimate resolve true more often
than thresholds set near it.
Two objections. First, the smuggled premise: C-17 presupposes a "reference class central
estimate" for this estimand. C-22 asserts no such reference class has ever been assembled, and §1
shows the candidate reference classes (Hayn & Hughes, ESMA, André) are measured in three mutually
non-convertible units. There is no central estimate for a threshold to be placed relative to.
Second, and more seriously: using C-17 as load-bearing means part of the verdict is derived from the
form of the question rather than from the accounting record. That is a heuristic about question
construction. It is not evidence about Canadian issuers, and it cannot discriminate between any of
Rivals 1–5.
4.5 C-14 — self-blocking, and undermines the ledger it appears in
C-14 (E3, no source) says small-sample medians move on single-issuer reclassification. I accept the
statistical point entirely — and it is fatal to the ledger that contains it. §1.4 shows the
reclassification is not hypothetical: Ballard's "first" impairment flips on a window boundary, and
§3 Rival 5 shows TELUS flips on the choice of filer. Two of five issuers are reclassifiable on
definitional grounds alone. But C-14 cannot be tested, because its falsifier requires the
population count that C-22 says does not exist. It is an unsourced prior asserting that the ledger's
statistic is unstable, which no record can confirm or refute.
4.6 C-48 — inadmissible and unsourced
inadmissible-records.txt lists C-48. It is E3, "model prior, no external source retrieved in this
run," with source "none" and no date. Its falsifier — TSX close × shares outstanding vs. BCE
consolidated equity at 2024 quarter-ends — is again arithmetic nobody performed. It should carry no
weight; I contest it for completeness.
4.7 The imported replication calibration cancels itself
C-11 (65% of 452 anomalies fail t≥1.96) and C-26 (a majority of factors replicate) are both in the
ledger, both SUPPORTING, both about asset-pricing factor replication. They point in opposite
directions and neither concerns accounting research. Whatever they were imported to do —
calibrate how much to trust C-45, C-10 and C-30 — they cancel, and they would not transfer to
accounting-archival work in any case. The ledger's reliability discount on its own academic inputs
nets to zero.
CONTESTED
C-02 — LOAD-BEARING, E2. An annual non-recognition rate cannot bound a quarterly median from
below (a Q4 indicator and a Q1 charge is a one-quarter lag counted as non-recognition). Neither
ESMA/2013/2 nor André/Filip/Paugam reports any recognition-timing distribution, in quarters or
otherwise — I retrieved both and confirmed this. *Rival: the median is undefined because the
survival function of indicator-firms never crosses 0.5 (Rival 4).*
C-04 — LOAD-BEARING, E2. Equivocates between mandatory quarterly indicator assessment and
loss recognition. Refuted by C-13, from the same web page: an indicator triggers estimation of
recoverable amount, not a charge. "Departure from required timing" is unsupported. *Rival: quarterly
assessment was performed and correctly produced nothing (Rival 1) — as C-27 and C-16 both document.*
C-12 — LOAD-BEARING, E2. "Zero quarters" is a disclosed lag, not a measured one. Its own
falsifier — September 30, 2022 net assets vs. exchange-sourced market capitalisation — was never
run. Ledger figures plus retrieved market-cap data require a ~35% single-quarter collapse in
Lightspeed's market capitalisation for the record to survive; and C-27 shows a goodwill test already
being performed at September 30, 2022. Rival: disclosure-date censoring (Rival 2).
C-14 — SUPPORTING, E3, no source. Untestable: its falsifier needs a population count that C-22
declares nonexistent. Accepted as statistics, rejected as a ledger record — and it undermines the
very statistic the ledger is building. Rival: none needed; the record cannot be evaluated.
C-17 — LOAD-BEARING, E3, no source. Smuggles the premise that a "reference class central
estimate" exists; C-22 denies it and §1.2 shows the candidate classes are measured in
non-convertible units. Substitutes a heuristic about question form for evidence about issuers, and
discriminates between none of the named rivals. Weakest load-bearing member in the ledger.
C-22 — LOAD-BEARING, E2. An unbounded negative existential resting on one run's retrieval and a
single Kroll landing page, with cutoff-failures.txt and flags.txt both empty, so no
negative-result diagnostics were logged at all. *Rival: the pairing exists in a commercial database
extract or a non-indexed regulator annex that this run did not reach.*
C-34 — LOAD-BEARING, E2. Two independent defects. (i) The claim asserts *market enterprise
value* while its falsifier tests market capitalisation — quantities separated by Corus's
C$1,246.1M of long-term debt; a falsifier measuring the wrong quantity cannot discriminate.
(ii) "Earliest disclosed" is treated as "earliest occurring," but the FY2023 MD&A never
claims to be a census of Corus's history. Retrieved data run against the record: equity of
C$1,067.5M at August 31, 2021 against market capitalisation of C$991.8M at December 30, 2021 puts
the market-capitalisation crossing at least three quarters earlier than August 31, 2022.
C-40 — LOAD-BEARING, E2. Same defect as C-12 and same rival. The "match" between trigger date
and recognition date is a match between two dates disclosed in the same filing, which is what
Rival 2 predicts regardless of the true crossing.
C-43 — SUPPORTING, E2. Falsified by its own source. The Grant Thornton page states IAS 36
requires assessment "at the end of each reporting period"; C-43 asserts IAS 36 mandates "only annual
goodwill testing at a fixed date." Cannot stand alongside C-04, which it contradicts and which
contradicts it.
C-48 — SUPPORTING, E3, no source, no date, and named in inadmissible-records.txt. Its
falsifier is arithmetic nobody performed.
C-16 × C-31 (joint inference) — Not a challenge to either record; both are true as written and
both are conceded individually. The contest is to any use of them as one issuer's impairment
history. TELUS Corporation recognised nothing for the TELUS digital experience CGU in 2024 while
TELUS Digital recognised US$224M in Q2 2025 for the same underlying business. "The issuer's first
impairment" is entity-relative and CGU-relative, and the ledger's own two records answer differently
(Rival 5).
**Contested: 10 records of 48, plus one joint inference. Every one is an E2 inference or an E3
model prior. No E0 or E1 primary record is contested.**
NOT CONTESTED
NOT CONTESTED — C-01 — A claim about what a Reuters wire item reported, at E1, doing no work on
the lag estimand in either direction. Attribution claims of this shape are settled by the item's own
text.
NOT CONTESTED — C-03 — Corus's fiscal 2023 MD&A does disclose August 31, 2022 among the dates at
which carrying value exceeded market enterprise value; the FY2022 year-end release I retrieved
independently uses the same enterprise-value language for the same date. Settled *as a statement of
what Corus disclosed*. Only C-34's inference from it is contested.
NOT CONTESTED — C-05 — Retrieved and verified verbatim: the page lists *"Carrying amount of the
net assets of the entity is more than its market capitalisation"* as an external indicator. Settled.
NOT CONTESTED — C-06 — The US$748.7M charge in the quarter ended December 31, 2022 is
corroborated three ways: the SEC release, the FY2023 audited statements, and the ledger's own equity
roll-forward (US$3,304.4M → US$2,539.5M, a movement of US$764.9M dominated by the charge). Settled.
NOT CONTESTED — C-07 — Same basis as C-03. The disclosed set of period-ends is settled; what is
contested is reading "disclosed" as "exhaustive."
NOT CONTESTED — C-08 — A dated, attributed forecast quotation at E1. Settled as reporting — and
it supports Rival 3, not the leading account.
NOT CONTESTED — C-09 — Retrieved and verified: 235 entities across 23 jurisdictions,
€794.9bn of goodwill, *"As of 31 December 2011, 43% of the sample showed a market capitalisation
below equity."* Exact match. Settled.
NOT CONTESTED — C-10 — Retrieved and verified: *"Only 20 to 25% of firms recognize impairments
depending on the measure of economic impairment."* Settled as a report of the paper's finding.
NOT CONTESTED — C-11 — Accurately reports Hou/Xue/Zhang's headline. Settled as to content; §4.7
concerns only its use.
NOT CONTESTED — C-13 — Retrieved and verified: identifying an indicator requires estimating
recoverable amount, which "then determines whether an actual loss exists." Settled — and it is the
Null's central exhibit, not the ledger's.
NOT CONTESTED — C-15 — Settled as to what Lightspeed's release states. The dispute is entirely
about whether a disclosed trigger date is the true crossing date (C-12, C-40).
NOT CONTESTED — C-16 — Retrieved and verified verbatim from the Form 40-F exhibit 99.4: *"the
recoverable amount of TELUS digital experience exceeded its carrying value as of the measurement
date and, therefore, no impairment was recognized."* Settled, and it establishes Rival 1 as fact
rather than possibility.
NOT CONTESTED — C-18 — 150,315,764 subordinate voting shares at September 30, 2022 is a share
capital note figure with a tight falsifier. Settled as to subordinate voting shares — noting only
that it is not a total share count and so cannot on its own support a market-capitalisation
computation, which is why C-12's falsifier remained unrun.
NOT CONTESTED — C-19 — CSA Staff Notice 51-364's list is a matter of public record. Settled.
NOT CONTESTED — C-20 — Verified verbatim; a near-duplicate of C-05. Settled as to the guidance
page. (Its falsifier's reference to IAS 36 ¶12 goes to a text the ledger never sources — a
citation-hygiene note, not a challenge to the claim.)
NOT CONTESTED — C-21 — Same basis as C-15. What Lightspeed stated is settled.
NOT CONTESTED — C-23 — Same event as C-06, from the audited-statement side. Settled.
NOT CONTESTED — C-24 — Retrieved and verified: the three proxies are equity market value minus
book value less than goodwill; market-to-book below one; negative EBITDA. Exact match. Settled.
NOT CONTESTED — C-25 — Independently corroborated outside the ledger's single Corus PDF: the
fiscal 2022 Q4 release states a non-cash goodwill impairment charge of $350.0 million in the
Television segment. Settled with genuine source independence.
NOT CONTESTED — C-26 — Accurately reports Jensen/Kelly/Pedersen. Settled as to content.
NOT CONTESTED — C-27 — Settled, and it is affirmative evidence against C-12 and C-40: a
goodwill test was performed as at September 30, 2022 and produced no impairment, one quarter before
recognition.
NOT CONTESTED — C-28 — An E0 audited-statement disclosure with a tight falsifier; no retrieval
failure was logged against it. Settled.
NOT CONTESTED — C-29 — Retrieved and verified: 47% of below-book issuers recognised goodwill
impairment losses in 2011. Settled as a figure. What it means is Rival 1's territory, not this
record's.
NOT CONTESTED — C-30 — Retrieved and verified verbatim from the abstract: *"For a third of the
companies examined, the delay can extend up to ten years."* Settled as a report of the paper.
NOT CONTESTED — C-31 — A dated, attributed results release carried by multiple wires. Settled as
reporting.
NOT CONTESTED — C-32 — Settled as to the Q3 FY23 note's language.
NOT CONTESTED — C-33 — Retrieved and verified verbatim: *"IAS 36 requires an entity to assess at
the end of each reporting period whether there is any indication that an asset or CGU may be
impaired."* Settled — and this is the record that falsifies C-43.
NOT CONTESTED — C-35 — A comparative-column fact with a tight falsifier. Settled.
NOT CONTESTED — C-36 — A dated results release stating a full-year figure and its attribution.
Settled, and consistent with C-41's Q3 figure.
NOT CONTESTED — C-37 — Accurate wire coverage. Settled as reporting, noting only that it
corroborates C-41 by restating it rather than by independent verification.
NOT CONTESTED — C-38 — US$3,304,419 thousand at September 30, 2022. Settled — and it is the
input to the arithmetic that C-12's falsifier calls for and the ledger never performed.
NOT CONTESTED — C-39 — The two equity figures are settled, and their difference against C-38
confirms the charge's magnitude independently.
NOT CONTESTED — C-41 — A dated results release with a segment attribution. Settled.
NOT CONTESTED — C-42 — Same event as C-06 and C-23, from the FY2023 audited statements. Settled.
NOT CONTESTED — C-44 — An E0 goodwill-note figure and measurement date with a tight falsifier.
Settled.
NOT CONTESTED — C-45 — Retrieved and verified verbatim: *"goodwill write-offs lag behind the
economic impairment of goodwill by an average of three to four years." Settled *as a report of
what Hayn & Hughes found, in Hayn & Hughes's units, for their US sample**. §1.2 and Rival 3 concern
transporting it, not stating it.
NOT CONTESTED — C-46 — An E0 comparative goodwill reconciliation figure. Settled — and it is
what makes "Ballard's first impairment" ambiguous, which is a defect in the estimand, not in this
record.
NOT CONTESTED — C-47 — Independently corroborated by the FY2022 Q4 release. Settled.
Not contested: 38 of 48. This includes every E0 and E1 record in the ledger.
What would move me
Not more documents. Five things, none of which the ledger contains:
1. Recoverable amounts for non-impairing indicator-firms. One table of VIU or FVLCD against
carrying amount for firms that crossed and did not impair collapses Rival 1. Without it, delay
and correct accounting are observationally identical in this ledger.
2. Exchange-sourced crossing dates. Quarter-end closing price × total shares outstanding against
audited net assets, for each issuer, for every quarter-end in the window. The ledger names this
arithmetic in four separate falsifiers (C-12, C-34, C-40, C-48) and performs it zero times.
3. A stated conversion between the three "economic impairment" constructs, or an abandonment of
the imported durations.
4. A resolution of C-04 against C-43 sourced to IAS 36 and IAS 34 themselves, not to a Big-Four
summary page. The normative baseline for "delay" is currently undefined.
5. A defined population — an enumeration, or an explicit rule for entity level, CGU level,
window boundaries and "first." Ballard and TELUS both flip on definitions alone.
Bottom line
The ledger establishes, robustly, that several Canadian issuers took goodwill impairments in
2022–2025, that IAS 36 requires quarterly indicator assessment and annual goodwill testing, that an
indicator does not compel a charge, and that in Europe in 2011 and in the US in the 1990s
recognition frequently did not coincide with market signals. It does not establish a lag, because
it never measures one: it has two disclosed lags, both zero, both from the issuer's own narrative,
both with the confirming arithmetic left unrun; and it has three imported durations measured in
constructs that do not convert to the estimand's units. The gap between those two bodies is bridged
by four E2 inferences and three E3 model priors, of which one is falsified by its own source, one is
an arithmetic non-sequitur, one is untestable by construction, one measures a different quantity
than the claim it is attached to, and one is not about goodwill at all.
On the estimand: undetermined, and specifically undetermined between "Canadian issuers delay"
and Rival 1 ("recoverable amount lawfully exceeded carrying amount"), Rival 2 ("disclosed lag is
zero by construction"), and Rival 3 ("the foreign, older literature does not transfer to
post-2022 Canada"). Nothing in the 48 records discriminates among these four.
Sources retrieved this run (all pre-cutoff)
- [Grant Thornton International — IAS 36: if and when to undertake an impairment review](https://www.grantthornton.global/en/insights/articles/IFRS-ias-36/ifrs-ias-36-If-and-when-to-undertake-an-impairment-review/)
- [Hayn & Hughes, Leading Indicators of Goodwill Impairment (SSRN 850705)](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=850705)
- [André, Filip & Paugam, Examining the Patterns of Goodwill Impairments in Europe and the US (SSRN 2867340)](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2867340)
- [ESMA/2013/2, Review of practices related to goodwill impairment](https://www.esma.europa.eu/sites/default/files/library/2015/11/2013-02.pdf)
- [TELUS Corporation 2024 audited consolidated financial statements (Form 40-F ex. 99.4)](https://www.sec.gov/Archives/edgar/data/868675/000110465925012533/tu-20241231xex99d4.htm)
- [Corus Entertainment fiscal 2022 fourth quarter and year end results](https://www.prnewswire.com/news-releases/corus-entertainment-announces-fiscal-2022-fourth-quarter-and-year-end-results-301655889.html)
- [Corus Entertainment market cap history — StockAnalysis](https://stockanalysis.com/quote/tsx/CJR.B/market-cap/)
- [Lightspeed Commerce market cap history — StockAnalysis](https://stockanalysis.com/stocks/lspd/market-cap/)
- [Lightspeed Commerce market cap history — CompaniesMarketCap](https://companiesmarketcap.com/lightspeed-pos/marketcap/)
Quarter-end closing-price series for LSPD and CJR.B could not be retrieved (403 / robots-disallowed /
no 2022 coverage on three providers). That gap is stated explicitly wherever it bears on a contest,
and no contest is rested on data I could not obtain.
CONCESSION: MOVED — I came to contest the ledger's breadth and I concede its substrate entirely: I retrieved five of its sources directly and every E0 and E1 figure I could check was exact (ESMA 43%/235 and 47%; André's 20–25% and all three proxies; Hayn & Hughes's "three to four years" and "a third … up to ten years"; Grant Thornton's indicator list and each-reporting-period language; TELUS's no-impairment finding; Corus's $350.0M Television charge, which I corroborated outside the ledger's single PDF). Two concessions cost me real ground: C-33 is verbatim correct, which gives C-04 the best premise it could have had, and the Lightspeed charge is triangulated three ways including by the ledger's own equity roll-forward. I drop any thinness attack on the documentary record and I contest 10 of 48. What did not move is the bridge. Not one record in this ledger reports a recoverable amount for a firm that had an indicator and booked nothing — and that single absent quantity is what separates "issuers delay" from Rival 1, which TELUS's 2024 statements and Lightspeed's September 2022 test show is not hypothetical but documented, twice, inside this ledger. Both Canadian lags are disclosed rather than measured, both are zero, and the arithmetic the ledger itself names as their falsifier was never run; on Corus that arithmetic appears to run the wrong way by three quarters or more, and the "match" survives only by switching from market capitalisation to enterprise value across C$1.25bn of net debt. C-02 converts two prevalence statistics into a duration statistic that neither source reports and that a Q4 indicator with a Q1 charge already refutes. The estimand remains undetermined, and it is undetermined between four named, live explanations that these 48 records do not separate.
ROUND 2 — 03-attacks/null-r2.md
THE NULL — ROUND 2
Cutoff 2026-08-22 · Mode LIVE · Records in subset 20
flags.txt none · cutoff-failures.txt none · inadmissible-records.txt none
POSITION
I retrieved against 13 of the 20 records and hit primary documents in every case.
The ledger's facts are unusually clean — several verify to the digit and to the
comma. I am conceding 14 of 20 records outright, and I concede them because the
evidence settles them, not as a courtesy.
What does not survive is the inference the facts are recruited for. The ledger
is assembled to support a lag thesis: that S&P/TSX Composite non-financial issuers
recognize goodwill impairment materially later than the point at which net assets
crossing above market capitalization signalled it, calibrated against Hayn & Hughes's
three-to-four-year figure. Six records fail at claim level. Three further failures are
ledger-level and belong to no single record.
The decisive structural fact: every record in this subset carries sources: 1.
Not one claim in twenty is independently corroborated. Where the subset appears to
have breadth, shared-sources.txt shows the breadth is bookkeeping — and in one case
two LOAD-BEARING records are the same sentence in the same document.
CONTESTED
C-14 — the measurement does not exist
Thinness (terminal) · Live rival named
Tier E3. Source: "model prior, no external source." It is the only record in the
subset that attempts the quantity the whole ledger is built to produce, and it rests on
nothing external at all.
The record states its own defeat: the population is three (Lightspeed, Ballard,
TELUS Digital), "of which two are reclassifiable on definitional grounds alone."
An effective n of one. A median over that is not a measurement; it is a single
observation with a confidence interval spanning the hypothesis and its negation.
I confirmed one of the two reclassifications from documents already inside the ledger:
**Ballard impaired goodwill in 2023 and 2024** — C-46 (US$23,991 thousand, Ballard
Motive Solutions, discontinued operations) and C-28 (2024, the goodwill written to nil).
"First-time impairer" is therefore already indeterminate for a third of the population
without leaving the ledger. Whichever definition rescues Ballard as a 2024 first-timer
— continuing operations only? segment level? excluding discontinued operations? — is a
choice the ledger never states, and it is outcome-determinative at n=3.
The falsifier is aimed at the wrong risk. It reads: *"The bounded population of
S&P/TSX Composite non-financial first-time impairers over 2022 to 2025 exceeds one
hundred issuers."* That tests the universe. The live risk is the retrieved
population — that this run found three because three is what one run's retrieval
surfaces. No stated falsifier can catch that, so the record cannot be falsified by the
fact that would actually undo it.
RIVAL — retrieval-by-outcome. The three issuers were located because they
disclosed both a crossing and a subsequent impairment. Issuers that crossed and never
impaired are invisible; issuers that impaired without a crossing are invisible. A lag
estimated on the survivors of that filter is conditioned on the outcome it claims to
explain. Nothing in the ledger distinguishes "issuers delay recognition" from
"we only counted issuers who eventually recognized." C-27 is a live instance of the
first missing cell — Lightspeed crossed at September 30 2022 and recorded no
impairment at that date — and it is in the ledger only because Lightspeed impaired
one quarter later.
C-04 — the standard is quoted with the load-bearing half removed
Ill-posedness · smuggled premise
The claim states IAS 36 requires assessment *at the end of each reporting period whether
an impairment indicator exists*, and estimation of recoverable amount where one does.
True — that is IAS 36.9. It is also incomplete in exactly the place the thesis lives.
The record's own cited source carries the missing half verbatim. Grant Thornton,
the page at the URL in this record:
> "annually for the following types of assets, **irrespective of whether there is an
> indication of impairment**: intangible assets with an indefinite useful life…
> goodwill acquired in a business combination."
Name the premise: that recognition is triggered by an indicator, which is what
makes "crossing-to-recognition lag" a coherent quantity to measure. If goodwill is
tested annually regardless of any indicator, a gap between a crossing and a recognition
is not evidence that a test was deferred — the test ran anyway. Lightspeed's own
release says so: *"The Company conducts its annual goodwill impairment test every
December 31."*
The falsifier is aimed at the wrong risk. It seeks *"a provision permitting deferral
of impairment testing to a later period after an indicator arises."* No such provision
is needed to produce a gap. A test that runs on schedule, is performed correctly, and
concludes that recoverable amount exceeds carrying amount produces the identical
observable. The falsifier guards a door nobody used.
C-15 and C-21 — one sentence, entered twice, both LOAD-BEARING
Correlation · the breadth is bookkeeping
The two records carry verbatim-identical claim text and the same single URL
(sec.gov/…/earningsreleasefy23q3.htm, flagged in shared-sources.txt). Both are
marked LOAD-BEARING. Only the falsifiers differ. Two entries of apparent support are
one sentence in one document; a document does not corroborate itself by being cited
twice, and a ledger that counts it twice has overstated its own base by one.
I verified the sentence. It is there, exactly:
> "The Company's net assets exceeded its market capitalization as at December 31, 2022
> which was an impairment trigger for the Company."
But both records are silent on what else that release contains. The same release
records a US$748.7 million goodwill impairment at that same December 31 2022
date — confirmed against the FY2023 audited statements (*"Goodwill impairment 16
748,712"). Both records instead foreground the negative rider — "it does not establish
that date as the first period-end at which that condition held"* — while omitting that
for this issuer the crossing and the recognition are simultaneous.
Set against C-27 (crossing present at September 30 2022, test performed under a fair
value less costs of disposal model, *"no impairment of goodwill as at September 30,
2022"*), Lightspeed's actual crossing-to-recognition interval is one quarter. That
is the cleanest measured interval anywhere in this subset, and it sits roughly an order
of magnitude from the three-to-four-year anchor the ledger imports at C-45. Two
LOAD-BEARING records point at this issuer and neither one reports it.
C-16 — two entities compared as one, on a source that reaches only half of it
Live rival named · single-source over-reach
I verified the first half against the cited 40-F exhibit, verbatim:
> "Management determined that the recoverable amount of TELUS digital experience
> exceeded its carrying value as of the measurement date and, therefore, no impairment
> was recognized."
and the critical audit matter, which states the recoverable amount was determined
*"based on a fair value less costs of disposal calculation, which uses discounted cash
flow projections."* No headroom is quantified anywhere in the document.
Admissibility. The record has one source — TELUS Corporation's audited
statements — and that document cannot carry the second half of the claim at all.
TELUS Digital's market capitalization, its owners' equity, and its Q2 2025 charge
are in other documents entirely. I independently confirmed owners' equity of
US$1,945M at December 31 2024; the US$1.08bn market capitalization I could not
confirm from the cited source, because it is not in the cited source. This is the
inverse of the shared-sources.txt pattern: not several records resting on one
document, but one record reaching across three and citing one.
The comparison is not apples-to-apples. It sets TELUS Corporation's **CGU-level
carrying value, tested by discounted cash flow**, against TELUS Digital's
standalone consolidated owners' equity and its quoted price. Different
reporting entities, different carrying bases, different measurement objects. The record
presents the juxtaposition so the contradiction appears self-evident. It is not.
RIVAL — a minority float is not the fair value of the whole. The record's own
premise is that TELUS Corporation consolidates this business (that is why the CGU sits
in TELUS Corp's statements). A consolidated subsidiary's listed shares are a **minority
public float in a controlled entity**. Neither IAS 36 nor IFRS 13 treats a depressed
minority-float quotation as the fair value of the CGU. So:
- (a) management's DCF was optimistic and recognition was late; or
- (b) the traded price of a thin minority float sat below the entity's fair value,
and the test was correct at the measurement date.
Every fact in the ledger is consistent with both. No record measures this CGU's
recoverable amount independently of management, and no record reports the headroom.
Nothing discriminates.
The falsifier tests only the uncontested half — *"The 2024 statements record a
goodwill impairment loss for the TELUS digital experience unit."* They do not, and I
confirmed it. That leaves the actual load — the implication that the 2024 conclusion was
wrong — untested by construction.
C-19 — a checkable number, checked, wrong
A stated falsifier passing is not a record being correct
The claim: CSA Staff Notice 51-364 lists impairment of non-financial assets as one of
thirteen areas. The notice's passage, retrieved verbatim, enumerates twelve:
known trends, events and uncertainties · liquidity and capital resources · debt
covenants · risk factor disclosure · impairment of non-financial assets · going
concern · events after the reporting period · significant judgement and measurement
uncertainties · expected credit losses · financial instrument risk disclosure · non-GAAP
and other financial measures · material change reporting.
Count it: twelve.
Stakes for the thesis: near zero — impairment is on the list, which is the part that
matters. Stakes for the ledger's method: not zero. *The falsifier reads "The notice
omitting impairment of non-financial assets from that list."*** That falsifier passes.
It was always going to pass. It cannot see the one thing in the record that is wrong.
This is the smallest possible demonstration of the pattern running through C-04, C-07,
C-16 and C-21: the stated falsifier tests the half that is safe.
CONTESTED — LEDGER-LEVEL
Not defects of any single record; defects of the set.
L-1 — "crossing" runs three different measures under one word
**Ill-posedness. Name the term: crossing.**
- C-07 / Corus: carrying value vs. market enterprise value
- C-15 / C-21 / C-27 / Lightspeed: carrying amount of net assets vs. **market
capitalization**
- C-16 / TELUS: market capitalization vs. owners' equity
Enterprise value includes net debt; market capitalization does not. For a levered
broadcaster the two tests can cross in different quarters, or one can cross and the
other never. These are not three readings of one variable — they are three variables.
Any lag series that mixes them is measuring interval between events of different kinds.
The ledger never notices the substitution, because one word covers all three.
L-2 — the BCE block is carried under a shifted term
**Ill-posedness. Name the term: impairment.**
C-36, C-37 and C-41 all verify — $2,190M full-year 2024, $2,113M Q3 2024, Bell Media,
correct author, correct wire, correct dates. I contest none of them as claims.
But every one is about "non-cash asset impairment charges" on **Bell Media's TV and
radio properties. The word goodwill** does not appear in the Q3 2024 release at all.
Everywhere else the ledger's subject is goodwill impairment (C-14, C-22, C-45, C-46,
C-27, C-35, C-28). Broadcast licences are indefinite-life intangibles — tested under
IAS 36 alongside goodwill, and not goodwill. Nothing in the ledger establishes that
BCE's $2,190M contains any goodwill component.
Compounding it: C-37 and C-41 are the same event, same date — a wire story and the
release it covers. C-37's own text says so ("not as independent verification of it"),
and I credit that candour; it is the most honest sentence in the subset. But it means
the largest single-issuer block here — three of twenty records — is one issuer, one
event, one primary document, entered three times, under a term that has moved.
L-3 — two records support opposite propositions and nothing adjudicates
Live rivals, inside the ledger.
Both verified verbatim against their abstracts:
- C-11 (HXZ): *"65% of the 452 anomalies in our data library… cannot clear the
single test hurdle of the absolute t-value of 1.96."*
- C-26 (JKP): *"The majority of asset pricing factors: (1) can be replicated…
(3) work out of sample in a new large data set covering 93 countries."*
Both are marked SUPPORTING. They support incompatible propositions about whether
documented effects survive scrutiny — the disagreement is methodological (t-hurdles with
microcap mitigation vs. Bayesian shrinkage across factors) and it is unresolved in the
literature. The ledger contains no record adjudicating between them. Whatever prior
is being imported from this pair is being imported from a live dispute, and the ledger
cannot tell you which way it points.
Secondary: all three SSRN records (C-11, C-26, C-45) cite abstract landing pages,
not papers — one host, one artifact type, mutable by their authors. C-11's page carries
a documented revision history (May 3 2017 → July 27 2017 → October 31 2018) across which
the headline numbers moved. The cited October 31 2018 version matches the claim exactly,
so C-11 stands as filed; but three records are being read off pages that can change
under them, and shared-sources.txt correctly flags the concentration.
NOT CONTESTED
C-07 — Verified verbatim in the cited MD&A: *"The Company's share price has continued
to decline meaningfully from August 31, 2021, which resulted in the Company's carrying
value being greater than its market enterprise value at August 31, 2022, May 31, 2023,
and August 31, 2023."* The rider is also correct — November 30 2022 and February 28 2023
appear nowhere in that context. The record accurately reports both what the document
says and what it does not. *(It is also my cleanest thinness exhibit: the ledger's own
record concedes that two of the four intervening quarter-ends are simply unknown, which
is fatal to any quarter-end-granular crossing series for this issuer.)*
C-08 — Verified. Carla Nunes, managing director at Kroll, CFO Dive, April 6 2022,
Maura Webber Sadovi: "Relative to 2021 I expect impairments to go up." SUPPORTING, and
it supports only that a practitioner forecast a rise — which is exactly what it claims.
No over-reach.
C-11 — Verified verbatim against the cited October 31 2018 version: 452 anomalies,
65%, |t| ≥ 1.96, microcaps mitigated via NYSE breakpoints and value-weighted returns.
Every element of the claim matches the abstract word for word.
C-22 — I tried to falsify this and could not. Its falsifier is well-formed (a
pre-cutoff study, extract, or regulator publication reporting that issuer-level pairing
would kill it), and it held: Kroll's Canadian goodwill impairment studies stop at 2014,
and I found no publication pairing S&P/TSX Composite non-financial first goodwill
impairments with crossings at quarter-end granularity for 2022–2025. The evidence
settles it. Concede what it means: if no such dataset exists, then the only dataset
behind the lag thesis is C-14's — n=3, tier E3, model prior, no external source, two of
three reclassifiable. C-22 and C-14 read together are the ledger stating in its own
voice that its load-bearing empirical base does not exist and that it did not build one.
C-26 — Verified. The abstract says the majority of factors replicate and work out of
sample across 93 countries. The record's added mechanism wording ("hierarchical",
"cross-factor shrinkage", "non-zero posterior means") is accurate to the paper though
not to the abstract page cited. The finding as stated is right.
C-27 — Verified verbatim: *"The carrying amount of the Company's net assets exceeded
the Company's market capitalization as at September 30, 2022. This triggered an
impairment test" · "using a fair value less costs of disposal model" · "which
demonstrated no impairment of goodwill as at September 30, 2022."* Every clause of the
claim is in the document. *(This is also the single most important record in the subset
for the opposing case — see C-14 and C-04 above.)*
C-28 — Verified. Three indicators, exactly as named: decline in the Corporation's
market capitalization in 2024; the global corporate restructuring initiated in September
2024; indicators of slowing hydrogen and fuel cell policy implementation and market
adoption. Correct count, correct content, correct ordering.
C-31 — Verified verbatim: US$224 million for the quarter ended June 30 2025, and
*"The recoverable amount was principally affected by changes in key valuation
assumptions including higher weighted average cost of capital, lower perpetual growth
rate and lower cash flow forecasts arising from pricing pressure on margins."* The
record's negative rider is also correct: the release names no market-capitalisation
trigger. This record is scrupulous and it is the ledger's own defeater. The one
clean recognition event in the subset attributes causation to discount rates and forecast
revisions — a common-cause structure, in which rates and outlook move price and
recoverable amount together on different lags, not a lagged-recognition structure in
which management sat on a known fact. Nothing in the ledger separates the two.
C-35 — Verified. FY2023 statements show "Goodwill impairment 16 748,712" for the
year ended March 31 2023 and "Goodwill impairment 16 —" in the FY2022 comparative
column. No goodwill impairment before April 1 2022.
C-36 — Verified to the digit: *"non-cash asset impairment charges totalling $2,190
million, mainly related to Bell Media's TV and radio properties."* Amount and attribution
both exactly as claimed. *(Admission to a goodwill ledger is contested at L-2; the claim
itself is settled.)*
C-37 — Verified: Sammy Hudes, The Canadian Press, BNN Bloomberg, November 7 2024,
$2.11 billion, Q3 2024, Bell Media. And the record's own characterisation is correct —
it is wire coverage of BCE's release of the same date, quoting BCE's CEO and CFO, not
independent verification. The record says this about itself, unprompted, and it is right.
C-41 — Verified: *"$2,113 million, mainly related to Bell Media's TV and radio
properties"* for the quarter ended September 30 2024. Exact.
C-45 — Verified verbatim: *"On average, goodwill write-offs lag behind the economic
impairment of goodwill by an average of three to four years."* The record is careful and
correct, and it volunteers its own limits — the authors' own fundamentals-based
construct, a US sample framed around SFAS 142. I concede the claim entirely. *(What it
is asked to bear is contested elsewhere: a lag derived under a two-step US reporting-unit
test on pre-2005 data, using the authors' constructed proxy for "economic impairment," is
transported to a one-step IAS 36 CGU test with a mandatory annual cycle in Canada
2022–2025, with no record in the ledger bridging the regimes — and the one interval the
ledger actually measures is one quarter.)*
C-46 — Verified: US$23,991 thousand, Ballard Motive Solutions U.K. restructuring and
effective closure, presented in discontinued operations. Exact to the thousand.
WHAT WOULD MOVE ME
Not more records. Four specific things, none of which is in the subset:
1. A denominator. The count of S&P/TSX Composite non-financial issuers that crossed
and did not subsequently impair. Without it C-14 is conditioned on its outcome.
2. Disclosed headroom for the TELUS digital experience CGU at December 31 2024. It
would separate rival (a) from rival (b) at C-16 in one number. The document does not
quantify it.
3. One stated definition of "first-time impairer" applied before the population was
assembled rather than after — enough to settle whether Ballard is in or out.
4. One measured interval other than Lightspeed's. The ledger contains exactly one
crossing-to-recognition pair with both endpoints documented, and it is one quarter.
CONCESSION: MOVED — I expected the breadth to be illusory and it partly is: every record carries a single source, C-15 and C-21 are one sentence entered twice as two LOAD-BEARING records, C-28 and C-46 are one PDF, and C-37 and C-41 are one event. But I could not break the facts. Thirteen records I retrieved against verified against primary documents, several to the digit and the comma — C-27, C-31, C-45 and C-07 verbatim including their negative riders — and C-22's falsifier held under my own attempt to kill it. This is a ledger of accurate claims, and I concede 14 of 20 without reservation. I moved further than that: C-19 aside, the records I contest are mostly not wrong, they are insufficient, and I have said so rather than pretending otherwise. What I am not moved on is the verdict. The thesis needs a lag; the ledger measures one interval, Lightspeed's, and it is one quarter against an imported anchor of three to four years. Its own C-31 attributes the only clean recognition to WACC and forecast revisions and expressly names no market-cap trigger — common cause, not delayed recognition, and nothing here separates them. Its own C-22 says the required dataset does not exist; its own C-14 says the substitute is three issuers, tier E3, no external source, two reclassifiable. C-04 omits the annual-test rule that dissolves the framing, and C-16's contradiction evaporates once a minority float in a consolidated subsidiary stops being read as the fair value of a CGU. On the facts: moved. On the question: undetermined, and the ledger says so itself in three places.
ROUND 3 — 03-attacks/null-r3.md
NULL — ROUND 3
Run parameters. Cutoff 2026-08-22. MODE LIVE. Inputs: stripped.md (19 records), flags.txt = none, cutoff-failures.txt = none, inadmissible-records.txt = none, shared-sources.txt (3 host/document collisions). Retrieval performed against the open web within cutoff.
Position. I contest 4 of 19 records. I concede 15. Three of the four contested are LOAD-BEARING; the fourth is SUPPORTING but is the ledger's only quantitative bridge between two rival literatures.
PRELIMINARY: WHAT I CHECKED AND WHAT IT COST ME
I entered intending to run the correlation argument off shared-sources.txt. Retrieval mostly took it away from me. I record that before arguing, because it bounds what follows.
The correlation file is largely a false positive. papers.ssrn.com/sol3/papers.cfm → C-11, C-26, C-45 is a host-level collision, not a source collision: three different abstract_ids (3275496, 3774514, 850705), three different author teams, three different literatures, two decades apart. There is no single dataset, no single author, no single originating account behind this subset. I will not pretend otherwise.
What the collision does expose is a shared artifact-class defect, and this one is real: all three records cite an SSRN landing page, tier it E0, and assert content that lives in the paper body rather than the abstract. I verified the abstracts directly:
- HXZ (3275496) abstract does contain "65% of the 452 anomalies … cannot clear the single test hurdle of the absolute t-value of 1.96," with "microcaps mitigated via NYSE breakpoints and value-weighted returns." C-11's first sentence is verbatim-supported.
- JKP (3774514) abstract contains neither "82.4 percent" nor "153 factors." Both are true of the published Journal of Finance (2023) version, which I verified independently — but they are not in the cited artifact.
- Hayn & Hughes (850705) abstract does contain the three-to-four-year average and the one-third/ten-year tail. It does not contain the sample-composition or SFAS 142 sensitivity-test descriptions C-45 attributes to it.
Add that SSRN serves the current revision at a fixed abstract_id: the JKP page itself reads "posted March 5, 2021; last revised March 7, 2022." So the ledger's recorded dates do not pin the text that was read. This is a systematic tier-inflation across three records. It is not, however, an evidence failure in two of the three cases, because I was able to settle the underlying propositions by retrieval. I score it as a caveat, not a contest, except where the gap is load-bearing.
The other two collisions. ballard.com/Q4-24-Financial-Statements-FINAL.pdf → C-28, C-46 is one audited filing carrying two genuinely independent facts; I verified both. Correlated failure risk exists but is low, and neither claim exceeds the document. sec.gov/…/earningsreleasefy23q3.htm → C-15, C-21 is genuine duplication — and C-21 discloses and neutralises it itself.
One thinness observation stands unqualified. Every record in this subset carries sources : 1. Nineteen records, nineteen single-source claims, zero corroboration anywhere. There is no record in this subset "whose several sources turn out to be one," because no record has several sources. I calibrate this honestly: for an E0 claim of the form "filing F states S," F is the best available evidence and a second source adds little. Single-sourcing bites only where the claim exceeds the document. That is precisely the boundary my four contests sit on, and it is the only place I press it.
CONTESTED
C-07 — Corus — LOAD-BEARING, E0
Grounds: arithmetic impossibility; misattribution; selection that manufactures an interval; live rival.
Two halves, opposite fates. The quotation half is verbatim-correct — I pulled the cited MD&A and the sentence reads exactly as recorded, naming August 31 2021 as reference point and August 31 2022, May 31 2023 and August 31 2023 as the period-ends at which carrying value exceeded market enterprise value. I concede that half without reservation, including the record's careful distinction between carrying value vs. market enterprise value and IAS 36 ¶12(d)'s net assets vs. market capitalisation. That distinction is correct, material, and to the ledger's credit.
The second half fails.
1. It does not add up. The record states "$590.0M in the quarter ended May 31 2023, of which $295.2M goodwill, $219.8M broadcast licences and $175.0M brands." Those three components sum to $690.0M — the fiscal-year total the same sentence names — not $590.0M. The "of which" is arithmetically impossible on its own face.
2. The source assigns those components to the year, not the quarter. I retrieved the cited PDF. It records "non-cash impairments in the Television CGU to goodwill of $295.2 million," "broadcast licences of $219.8 million" and "brands and trade marks of $175.0 million" for the year ended August 31, 2023. It gives the $590.0M / $100.0M quarterly split but provides no component-level quarterly allocation. Corus's Q3 release likewise presents $590.0M as a single undifferentiated figure. The ledger holds one Corus source, and that source does not state when the goodwill was recognised.
3. The omission runs one way. Corus recognised a $350 million goodwill impairment in the Television segment in the quarter ended August 31, 2022 — the same period-end as the first crossing this record names. C-07 names that crossing and then reports recognition at May 31 2023, three quarters later, while the recognition that coincides with the crossing goes unmentioned.
Named rival. Leading explanation: Corus's carrying value crossed at August 31 2022 and goodwill recognition followed at May 31 2023 — a three-quarter interval. Rival: Corus recognised at the crossing, twice, with zero interval each time — $350M of goodwill in the quarter the crossing first occurred (Aug 31 2022), and the fiscal 2023 charges as a second and distinct impairment responding to further deterioration. Under the rival there is no lag to measure at all. Nothing in the ledger discriminates: it holds one document, that document does not allocate goodwill to a quarter, and it does not carry the fiscal 2022 recognition into the series it constructs.
C-11 — Hou/Xue/Zhang + Chen/Zimmermann — SUPPORTING, E0
Grounds: denominator equivocation; two incompatible generating processes for one number; single source covering two papers.
The first sentence is verbatim-supported by the cited abstract and I concede it. The second sentence is where the record does its actual work, and it misdescribes the relation between the two studies.
I retrieved the Chen & Zimmermann abstract (FEDS 2021-037). Verbatim: "Our 319 characteristics draw from previous meta-studies… For the 161 characteristics that were clearly significant in the original papers, 98% of our long-short portfolios find t-stats above 1.96… The remaining 114 characteristics were insignificant in the original papers or are modifications of the originals created by Hou, Xue, and Zhang (2020)."
The 161 is a subset of Chen & Zimmermann's own 319, not a restriction of HXZ's 452. The 98% is computed on C&Z's own long-short portfolios. C-11's phrasing — "the denominator includes… ; restricted to the 161 … puts 98 percent above |t| = 1.96" — reads as one denominator filtered down to a clean core, with 65% and 98% as two views of a single construction. They are not. They are two teams' portfolios, built on overlapping but non-identical characteristic universes, under different construction rules.
Named rival. Leading explanation (the one the record's framing invites): the 65%→98% gap is driven by denominator composition — originally-insignificant characteristics and HXZ-created modifications dragging the failure rate down; strip them and the literature is fine. Rival: the gap is driven by portfolio construction — HXZ uniformly impose NYSE breakpoints and value-weighting, which C-11 itself names, and that screen is what kills microcap-driven anomalies regardless of denominator. Under the rival, C&Z's 98% says nothing whatever about how those same 161 characteristics would fare under HXZ's construction, and the literature's health remains open.
The number that would discriminate — HXZ's failure rate restricted to originally-significant characteristics, computed under HXZ's own construction — appears in neither cited artifact and is nowhere in the ledger. The ledger holds two abstracts and cites only one of them: C-11 lists sources : 1, the HXZ page, which cannot support the Chen & Zimmermann half at all.
C-15 — Lightspeed — LOAD-BEARING, E0
Grounds: ill-posed measurand; an interval whose right endpoint is exogenous to its left; both endpoints underdetermined.
Every quotation in this record is correct. I retrieved the release: "The Company's net assets exceeded its market capitalization as at December 31, 2022 which was an impairment trigger for the Company"; "The Company conducts its annual goodwill impairment test every December 31"; the ($748.7) million charge. C-27 is likewise verbatim-correct on the September 30 2022 test. I concede all of it.
I contest the record's final sentence: "The documented crossing-to-recognition interval is one quarter, with both endpoints in issuer filings."
Name the term: "crossing-to-recognition interval." It equivocates between a calendar distance and a responsiveness lag, and only the first is documented. The record itself supplies the reason: it observes, correctly, that under IAS 36 ¶96 the December 31 test "would have occurred irrespective of the trigger." That concession is fatal to the measurement. An interval whose right endpoint is fixed by a standing annual calendar cannot measure responsiveness to an indicator — the endpoint was set before the crossing existed and would have arrived on the same date had no crossing ever occurred.
The left endpoint is no firmer. C-35 — verified, and correct in its own self-limitation — establishes only that no goodwill impairment was recognised before April 1 2022, and expressly disclaims establishing whether a crossing occurred or a test was performed in FY2022. The ledger holds no record at June 30 2022 (Q1 FY2023). September 30 2022 is the earliest documented crossing, not the established first crossing.
Named rival. Leading explanation: a crossing at September 30 2022 produced recognition one quarter later — delayed recognition. Rival: zero lag plus fresh information. The September 30 test was performed promptly, on the facts then available, and passed; the December 31 recognition followed a further and distinct deterioration and landed on a date fixed years earlier by the annual-test policy. The ledger's own material leans toward the rival — C-21 quotes the FY2023 statements attributing the trigger to changes occurring "During the three months ended December 31, 2022."
What would discriminate is the headroom at September 30 2022: a razor-thin pass supports the leading explanation, a comfortable one supports the rival. The ledger records the September 30 conclusion and not one number behind it. Absent that figure, "one quarter" is a distance between two calendar dates, one of which was never a function of the other.
C-22 — retrieval negative — LOAD-BEARING, E2
Grounds: a load-bearing negative whose search space excludes exactly where the object would live; self-referential evidence.
I do not contest this record's literal wording, which is scrupulous — it says the retrieval did not surface the pairing, and expressly declines to claim none could be built. I contest that a record of this construction can be LOAD-BEARING.
Its entire evidentiary content is the negative result of one agent's search session within one run, tiered E2, sourced to "Inference from retrieval … in this run." It is self-referential: the ledger is cited as evidence about the ledger. And the record names its own exclusions — Audit Analytics, S&P Capital IQ, Compustat/WRDS extracts, and SEDAR+ full-text search — which are, precisely and without close competition, the four places an issuer-level Canadian impairment-and-market-capitalisation pairing would be constructed. A negative result whose search space omits the object's most probable locations carries no weight against the object's existence.
Named rival. Leading explanation: no such external base exists, so the ledger's construction is genuinely novel. Rival: the base exists, or is trivially constructible, inside the four excluded sources — SEDAR+ full-text search over Canadian issuer MD&A alone would surface ¶12(d) language at quarter-end granularity across the S&P/TSX Composite. Nothing in the ledger discriminates, and nothing could, because the ledger's instrument was pointed away from both rivals' decisive terrain.
Secondary: the record's positive sub-claim that Duff & Phelps/Kroll's Canadian Goodwill Impairment Study is "the nearest precedent construction for this population" is loose on two counts. I verified the study's existence and cadence — the 2013 edition states it is "in its second year of publication," confirming a 2012 first edition, and 2014 is the last Canadian edition I can find, so "discontinued after 2014" holds. But the 2013 edition was prepared "in partnership with the Canadian Financial Executives Research Foundation (CFERF)," not FEI Canada directly, so the record's "(the 2012 jointly with FEI Canada)" singles out a relationship that was not unique to 2012. More substantively, that study reports aggregate goodwill impairment among Canadian public companies — it is not an indicator-to-recognition pairing at all, so calling it the "nearest precedent construction" for the described object compares two different objects.
NOT CONTESTED
NOT CONTESTED — C-04 — The paragraph numbers are right and the inference is valid. I verified ¶9, ¶10, ¶12(d) and ¶96 against the cited Grant Thornton guidance; each matches the record's characterisation, including ¶12(d)'s "carrying amount of the net assets of the entity is more than its market capitalisation." ¶99 is not discussed in the cited source — a sourcing gap I note — but the record's statement of it is a correct rendering of the standard and I can find no reading on which it fails. The closing inference ("an interval … is not by itself evidence that a test was deferred") is a validity claim, not an empirical one, and it is sound: an interval is consistent with both a deferred test and a completed test that passed. I would quibble only that "not by itself evidence" is strictly too strong where "not dispositive" is exact. The record does the Null's own work and I decline to attack it.
NOT CONTESTED — C-08 — Verbatim confirmed at the source. CFO Dive, April 6 2022, Maura Webber Sadovi, Carla Nunes as Kroll managing director: "Relative to 2021 I expect impairments to go up." Author, date, outlet, title and direction all match. The stated falsifier (a forecast of flat or declining levels) is squarely absent.
NOT CONTESTED — C-16 — Verbatim confirmed, including the negative. TELUS Corporation's 2024 Form 40-F exhibit 99.4 states that "Management determined that the recoverable amount of TELUS digital experience exceeded its carrying value as of the measurement date and, therefore, no impairment was recognized," on a fair value less costs of disposal calculation using discounted cash flow projections, and the auditor designated goodwill impairment a critical audit matter. I confirmed the record's own caveat too: no headroom figure is quantified anywhere in the CAM or Note 18(d). The record states what the document states and disclaims what it does not.
NOT CONTESTED — C-19 — The full list matches, in order. I pulled the notice. The twelve items following "Some areas that may be impacted by the current economic environment include" appear exactly as enumerated, with impairment of non-financial assets in position five, and the "Some areas … include" framing confirms non-exhaustiveness. There is nothing left open here.
NOT CONTESTED — C-21 — Settled by the same document that settles C-15, and self-neutralised. The release sentence and the fuller FY2023 narrative both verified. The record discloses that it is a second entry of one sentence in one document at one URL, weights itself at zero, and states that its LOAD-BEARING tag adds nothing. That is the correct disposition and it is the ledger's own. I note only that the tag still counts one record toward a load-bearing base of ten; the ledger's effective load-bearing base is nine. Contesting a duplicate the ledger has already voided would be padding.
NOT CONTESTED — C-26 — The numbers are correct; I settled them myself. The published Journal of Finance version confirms 82.4%, a data set of 153 factors, the posterior threshold z̄ = 1.96, the equivalent statement that the posterior probability of the null falls below 2.5%, and the 93-country out-of-sample set. The record is also correct and unusually careful in keeping the replication rate and the out-of-sample result as distinct findings and asserting no conjunction. Caveat recorded, not scored: the cited artifact — the 2021 SSRN abstract page — contains none of these figures, and the page shows a March 2022 revision, so the recorded date does not pin the text. The proposition is settled; the ledger's route to it is not the one it names.
NOT CONTESTED — C-27 — Verbatim confirmed at the primary filing. The Q2 FY2023 interim statements state the test reason ("The carrying amount of the Company's net assets exceeded the Company's market capitalization as at September 30, 2022"), the fair value less costs of disposal model built on discounted cash flows, and the conclusion of no goodwill impairment at that date. Every element of the claim is present and the falsifier is absent. This record is the cleanest thing in the subset.
NOT CONTESTED — C-28 — Three indicators, market capitalisation first. The FY2024 goodwill note names exactly the three the record names: the decline in the Corporation's market capitalization in 2024, the global corporate restructuring initiated in September 2024, and indicators of slowing hydrogen and fuel cell policy implementation and market adoption. Count, content and ordering all confirmed. Sharing a PDF with C-46 does not damage it — one filing carries independent facts, and this fact is not one C-46 depends on.
NOT CONTESTED — C-31 — Verbatim confirmed, including the disclaimed absence. US$224 million non-cash goodwill impairment for the quarter ended June 30 2025, and the attribution quoted exactly: "higher weighted average cost of capital, lower perpetual growth rate and lower cash flow forecasts arising from pricing pressure on margins." I also confirmed the negative — the release does not identify any indicator or explain what prompted the test at that date. Caveat recorded, not scored: the record's stated falsifier points at TELUS Digital's Q2 2025 interim statements, a document the ledger does not hold, so the falsifier cannot be executed from within the ledger.
NOT CONTESTED — C-35 — Confirmed, and correctly self-limited. The FY2023 audited statements show nil goodwill impairment in the March 31 2022 comparative against $748,712 thousand for FY2023, and carry no FY2022 testing or indicator disclosure. The record's epistemic restraint — that this establishes only the absence of a recognised loss before April 1 2022 and nothing about crossings or tests — is exactly right, and it is the sentence on which my C-15 contest partly rests.
NOT CONTESTED — C-36 — Verbatim confirmed and correctly deflated. The FY release states "non-cash asset impairment charges totalling $2,190 million, mainly related to Bell Media's TV and radio properties to reflect a further decline in demand and spending in the traditional advertising market," with no goodwill component and no further decomposition. The record's inference that broadcast licences are indefinite-life intangibles tested under IAS 36 but are not goodwill, and that the record therefore supports no goodwill observation, is correct and is the Null's own conclusion. Caveat recorded, not scored: grouping this February 2025 full-year release with the November 2024 Q3 release as "one disclosure event" merges two documents three months apart reporting different totals, and silently absorbs the $77M difference between $2,190M and $2,113M, which no record in the subset addresses. The merge errs toward deflation, so I let it stand.
NOT CONTESTED — C-37 — Every checkable element confirmed. Sammy Hudes, The Canadian Press, November 7 2024, "$2.11 billion in asset impairment charges, mainly related to Bell Media's TV and radio properties." The word "goodwill" appears nowhere in the article. The independent content is exactly as described: the CRTC fibre-access decision, shares closing at $38.94 — below $40 for the first time since 2012 — and Scotiabank's Maher Yaghi on churn. The record identifies itself as derivative wire coverage and excludes itself from source totals, which is correct. I note only that "no goodwill component is stated anywhere in the reporting" is scope-ambiguous between the press chain and BCE's own reporting; on the press-chain reading, which is the natural one, it is verified.
NOT CONTESTED — C-41 — Verbatim confirmed. The Q3 2024 release shows impairment of assets of $2,113 million for the quarter, "mainly related to Bell Media's TV and radio properties," attributed to a further decline in advertising demand and spending, with no goodwill component and no allocation. The record's reading of "mainly related to" as leaving an unquantified remainder is correct, and its refusal to extract a goodwill observation is the right call on this evidence.
NOT CONTESTED — C-45 — The headline finding is verbatim in the cited abstract. "Goodwill write-offs lag behind the economic impairment of goodwill by an average of three to four years. For a third of the companies examined, the delay can extend up to ten years." The record's arithmetic point holds on inspection: a third of observations near ten years against a mean of three to four forces the remaining two-thirds to average roughly a quarter of a year, so a median well below three years is not merely compatible but implied. The record names its own key equivocation — "economic impairment" measured on the authors' own fundamentals-based construct, not a regulatory or accounting quantity — and declares itself non-transportable to a one-step IAS 36 CGU test with a mandatory annual cycle. Caveats recorded, not scored: the sample-composition and SFAS 142 sensitivity-test clauses are body-of-paper content absent from the cited abstract page; and the cross-reference to C-30 cannot be checked from within this subset. Neither caveat threatens the claim, and the claim's own conclusion is a limitation rather than an assertion.
NOT CONTESTED — C-46 — Amount, description and presentation all confirmed. The FY2024 statements record goodwill impairment charges of $23,991,000 in the 2023 comparative "related to the restructuring and effective closure of operations of Ballard Motive Solutions in the U.K.," cross-referenced to the discontinued-operations note, sitting inside a net loss from discontinued operations of $33,506,000 that comfortably contains it alongside the $2,266,000 intangible write-down. My first read of this PDF returned the opposite answer on presentation and the targeted re-read overturned it; I record that because the reversal is the reason I am confident, not in spite of it. The record's accounting caution is also correct: comparatives are re-presented on discontinued-operations classification, so this filing establishes presentation in itself and cannot establish presentation as originally reported. And its substantive point stands — a closure-driven write-off inside discontinued operations is not an indicator-triggered CGU test outcome and does not belong in a market-capitalisation-indicator-to-recognition series.
SUMMARY OF POSITION
| | Count | Contested |
|---|---|---|
| LOAD-BEARING | 10 (9 effective; C-21 self-voided) | 3 — C-07, C-15, C-22 |
| SUPPORTING | 9 | 1 — C-11 |
| Total | 19 | 4 |
Tiers: E0 = 13, E1 = 4, E2 = 2. Both E2 records are LOAD-BEARING; I concede one (C-04) and contest the other (C-22).
The subset's E0 primary-document quotations are, with a single exception, exactly what they claim to be. I checked fifteen of them against their own cited filings and fourteen came back verbatim. That is a genuinely strong ledger at the level of what documents say, and several records — C-04, C-21, C-35, C-36, C-37, C-41, C-45, C-46 — perform their own deflation, disclaim their own reach, or refuse an inference the surrounding thesis would have wanted. A ledger that argues against itself in eight places is not one I can dismiss.
The failure is not distributed. It sits at exactly one seam: where a record stops reporting a document and starts constructing a quantity. C-07 builds a component allocation the document does not make and omits the recognition that coincides with the crossing it names. C-15 builds an "interval" whose right endpoint is fixed by a standing annual calendar and whose left endpoint is merely the earliest documented one. C-11 builds a filtered denominator across two teams' incompatible constructions. C-22 builds a negative from a search that excluded the four places the object would live. In all four, a rival explanation is available, and the ledger holds nothing that separates it from the leading one.
The three quantities that would settle all four are absent and specific: Lightspeed's headroom at September 30 2022; Corus's goodwill allocation between the quarters ended May 31 and August 31 2023, together with the fiscal 2022 $350M charge; and HXZ's failure rate restricted to originally-significant characteristics under HXZ's own construction. None is unobtainable. Two are almost certainly in documents adjacent to ones the ledger already holds.
CONCESSION: MOVED — I came to argue correlation and left unable to. shared-sources.txt is a host-level artifact, not a common origin: three SSRN abstract_ids are three papers, not one, and no dataset, author or account stands behind this subset. The Lightspeed duplication was disclosed and voided by the ledger before I reached it, and the Ballard PDF carries two independent facts I verified separately. Fifteen of nineteen records survived direct retrieval against their own primary sources, four of them — C-19's twelve-item list, C-27's test disclosure, C-37's article, C-45's abstract — word for word, and eight records volunteer limitations I had prepared to impose on them. On thinness I am also moved: single-sourcing is not a defect for a claim of the form "filing F states S," and I withdraw it as a general argument, retaining it only where a claim exceeds its document. I hold four. C-07 I hold hardest, and on new evidence rather than on doubt: the components are misallocated by arithmetic the record cannot survive, and Corus recognised $350M of Television goodwill in the very quarter of the first crossing the record names — a fact that does not weaken the lag conclusion so much as invert it. C-15 and C-22 I hold on structure, not on error: an interval cannot measure responsiveness when a standing annual test date fixes its right endpoint, and a negative cannot bear load when its search space excluded SEDAR+, Compustat/WRDS, Capital IQ and Audit Analytics. C-11 I hold because 65% and 98% are not two views of one denominator and the ledger has no number that decomposes the gap. Undetermined, on these four, is the finding — not the fallback.
ADVOCATE OUTPUT
ROUND 1 — 04-advocate/advocate.md
THE ADVOCATE — salvage pass on the stripped ledger
Run parameters. Evidence cutoff 2026-08-22. Mode LIVE. Inputs: stripped.md (48 records),
inquisitor.md (47 records attacked; C-38 unattacked), null.md (10 records contested plus one
joint inference; 38 conceded). All retrieval below was performed this run against pre-cutoff sources.
0. What I did, and what it cost the ledger
Both attackers converge on one demand: the ledger names an arithmetic in four falsifiers (C-12,
C-34, C-40, C-48) — quarter-end market capitalisation against audited net assets — and performs it
zero times. The Null lists it first among the five things that would move it. I performed it, at
annual granularity, for three of the ledger's five issuers, using audited equity from the issuers'
own statements and market capitalisation from price aggregators.
It runs against the ledger. Three Canadian, IFRS, in-window observations:
| Issuer | Crossing (net assets > market cap) | Recognition | Lag |
|---|---|---|---|
| Lightspeed | Disclosed by the issuer at 30 Sep 2022; tested, no impairment | 31 Dec 2022 | 1 quarter |
| Ballard | Absent at 31 Dec 2023 (mcap ≈US$1.10–1.11bn vs equity US$991.2M); present during 2024 | Testing at 30 Sep 2024 | ≤ 2 quarters |
| TELUS Digital | Absent at 31 Dec 2023 (mcap US$2.35bn vs owners' equity US$2,037M); present at 31 Dec 2024 (mcap US$1.08bn vs US$1,945M) | 30 Jun 2025 | 2–5 quarters |
Sources: Lightspeed FY23-Q2 statements, goodwill note (retrieved verbatim this run); Ballard Q4-24
audited statements (equity US$991,216k at 31 Dec 2023; US$672,990k at 31 Dec 2024) and Q2-24 interim
statements (equity US$922,132k, 299,408,379 shares at 30 Jun 2024); companiesmarketcap /
stockanalysis market-cap histories for BLDP and TIXT; TELUS Digital Q4-2024 and Q2-2025 results
releases (owners' equity US$2,037M / US$1,945M / US$1,694M).
Three consequences, all adverse to the ledger and all conceded below:
1. The ledger's two zero-quarter records are false. Lightspeed's own Q2 FY2023 goodwill note —
cited three times in this ledger and never quoted — says the crossing was live at 30 September
2022. C-12 and C-40 are conceded outright.
2. The central bridge does not survive. No measured Canadian lag here exceeds four quarters and
the central value is one to two. C-02 is conceded.
3. What the measurement buys back is Rival 2. A disclosed crossing appearing in a no-charge
filing a full quarter before recognition, plus two crossings established from audited equity and
exchange data with no issuer narrative involved, is exactly the evidence Rival 2 says cannot
exist. That rival is defeated. See §2.
Method caveat, stated once. Market capitalisation at year-end comes from secondary aggregators,
and the comparison is annual, so it brackets rather than pins the crossing quarter for Ballard and
TELUS Digital. I could not retrieve a pre-cutoff quarter-end closing-price series for BLDP, TIXT or
LSPD (403 / paywalled / no coverage on five providers — the same wall the Null hit). No verdict
below rests on a price I did not retrieve. The bounds above use only figures I obtained.
1. Verdicts
C-01 │ REBUILT: IOSCO's Recommendations on Accounting for Goodwill (FR13/23, December 2023)
states that "the issue of 'too little, too late' goodwill impairment and the issue of insufficient
disclosure on goodwill impairment tests are priority issues in financial reporting"; the Reuters wire
cited in the ledger reports that position accurately. — The PROVENANCE attack is sound and the
remedy is to retrieve the primary, which I did. The record upgrades from a third-hand wire (E1) to
the regulator's own document (E0); the attributed characterisation is IOSCO's own words.
C-02 │ CONCEDED: This is the ledger's spine and it does not hold. Both attackers identify the
same fatal frame error and they are right: an indicator in Q4 followed by a charge in Q1 is a
one-quarter lag counted as annual non-recognition, so an annual non-recognition rate places no lower
bound on a quarterly median. Neither ESMA/2013/2 nor André/Filip/Paugam reports a recognition-timing
distribution, in quarters or in any other unit. Worse for the record, my own measurement runs the
other way: the three Canadian lags I could bound are one quarter, at most two, and two-to-five. What
survives is only this, and it is not the claim: *a majority of European issuers carrying a
market-based impairment indication did not recognise a goodwill impairment in the same annual
reporting period (ESMA 2011: 53%; André/Filip/Paugam 2006–2015: 75–80%).* The "median above four
quarters" content is unrecoverable from these sources.
C-03 │ HELD AS WRITTEN: Survives the PROVENANCE attack. Duplication with C-07 is a defect in the
ledger's bookkeeping, not in the claim — the Null verified the disclosure independently against
Corus's FY2022 year-end release and does not contest it. The correct remedy is to weight C-03 and
C-07 as one record, not to discount either.
C-04 │ REBUILT: IAS 36 requires an entity to assess at the end of each reporting period whether
an impairment indicator exists and, where one does, to estimate the recoverable amount of the asset
or CGU. — The words "makes a lag beyond one quarter a departure from required timing" are dropped;
they equivocate between assessment timing (mandatory, each reporting period) and recognition timing
(contingent on measurement), and both attackers catch it. The ledger's own C-13 and C-27 refute the
original: Lightspeed assessed, tested and recognised nothing at 30 September 2022, departing from
nothing. Note also that the source page says nothing about quarters; interim frequency is an IAS 34
matter the ledger never sources.
C-05 │ HELD AS WRITTEN: Verified verbatim by me and by both attackers — the page lists "Carrying
amount of the net assets of the entity is more than its market capitalisation" among the external
indicators. The attack is duplication with C-20 and tier inconsistency; that is ledger hygiene, and
the claim is exactly true.
C-06 │ HELD AS WRITTEN: Survives PROVENANCE. Duplication across C-23 and C-42 is a weighting
defect; the charge itself is triangulated three ways, including by the ledger's own equity
roll-forward (US$3,304.4M → US$2,539.5M). The Null concedes it explicitly.
C-07 │ REBUILT: Corus's fiscal 2023 annual MD&A names August 31 2022, May 31 2023 and August 31
2023 as period-ends at which carrying value exceeded market enterprise value. The list is a
disclosure set, not a crossing series: it is silent on the intervening quarter-ends of 30 November
2022 and 28 February 2023 and does not establish that the condition failed at them.
C-08 │ REBUILT: In April 2022 Kroll managing director Carla Nunes told CFO Dive she expected 2022
goodwill impairments among US public companies to rise relative to 2021. — Narrowed to what it is: a
dated ex ante forecast about US-GAAP filers under ASC 350, from an "early assessment" rather than a
published study. It carries nothing about IFRS issuers, outturns, or indicator-to-recognition lag,
and it is not independent of C-22's Kroll source. Conceding its scope matters: as the Null notes, it
supports Rival 3 rather than the leading account.
C-09 │ HELD AS WRITTEN: Verified exact by the Null against the primary ("As of 31 December 2011,
43% of the sample showed a market capitalisation below equity"; 235 entities). The attack is
non-independence with C-02 and C-29, which is a weighting instruction, not a defeat.
C-10 │ HELD AS WRITTEN: Verified exact ("Only 20 to 25% of firms recognize impairments depending
on the measure of economic impairment"). The attack is that C-10 and C-24 are one abstract split in
two — true, and again a weighting instruction.
C-11 │ REBUILT: With microcaps mitigated via NYSE breakpoints and value-weighted returns, 65% of
the 452 anomalies in Hou, Xue and Zhang's data library fail to clear |t| ≥ 1.96. — The construction
condition is restored, and I drop any use of the record as a general reliability discount: it is a
long-short factor-return result, not a property of published findings at large, and it does not
transfer to descriptive recognition frequencies read off filings.
C-12 │ CONCEDED: Decisive, and I verified it directly rather than taking either attacker's word.
Lightspeed's FY23-Q2 goodwill note reads: "During the six months ended September 30, 2022, the
Company's share price and therefore its market capitalization decreased. The carrying amount of the
Company's net assets exceeded the Company's market capitalization as at September 30, 2022. This
triggered an impairment test," and "The Company completed an impairment test of goodwill as at
September 30, 2022 using a fair value less costs of disposal model which demonstrated no impairment
of goodwill as at September 30, 2022." The disclosed market-capitalisation trigger is live one full
quarter before recognition. The disclosed lag is one quarter, not zero. The record is false as
written and the ledger cited that PDF three times without quoting the sentence that falsifies it.
C-13 │ HELD AS WRITTEN: Verified verbatim. The PROVENANCE attack is that the ledger draws
mutually inconsistent records off one page — correct, but the inconsistency is C-43's, not C-13's.
C-13 is the faithful reading; C-43 is the misread. See C-43.
C-14 │ REBUILT: Where the in-window population of first-time impairers is small, single-issuer
reclassification moves the reported median — and on this run's retrieval the population of
S&P/TSX-listed issuers in this ledger with a determinable crossing-to-recognition pair is three
(Lightspeed, Ballard, TELUS Digital), of which two (Ballard on the 2023 discontinued-operations
charge, TELUS on the choice of filer) are reclassifiable on definitional grounds alone. — The
original asserted an antecedent it never established; I have now established it by count rather than
by prior. The record no longer needs the population enumeration that C-22 says does not exist, so
the self-blocking pair the Null identifies is broken. It remains E3 as to the general statistical
proposition, and it now cuts against the ledger that contains it.
C-15 │ REBUILT: Lightspeed's Q3 FY2023 release states that at 31 December 2022 the carrying
amount of its net assets exceeded its market capitalization and that this triggered an impairment
test. It does not establish that 31 December 2022 was the first period-end at which that condition
held — Lightspeed's own Q2 FY2023 statements say the identical condition triggered a test at
30 September 2022.
C-16 │ REBUILT: TELUS Corporation recognized no goodwill impairment for the TELUS digital
experience CGU in 2024 because management determined the recoverable amount exceeded carrying value
at the measurement date — a judgement the auditor designated a critical audit matter — while the same
business's listed vehicle carried market capitalization of US$1.08bn against owners' equity of
US$1,945M at 31 December 2024, and recognized a US$224M goodwill impairment two quarters later. —
Non-recognition is restated as what it is (a measurement outcome, not evidence of absent economic
impairment), and the two facts that make it discriminating are attached. See §2, Rival 1.
C-17 │ CONCEDED: LOAD-BEARING on zero external evidence, and I cannot repair that. The source
line is "model prior, no external source"; the date field is the run date; the claim is an empirical
proposition about how a countable class of resolved forecasts behaves, and no calibration archive,
resolution dataset or scoring history is cited or reachable. It also presupposes a "reference class
central estimate" that §1.2 of the Null shows does not exist in convertible units. Most seriously, it
is a claim about the form of the question rather than about Canadian issuers, so it discriminates
between none of Rivals 1–5. Both attackers are right and I have nothing to add in its defence.
C-18 │ HELD AS WRITTEN: Verified. I also close the Null's caveat that this "is not a total share
count and so cannot on its own support a market-capitalisation computation": the FY23-Q2 share
capital note shows no other class outstanding at 30 September 2022, and the arithmetic checks —
Lightspeed's independently reported market capitalisation of ≈US$2.15bn at 31 December 2022 divided
by 150.3M shares implies ≈US$14.30 per share, a coherent LSPD close. The figure does support the
computation, and I performed it.
C-19 │ REBUILT: CSA Staff Notice 51-364 lists impairment of non-financial assets as one of
thirteen areas that may be impacted by the current economic environment. — Narrowed: it sits in an
undifferentiated enumeration alongside going concern, debt covenants and expected credit losses, in a
notice covering fiscal years ended 31 March 2022 and 2021. It is not a finding, a priority
designation, a deficiency rate, or a statement about impairment timing.
C-20 │ HELD AS WRITTEN: Verified verbatim, as with C-05. The duplication is real and should be
collapsed; the citation-hygiene point — that the falsifier invokes IAS 36 ¶12, a text the ledger
never sources — is a genuine gap I concede without it touching the truth of the claim.
C-21 │ REBUILT: Same narrowing as C-15: the release states the condition and the trigger at
31 December 2022; it does not establish that date as the first crossing, and Lightspeed's Q2 FY2023
statements show it was not.
C-22 │ REBUILT: No published dataset pairs S&P/TSX Composite non-financial issuers' first
goodwill impairments with net-assets-above-market-capitalization crossings at **quarter-end
granularity for 2022–2025**. — This is my strongest salvage and it survives the Inquisitor's
counter-dataset. The Duff & Phelps / CFERF Canadian Goodwill Impairment Study is real and I
confirmed it on the Kroll hub, but it does not do what C-22 denies: the Kroll publications page lists
**three Canadian editions only — 2012 (13 Feb 2013), 2013 (19 Dec 2013) and 2014 (18 Dec 2014) — and
none later**; the series covers 2007–2013, pairs impairment with below-book trading at annual and
aggregate level (72% of impairing companies below book at the 2008 peak vs 52% of all Canadian
companies), and contains no quarter-end crossing construct and no "first impairment" construct. The
US series continues to a 2025 edition; the Canadian one stopped twelve years before the estimand
window opens. The Inquisitor's SELECTION attack lands against C-22 as originally worded — the
retrieval genuinely missed a relevant series on the very hub it cited, and I concede that — but the
narrowed claim is intact. I also drop the universal negative: the record now asserts a window- and
granularity-scoped gap, not a fact about all publications everywhere.
C-23 │ HELD AS WRITTEN: Duplication with C-06 and C-42 only. The Null concedes the figure.
C-24 │ HELD AS WRITTEN: Verified exact against the abstract — market-to-book below one, equity
market value minus book value below goodwill, negative EBITDA. The attack is that it splits one
abstract with C-10; weighting, not truth.
C-25 │ HELD AS WRITTEN: Survives the single-document attack outright, because the Null
corroborated it outside the ledger's Corus PDF against the fiscal 2022 Q4 release. This is one of
the few records in the ledger with genuine source independence.
C-26 │ REBUILT: Jensen, Kelly & Pedersen find that under a Bayesian hierarchical model of factor
replication with cross-factor shrinkage, the majority of asset-pricing factors have non-zero
posterior means, holding out of sample across 93 countries. — The appended clause "contradicting
broad claims that published quantitative findings collapse under re-measurement" is dropped: it
generalises a finance-factor estimator dispute to published quantitative findings at large, which the
abstract does not support. With C-11 likewise narrowed, I accept the Null's §4.7 conclusion — the
imported replication calibration nets to zero and should carry no weight against the accounting
records.
C-27 │ REBUILT: Lightspeed performed a goodwill impairment test as at 30 September 2022 because
the carrying amount of its net assets exceeded its market capitalization, using a fair value less
costs of disposal model, and the test demonstrated no impairment of goodwill at that date. — The
Inquisitor's charge of misreading by omission is correct and the omitted clause is load-bearing: I
retrieved the note and the trigger sentence is there. Reinstating it is what falsifies C-12 and C-40,
and I reinstate it rather than defend the omission.
C-28 │ REBUILT: Ballard's 2024 goodwill note names the decline in the Corporation's market
capitalization in 2024 as one of three indicators of potential impairment, alongside the global
corporate restructuring initiated in September 2024 and indicators of slowing hydrogen and fuel cell
policy implementation and market adoption. — I add a fact that strengthens what remains: the 2024
test itself was performed on "a modified market capitalization assessment," deriving enterprise value
from the September average share price plus a control premium. So for Ballard the recoverable amount
was itself computed from market capitalisation. The Inquisitor is right that the September
restructuring is a sufficient independent trigger and that this record cannot on its own date a
market-cap-indicator-to-recognition interval; the bracketing in §0 does that instead.
C-29 │ HELD AS WRITTEN: Verified exact by the Null against the primary. Non-independence with
C-09 and C-02 is a weighting instruction. What the 47% means is Rival 1's territory, addressed in §2.
C-30 │ HELD AS WRITTEN: Verified verbatim ("For a third of the companies examined, the delay can
extend up to ten years"). Splitting one abstract with C-45 is a weighting defect.
C-31 │ REBUILT: TELUS Digital recognized a US$224 million non-cash goodwill impairment for the
quarter ended 30 June 2025, attributed to a higher weighted average cost of capital, a lower
perpetual growth rate and lower cash flow forecasts arising from pricing pressure on margins; the
release names no market-capitalisation trigger. — I confirmed amount, period and attribution against
the release. The provenance objection is fair (a republication where the ledger goes primary
elsewhere) and the primary exists. The narrowed record cannot by itself supply a market-cap-indicator
date — but §0 supplies one from the balance sheet and exchange data instead, and that is what makes
TELUS Digital the ledger's longest measurable lag.
C-32 │ HELD AS WRITTEN: Accurate as to the Q3 FY23 note's language, which the Null verified. The
attacks are that it restates the same event under a different URL (weighting) and that the date field
is the balance-sheet date rather than the publication date (metadata). Neither touches the claim.
C-33 │ HELD AS WRITTEN: Verified verbatim by me and both attackers: "IAS 36 requires an entity to
assess at the end of each reporting period whether there is any indication that an asset or CGU may
be impaired." The PROVENANCE attack is that one page yields contradictory records — true, and C-33 is
the correct one. It is the record that falsifies C-43.
C-34 │ CONCEDED: Three independent defeats and I can rebut none of them. (i) The claim asserts
market enterprise value while its falsifier tests market capitalisation — quantities separated by
Corus's C$1,246.1M of long-term debt, so the falsifier cannot discriminate and the "match" survives
only by switching measures mid-inference. (ii) "Earliest disclosed" is not "earliest occurring"; the
FY2023 MD&A never claims to be a census, and the same MD&A dates the deterioration a year earlier.
(iii) Fatal, and I verified it directly: Corus recognized **$673.0 million of Television goodwill
impairment and $46.0 million of Radio goodwill impairment in the quarter ended 31 May 2020**, plus
$67.8 million of Radio broadcast licence impairment. Fiscal 2022 is therefore not Corus's first
goodwill impairment, 31 August 2022 is not its first crossing, and Corus does not belong in a
first-impairment population at all. The record contributes no lag observation, which is why §0 has
three issuers rather than four.
C-35 │ REBUILT: Lightspeed's FY2023 audited statements report no goodwill impairment loss in the
comparative fiscal year ended 31 March 2022. — The provenance objection is right that a comparative
column is not the primary record of the period, so the record should name its basis. Neither attacker
disputes the underlying fact and the Null concedes it.
C-36 │ REBUILT: BCE reported non-cash asset impairment charges totalling $2,190 million for 2024,
mainly related to Bell Media's TV and radio properties. — I verified the release: "non-cash asset
impairment charges totalling $2,190 million, mainly related to Bell Media's TV and radio properties."
The original dropped "and radio," and the record's own falsifier tests attribution, so it fired. The
figure is correct; the attribution needed the correction.
C-37 │ REBUILT: The Canadian Press reported that BCE recorded $2.11 billion of asset impairment
charges in Q3 2024, mainly related to Bell Media properties — as wire coverage of BCE's own release
of the same date, not as independent verification of it. — I concede the independence failure the
shared-source audit missed: C-36, C-37 and C-41 are one corporate disclosure event and should be
weighted as one, not three.
C-39 │ HELD AS WRITTEN: Both equity figures verified, and their difference against C-38
corroborates the charge's magnitude independently. The attack is that the record omits the
30 September 2022 quarter-end that straddles the two endpoints — a fair criticism of the ledger's
use of the figures, which §0 now remedies, not of the figures themselves.
C-40 │ CONCEDED: Same defeat as C-12, from the same retrieved sentence. Lightspeed's disclosed
trigger date is 30 September 2022 and its recognition date is 31 December 2022, so the disclosed lag
is one quarter. The record's own falsifier did not even need exchange data to fire — the issuer's
prior filing says it in words.
C-41 │ REBUILT: BCE's Q3 2024 results news release states $2,113 million of non-cash asset
impairment charges against Bell Media for the quarter ended 30 September 2024. — Accurate as to the
release; I concede it is a release rather than the interim financial statements its own falsifier
appeals to, and that it is one event with C-36 and C-37.
C-42 │ HELD AS WRITTEN: Third statement of one charge (weighting), from the audited-statement
side. The Null concedes it. Sharing a PDF with C-35 is hygiene.
C-43 │ CONCEDED: Falsified by the source it cites, and I verified the source. The Grant Thornton
page states that IAS 36 requires assessment "at the end of each reporting period"; the
annual-test-timing sentence sets a minimum frequency for goodwill, not a permission to defer
recognition once an indicator arises. IAS 36 does not "mandate only annual goodwill testing at a
fixed date," and no structural three-quarter deferral follows. The record contradicts C-04, C-13 and
C-33 drawn from the same page, and it is the one that must go.
C-44 │ HELD AS WRITTEN: Verified verbatim this run: goodwill impairment loss of $40,277 thousand,
written down to $nil, on testing performed as at 30 September 2024. The attack is that one Ballard
PDF carries three records — weighting, not truth.
C-45 │ REBUILT: Hayn & Hughes report that goodwill write-offs lag the economic impairment of
goodwill by an average of three to four years, in their own fundamentals-based construct, for a US
sample framed around SFAS 142. — Three narrowings, all conceded to the attackers. It is a mean, not a
median, and C-30's ten-year tail for a third of the sample makes the mean unrepresentative of the
median — arithmetically compatible with a median well below three years. Its "economic impairment"
construct is not the IAS 36 market-capitalisation indicator, and no conversion between them exists in
this ledger or in my retrieval. And it is US, pre/early-SFAS-142, and twenty years stale relative to
the estimand. It cannot underwrite a quarterly median threshold for Canadian IFRS issuers in
2022–2025, and I do not defend that use.
C-46 │ REBUILT: Ballard recognized a US$23,991 thousand goodwill impairment in 2023 relating to
the restructuring and effective closure of Ballard Motive Solutions in the UK, presented within loss
from discontinued operations. — Verified this run. A closure-driven write-off inside discontinued
operations is not an indicator-triggered CGU test outcome and does not belong in a
market-capitalisation-indicator-to-recognition series; the Inquisitor is right that the original
framing concealed the distinction.
C-47 │ HELD AS WRITTEN: Corroborated by the Null outside the ledger's single Corus PDF against
the FY2022 Q4 release. Restating C-25 with a different period descriptor is a duplication defect to
be collapsed in weighting.
C-48 │ CONCEDED: No source, no date, named in inadmissible-records.txt, and the four-observation
check it asserts — TSX close times shares outstanding against BCE's consolidated equity at each 2024
quarter-end — was never performed by anyone in this run, including me. It should carry no weight.
C-16 × C-31 (joint) │ REBUILT: For the TELUS digital experience business the impairment date is
entity-relative and CGU-relative: TELUS Corporation recognized nothing for that CGU in 2024 while
TELUS Digital recognized US$224M in Q2 2025, and IAS 36 permits testing at the level at which
goodwill is monitored. Any median over "issuers" must first fix the filer, and this ledger has no
such rule. — I concede this is a defect in the estimand, not in either record. What I add is that the
subsidiary-level figures are the ones that yield a measurable crossing (§0), so fixing the filer at
the listed subsidiary is not merely a convention but the choice that makes the estimand computable.
*(C-38 was unattacked by the Inquisitor and expressly not contested by the Null. It stands, and it is
the input to the arithmetic in §0.)*
2. The named rivals
I claim a rival only where I can point at evidence that separates it from the leading account.
RIVAL 1 — Recoverable amount lawfully exceeded carrying amount │ PARTLY DEFEATED
The Null's strongest card, and it is a real card: TELUS's 2024 critical audit matter and Lightspeed's
September 2022 test are both documented instances of the standard operating correctly and producing
nothing. I concede that lawful non-recognition happens and that this ledger contains two examples.
What discriminates is that **lawful non-recognition and untimely recognition are the same mechanism
under IAS 36, not competing explanations** — and the two securities-regulator bodies in this ledger's
own subject matter say so in terms. ESMA, in its comment letter to the IASB on the goodwill
Discussion Paper, states that "the so-called shielding effect are among the most significant
causes leading to the non-recognition (or non-timely recognition) of goodwill impairment"
(para. 26), and that "the result is that impairments are sometimes recognised too little, too late"
(para. 31). IOSCO's FR13/23 (Dec 2023) makes the same finding — "there may be circumstances where
impairment losses are not recognized adequately or on a timely basis" — and describes shielding as
the mechanism by which unrecognised internally generated goodwill masks impairment of acquired
goodwill. The IASB's own project papers record that "impairment losses on goodwill are sometimes
recognised too late" and that it is "not feasible to design a different impairment test … that is
significantly more effective."
That is decisive against the exculpatory form of Rival 1 — the reading on which a recoverable
amount above carrying amount means no economic impairment existed and the lag question is empty. Both
regulators say the opposite: recoverable amount lawfully exceeds carrying amount because headroom
shields the loss, and that is precisely why recognition is late.
A second, within-issuer discriminator, weaker: TELUS's recoverable amount exceeded carrying value at
the 2024 measurement date and the same CGU took US$224M six months later with no intervening
acquisition, while its listed vehicle sat at 56% of book. I flag the honest limit — the issuer
attributes the 2025 charge to genuinely changed assumptions (WACC up, growth down, pricing pressure),
which is consistent with Rival 1 — so I rest the discrimination on the regulator evidence, not on
this.
What I do not claim: none of this establishes a magnitude for the Canadian lag. Rival 1
survives as an explanation of any individual non-recognition. It no longer survives as a reason the
question is unaskable.
RIVAL 2 — Disclosure-date censoring (the zero-lag artifact) │ DEFEATED
Rival 2 says a disclosed lag of zero is uninformative because an issuer discloses the crossing in the
filing that records the charge. Three retrieved facts separate it from the leading account:
1. Lightspeed disclosed the crossing in a no-charge filing. The FY23-Q2 statements state the
crossing at 30 September 2022 and record no impairment. Rival 2 predicts this filing should be
silent. It is not.
2. Ballard's crossing is establishable without any issuer narrative: market capitalisation
≈US$1.10–1.11bn against audited equity of US$991,216 thousand at 31 December 2023 (no crossing),
against US$0.497bn and US$672,990 thousand at 31 December 2024 (crossing), with recognition on
testing at 30 September 2024.
3. TELUS Digital's likewise: US$2.35bn against owners' equity of US$2,037M at 31 December 2023
(no crossing); US$1.08bn against US$1,945M at 31 December 2024 (crossing); recognition
30 June 2025.
Neither (2) nor (3) is zero, and neither depends on what the issuer chose to narrate. The ledger's
zero-lag records are indeed artifacts — which is why I conceded C-12 and C-40 — but the general
claim that disclosed lags are zero by construction is false on this ledger's own issuers.
RIVAL 3 — Regime and jurisdiction non-transfer │ CONCEDED, against the ledger
I cannot separate this rival, and the evidence I gathered supports it. Every long-lag record in the
ledger is foreign and old: Hayn & Hughes (US, SFAS 142 framing, 2005), ESMA (Europe, 2011 crisis
balance sheets), André/Filip/Paugam (Europe and US, 2006–2015). The three Canadian in-window lags I
could bound are one quarter, at most two, and two-to-five. The imported durations do not transfer,
and where the Canadian record can actually be measured it runs materially shorter than the
literature the ledger imported to stand in for it.
RIVAL 4 — The median may be undefined │ PARTLY DEFEATED, narrowly
Within the population C-14 actually scopes — S&P/TSX Composite non-financial first-time impairers
2022–2025 — every member has a finite indicator-to-recognition lag by construction, so the median
exists over that population and the survival-function objection does not apply to it. The Null's
argument is sound about the unconditioned population of indicator-firms, which is a different
estimand.
That is the whole of my win here, and I concede the substantive half: the conditioning is exactly the
right-truncation the Null names. Firms that crossed and never impaired are excluded by construction,
their exclusion pulls the median down, and left-censored pre-2022 crossings pull it down again. The
statistic is computable and answers a narrower question than the framing implies — which, combined
with §0, is another reason the ">4 quarters" reading is unsupportable.
RIVAL 5 — Reporting-entity multiplicity │ CONCEDED
TELUS Corporation and TELUS Digital answer differently for the same business, Ballard's "first"
impairment flips on whether a discontinued-operations closure charge counts, and the ledger states no
rule for entity level, CGU level, window boundaries or "first." This is a defect in the estimand and
no evidence I retrieved repairs it. The most I can offer is the observation under C-16 × C-31: fixing
the filer at the listed subsidiary is the choice that makes the crossing measurable.
3. Where the evidence stops
It supports: that several Canadian issuers recognised goodwill impairments in 2022–2025; that
IAS 36 requires indicator assessment at the end of each reporting period and annual goodwill testing;
that an indicator compels estimation of recoverable amount, not a charge; that two securities
regulators and the IASB have found goodwill impairment under the impairment-only model is sometimes
recognised too little and too late, with shielding named as a principal cause; and that for the three
S&P/TSX-listed issuers in this ledger whose crossings I could bound, the indicator-to-recognition lag
was one quarter, at most two, and between two and five.
It does not support: a median indicator-to-recognition lag above four quarters for Canadian
issuers, in any form. That claim rested on C-02, C-12, C-34, C-40, C-43, C-17 and C-45's imported
mean — of which I have conceded six outright and narrowed the seventh out of load-bearing use. The
measurement the ledger never performed, now partly performed, points the other way: a central value
near one to two quarters, over a population of three, biased downward by the conditioning on
recognition and bracketed at annual rather than quarterly granularity.
What would still move it: quarter-end closing-price series for LSPD, BLDP, TIXT and CJR.B, which
would convert my annual brackets into exact crossing quarters; and disclosed recoverable amounts or
headroom for any non-impairing indicator-firm, which no filing I retrieved provides in figures.
Sources retrieved this run (all pre-cutoff)
- [IOSCO FR13/23, Recommendations on Accounting for Goodwill, December 2023](https://www.iosco.org/library/pubdocs/pdf/IOSCOPD753.pdf)
- [ESMA comment letter to the IASB on the goodwill Discussion Paper (ESMA32-61-413)](https://www.esma.europa.eu/sites/default/files/library/esma32-61-413_esmas_cl_to_iasb_dp_goodwill_and_impairment.pdf)
- [IFRS Foundation, Business Combinations—Disclosures, Goodwill and Impairment, GPF agenda paper 4 (March 2023)](https://www.ifrs.org/content/dam/ifrs/meetings/2023/march/gpf/ap4-business-combinations-disclosures-goodwill-and-impairment.pdf)
- [Grant Thornton International, IAS 36 — If and when to undertake an impairment review](https://www.grantthornton.global/en/insights/articles/IFRS-ias-36/ifrs-ias-36-If-and-when-to-undertake-an-impairment-review/)
- [Lightspeed Commerce, condensed interim consolidated financial statements, FY2023 Q2](https://s28.q4cdn.com/517092977/files/doc_financials/2023/q2/Financial-Statements-FY23-Q2-Final.pdf)
- [Ballard Power Systems, audited consolidated financial statements, year ended 31 December 2024](https://www.ballard.com/wp-content/uploads/2025/03/Q4-24-Financial-Statements-FINAL.pdf)
- [Ballard Power Systems, condensed interim financial statements, 30 June 2024 (SEC EX-99.1)](https://www.sec.gov/Archives/edgar/data/1453015/000162828024036632/bldp063024-ex991fs.htm)
- [Ballard Reports Q3 2024 Results](https://www.ballard.com/press-release/ballard-reports-q3-2024-results/)
- [Ballard Power Systems market capitalization history](https://companiesmarketcap.com/ballard-power-systems/marketcap/)
- [TELUS Digital, fourth quarter and full-year 2024 results](https://www.businesswire.com/news/home/20250213367125/en/TELUS-Digital-reports-fourth-quarter-and-full-year-2024-results-delivering-stable-operating-and-financial-performance-provides-outlook-for-2025)
- [TELUS Digital, second quarter 2025 results](https://www.nasdaq.com/press-release/telus-digital-reports-second-quarter-2025-results-incremental-improvement-revenue)
- [TELUS International (TIXT) market capitalization history](https://stockanalysis.com/stocks/tixt/market-cap/)
- [Lightspeed Commerce (LSPD) market capitalization history](https://stockanalysis.com/stocks/lspd/market-cap/)
- [Corus Entertainment fiscal 2020 third quarter results](https://www.prnewswire.com/news-releases/corus-entertainment-announces-fiscal-2020-third-quarter-results-301084272.html)
- [BCE 2024 Q4 and full-year results](https://www.prnewswire.com/news-releases/bce-reports-2024-q4-and-full-year-results-announces-2025-financial-targets-302369604.html)
- [Kroll goodwill impairment publications hub (Canadian editions 2012, 2013, 2014)](https://www.kroll.com/en/insights/publications/goodwill-impairment)
- [Duff & Phelps publishes 2013 Canadian Goodwill Impairment Study](https://www.kroll.com/en/newsroom/duff-and-phelps-publishes-2013-canadian-goodwill-impairment-study)
Quarter-end closing-price series for LSPD, BLDP and TIXT could not be retrieved (403, paywall, or no
2022–2024 coverage across five providers). Every crossing bound above is stated at the granularity
the retrieved data actually support, and no verdict rests on a price I did not obtain.
CONCESSION: MOVED — I performed the arithmetic both attackers said the ledger names and never runs, and it moved me against the ledger I was defending. Lightspeed's own FY23-Q2 goodwill note, cited three times in this ledger and never quoted, says the market-capitalisation crossing was live at 30 September 2022 and produced a test that recognised nothing; C-12 and C-40 are false as written and I concede them outright rather than narrow them. C-02, the record carrying the whole median-above-four-quarters conclusion, converts two annual prevalence statistics into a quarterly duration statistic that neither source reports and that a Q4 indicator with a Q1 charge already refutes — conceded. C-43 is falsified by the page it cites; C-34 dies three times over, decisively on my own retrieval that Corus wrote off $673.0M of Television goodwill in the quarter ended 31 May 2020, which removes it from any first-impairment population; C-17 is load-bearing on no source and about the shape of questions rather than about issuers; C-48 was never checked by anyone. Seven concessions, five of them LOAD-BEARING, including the spine. What I did rebuild, I rebuilt narrower: C-22 survives its refutation only because the Canadian Goodwill Impairment Study stopped at a 2014 edition covering 2007–2013 with no quarter-end or first-impairment construct, and C-14 survives only because I could finally count the population it presupposed — three. Against the Null I win two rivals and only two. Rival 2 falls to evidence, not assertion: Lightspeed disclosed a crossing in a filing that recorded no charge, and Ballard's and TELUS Digital's crossings are establishable from audited equity against exchange data with no issuer narrative involved — none of those lags is zero. Rival 1's exculpatory form falls to ESMA's and IOSCO's own finding that shielding is "among the most significant causes leading to the non-recognition (or non-timely recognition) of goodwill impairment," which makes lawful non-recognition the mechanism of lateness rather than an alternative to it. Rivals 3 and 5 I concede, and Rival 4 I split. But the number that came out of the measurement is the one that matters: one quarter for Lightspeed, at most two for Ballard, two to five for TELUS Digital. The ledger asked whether the median exceeds four quarters, and the best Canadian evidence I could assemble says it does not.
ROUND 2 — 04-advocate/advocate-r2.md
ADVOCATE — ROUND 2 (cutoff 2026-08-22 · MODE LIVE)
RETRIEVAL CONDITIONS (disclosed, because it bears on how much weight this document earns)
WebFetch returned session limit on every call this run, and direct curl to every host tried
(osc.ca, sec.gov, federalreserve.gov, journals.sagepub.com) was refused by the egress proxy with
CONNECT tunnel failed, response 403. My only working live channel was WebSearch, which in this
build returns titles and URLs and no page text. So my independent verification this round is:
(a) existence-and-title confirmation via WebSearch, and (b) cross-checking the Inquisitor's and the
Null's verbatim quotes against each other — two adversarial agents who retrieved separately.
Where both attackers quote the same string, I treat it as corroborated and I do not relitigate it.
Where they disagree (C-26, C-45) I say which side I take and why. Where I obtained genuinely
independent confirmation I say so and name it. Three such confirmations were obtained:
- Lightspeed US$748.7M — headline, Medicine Hat News, 2023-02-02: "Lightspeed Commerce
reports US$814.8M Q3 loss on US$748.7M goodwill impairment charge." Independent of both attackers.
- C-08's outcome — Kroll, "Goodwill Impairment Trends Through H1 2022" (Valuation Insights Q3
2022) confirmed live on kroll.com; BVR headline "Goodwill impairments YTD 2022 are up, per Kroll
analysis." The April 2022 forecast direction was borne out.
- Named gaps confirmed to exist: TELUS Digital's own FY2024 Form 20-F (TIXT_Q4_2024_20-F.pdf);
Chen & Zimmermann FEDS 2021-037; Chen, "Most claimed statistical findings in cross-sectional
return predictability are likely true" (arXiv 2206.15365); JKP published Journal of Finance
version; Hayn & Hughes published SAGE version; BCE 2024 consolidated financial statements on
SEDAR+; Corus fiscal 2023 Q1, Q2, Q3 and Q4/year-end releases; Ballard Q3 2024 release.
I could not re-verify page text myself. Discount me accordingly on any line where I am relying on
attacker quotation rather than on my own retrieval — which is most lines. I would rather the
Adjudicator know that than have it inferred.
PER-CLAIM
C-04 │ REBUILT: *IAS 36 ¶9 requires an entity to assess at the end of each reporting period
whether any indication of impairment exists, and, where an indication exists, to estimate the
recoverable amount of the asset or CGU. Separately and in addition, ¶10 requires goodwill acquired
in a business combination and indefinite-life intangibles to be tested for impairment annually
irrespective of whether any indication exists; ¶96 permits that annual test to be performed at
any time in the annual period provided it is performed at the same time each year, and ¶99 permits
a preceding period's detailed calculation to be carried forward where stated criteria are met.
¶12(d) names "the carrying amount of the net assets of the entity is more than its market
capitalisation" as an indicator. Recognition follows only where recoverable amount falls below
carrying amount.* — The original stated ¶9 and stopped there, which read the standard as
indicator-gated. Both attackers caught it and they are right; in a goodwill ledger the annual-test
limb is the load-bearing half, and the record's own cited Grant Thornton page carries it. I add the
consequence rather than hide it: **an interval between a ¶12(d) crossing and a recognised loss is
not by itself evidence that a test was deferred.** The annual test runs regardless, and a test that
runs on schedule and concludes recoverable amount exceeds carrying amount produces the identical
observable. Lightspeed is the live instance — its release states December 31 is its fixed annual
test date. The original falsifier ("a provision permitting deferral") guarded a door nobody used and
is withdrawn; the replacement falsifier is: *an IAS 36 or IAS 34 provision under which a ¶12(d)
indicator at an interim date obliges recognition rather than testing.* Tier should move E2 → E0 and
the source to the standard's own text, which is freely retrievable; a LOAD-BEARING claim about the
words of a standard should not rest on an E2 advisory summary filed as "Inference from."
C-07 │ REBUILT: *Corus's fiscal 2023 annual MD&A states that the Company's share price "has
continued to decline meaningfully from August 31, 2021, which resulted in the Company's carrying
value being greater than its market enterprise value at August 31, 2022, May 31, 2023, and August 31,
2023." Those are the period-ends the MD&A names; the document attributes all three to one continuing
share-price decline and does not state the condition's status at November 30 2022 or February 28
2023. The measure is carrying value against market enterprise value, which includes net debt, not
net assets against market capitalisation. The same document records $690.0M of fiscal 2023
impairments in the Television CGU — $590.0M in Q3 (quarter ended May 31 2023, of which $295.2M
goodwill, $219.8M broadcast licences, $175.0M brands) and $100.0M in Q4.* — Three changes, all
narrowing or additive. First, "the list is a disclosure set, not a crossing series" is withdrawn:
that was the claimant's characterisation and the document's own causal wording ("has continued to
decline") leans the other way. Second, "is silent on" becomes "does not state the condition's status
at," because the annual MD&A does not cover those quarter-ends and silence in a document that never
addressed them is not evidence; the Inquisitor's retrieval of the Q2 fiscal 2023 release (which
carries no carrying-value-versus-market language at all) is the evidence that should have been
produced and is not in the ledger. Third, the recognition sitting in the cited document is now
booked. That third point is not a favour to the ledger — it is what destroys C-14.
C-08 │ HELD AS WRITTEN: Survives both attacks. The Inquisitor's PROVENANCE attack (not
independent of C-22, both tracing to Kroll) is a ledger-level source-concentration defect, not a
defect in the claim: the claim is that a named Kroll managing director made a named forecast in a
named article on a named date, and Kroll being the origin of a Kroll forecast is not circularity.
The SELECTION attack (the ledger books the forecast and never the outcome) does not contradict the
claim either — and on my own retrieval the outcome confirms the forecast's direction: Kroll's
"Goodwill Impairment Trends Through H1 2022" and BVR's coverage both report 2022 year-to-date
impairments up. The Null does not contest it. The record over-reaches nowhere; it claims a forecast
and evidences a forecast.
C-11 │ REBUILT: *65 percent of the 452 characteristic portfolios in Hou, Xue and Zhang's own
data library — as HXZ construct them, with microcaps mitigated via NYSE breakpoints and
value-weighted returns — have absolute t-values below 1.96.* — The number is right and both attackers
verified it verbatim against the cited October 31 2018 version. What changes is the **denominator's
meaning**: "the 452 anomalies" invites the reading "65 percent of published anomalies fail," which is
not what the statistic says. Chen & Zimmermann (FEDS 2021-037) partition the library and report that
of the 161 characteristics clearly significant in their originating papers, 98 percent have |t| above
1.96 on their independent reproduction, the residual being characteristics that were insignificant in
the original papers or HXZ-created modifications of originals. Both figures can be true at once
because they are computed over different sets. The ledger should hold C-11 and Chen & Zimmermann,
not C-11 alone. Provenance concession: the cited page is an SSRN abstract landing page with a
documented revision history (2017-05-03 → 2017-07-27 → 2018-10-31) across which the headline numbers
moved; it should be re-cited to Hou, Xue & Zhang (2020), Review of Financial Studies.
C-14 │ CONCEDED: This cannot be rebuilt at any width, and it is the single most useful thing I
can tell the Adjudicator. Four independent reasons, any one of them sufficient. (1) No source.
The record's own line reads "model prior, no external source," self-dated to the cutoff — but what it
carries is not a prior, it is a factual population count and a named enumeration of three issuers,
i.e. a retrievable fact filed as an unauditable intuition. (2) Falsified from inside the ledger.
Corus is a fourth issuer with both endpoints in a single document the ledger already holds: C-07's
own source names the crossings (August 31 2022, May 31 2023, August 31 2023) and records $295.2M of
goodwill within $590.0M of Q3 fiscal 2023 charges. "Three" is wrong on the ledger's own evidence.
(3) "First-time impairer" is undefined and outcome-determinative at this n. Ballard impaired
goodwill in 2023 (C-46, US$23,991 thousand) and in 2024 (C-28's note, $40,277 thousand of corporate
goodwill written to nil on tests performed at September 30 2024). Whether Ballard is a 2024
first-timer turns on a definition — continuing operations only? segment level? excluding discontinued
operations? — that the ledger never states and that was not fixed before the population was
assembled. (4) The Null's rival is undefeated and I have nothing that discriminates. Retrieval-by-
outcome: these issuers were located because they disclosed both a crossing and a subsequent
impairment. Defeating that rival requires a denominator — the count of S&P/TSX Composite non-financial
issuers that crossed and did not subsequently impair. The ledger contains no such count, this
run's retrieval produced none, and my retrieval channel this round could not build one. C-27 and
Corus each supply a within-issuer instance of a crossing that produced no contemporaneous recognition,
which shows the event-level filter is not absolute; that is not a denominator and I will not dress
it up as one. Separately, the stated falsifier is aimed at the wrong risk — it tests whether the
universe exceeds one hundred issuers, while the live risk is the retrieved population, and a
population of four defeats "three" without ever approaching a hundred. A median over an effective n
of one is not a measurement. The record should be struck, not narrowed.
C-15 │ REBUILT: *Lightspeed's Q3 FY2023 release states: "The Company's net assets exceeded its
market capitalization as at December 31, 2022 which was an impairment trigger for the Company." The
same release records a US$748.7 million non-cash goodwill impairment charge at that same date, and
states that "The Company conducts its annual goodwill impairment test every December 31" — so under
IAS 36 ¶96 the December 31 2022 test would have occurred irrespective of the trigger. December 31
2022 was not the first period-end at which the condition held: the Q2 FY2023 interim statements
(C-27's source) record the same condition at September 30 2022 with a completed fair-value-less-
costs-of-disposal test showing no impairment. Lightspeed's crossing-to-recognition interval is
therefore one quarter, with both endpoints documented.* — The original hedged on an open question
that its own ledger closes, and read one document only for its trigger sentence while the recognition
sat in the same paragraph. I have independent confirmation of the $748.7M figure (Medicine Hat News
headline, 2023-02-02) and both attackers confirm it against the FY2023 audited statements
("Goodwill impairment 16 748,712"). I book the consequence rather than bury it: **this is the only
crossing-to-recognition interval anywhere in the subset with both endpoints documented, and it is one
quarter against C-45's imported anchor of three to four years — roughly an order of magnitude.** That
cuts against the ledger's thesis. It is what the evidence says.
C-16 │ REBUILT — and the Null's rival stands undefeated: *(i) TELUS Corporation's 2024 audited
consolidated statements record that "Management determined that the recoverable amount of TELUS
digital experience exceeded its carrying value as of the measurement date and, therefore, no
impairment was recognized," on a fair value less costs of disposal calculation using discounted cash
flow projections, a judgement the auditor designated a critical audit matter; the document quantifies
no headroom. (ii) Separately and from other documents, the listed subsidiary TELUS Digital reported
owners' equity of US$1,945M at December 31 2024 and recognised a US$224M goodwill impairment for the
quarter ended June 30 2025.* — The juxtaposition is withdrawn entirely. It compared a CGU carrying
amount inside TELUS Corporation's CAD consolidated statements against a separate reporting entity's
USD consolidated owners' equity and its quoted price: different entities, different carrying bases,
different measurement objects. Under IAS 36 ¶12(d), net assets above market capitalisation is an
indicator obliging a test, not a measure of recoverable amount — the original claim silently
used it as the latter. Provenance concession: sources = 1, and that source carries at most half the
claim; the US$1.08bn market capitalisation is not in the cited document and the Null could not confirm
it from there. That figure should be dropped or separately sourced. On the Null's rival I concede.
The two live explanations — (a) management's DCF was optimistic and recognition was late, versus
(b) a thin minority float in a controlled subsidiary traded below the entity's fair value and the
2024 test was correct — are both consistent with every fact in the ledger, and restating (a) more
confidently would prove the Null's point rather than answer it. Two facts lean toward (b) without
settling it: TELUS was buying additional TELUS Digital shares in October 2024, and under IFRS 13 a
depressed minority-float quotation is not the fair value of a controlled CGU. Two things would
actually discriminate and neither is in the ledger: **disclosed headroom for the TELUS digital
experience CGU at December 31 2024, and the per-share consideration TELUS paid to acquire the
remaining TELUS Digital shares** measured against the float price at the same date — a control
premium materially above float would favour (b), a price at or below it would favour (a). Also absent:
TELUS Digital's own FY2024 Form 20-F, which carries a second board's and a second auditor's
subsidiary-level goodwill conclusion for the identical year. I confirmed that document exists; I could
not read it. The original falsifier tested only the uncontested half and is replaced by: *disclosed
headroom for the TELUS digital experience CGU at December 31 2024 sufficient to absorb the subsequent
US$224M charge.*
C-19 │ REBUILT (count conceded wrong): *CSA Staff Notice 51-364 names impairment of non-financial
assets among a non-exhaustive list of areas that may be impacted by the current economic environment
("Some areas that may be impacted by the current economic environment include …"). — *The
count of thirteen is conceded as false.** Two agents counted the same sentence independently and both
got twelve; on the only alternative parse (splitting "known trends, events and uncertainties" into
three) the count is fourteen. No parse yields thirteen. Because the sentence says "include," the list
is expressly open and no count is a finding of the notice at all — the precision was manufactured,
and dropping it makes the record both true and stronger. Method concession, which matters more than
the number: the stated falsifier ("the notice omitting impairment of non-financial assets") was always
going to pass and could never see the one thing in the record that was wrong. That is the pattern the
Null identifies across C-04, C-16 and C-21, and it is real. Replacement falsifier: *the notice's
enumerated list omits impairment of non-financial assets, or the sentence is exhaustive rather than
illustrative.*
C-21 │ CONCEDED: Verbatim-identical claim text to C-15, on the identical URL, entered as a second
record differing only in falsifier wording, and both marked LOAD-BEARING. A document does not
corroborate itself by being cited twice. C-21 supplies no evidence C-15 does not already supply, and
the ledger's evidentiary base for this proposition is one sentence in one press release. The record
should be merged into C-15 and struck, and the ledger's LOAD-BEARING count for this proposition
reduced by one. Secondary concession: even as a single record it cites the weaker of two documents
the ledger already holds — the FY2023 audited statements (C-35's source) carry a fuller trigger
narrative and the outcome.
C-22 │ REBUILT: *Retrieval in this run — across Kroll's goodwill-impairment publications hub,
IOSCO, Reuters and issuer filings — did not surface any dataset pairing S&P/TSX Composite
non-financial issuers' first goodwill impairments with net-assets-above-market-capitalization
crossings at quarter-end granularity for 2022–2025. The nearest precedent construction for this
population is Duff & Phelps/Kroll's Canadian Goodwill Impairment Study, published in 2012, 2013
and 2014 editions (the 2012 edition jointly with FEI Canada) and discontinued after 2014.* — The
universal negative ("No published dataset pairs…") is withdrawn and replaced by a retrieval-bounded
statement, because a directory page listing study editions cannot evidence what datasets exist
anywhere. I credit the Null here: it genuinely tried to falsify this and could not, and Kroll's
Canadian series does stop at 2014 — I confirmed the publications hub lists Canadian editions
2012–2014 only. But the Inquisitor is also right that the named search space excludes every venue
where such a pairing would most plausibly live: **Audit Analytics, S&P Capital IQ, Compustat/WRDS
extracts, and SEDAR+ full-text search.** Rebuilt this way the record still does the ledger's work —
it establishes that the ledger did not inherit an external base — but it can no longer be read as
establishing that no such base could be built. Read together with C-14 as I have conceded it, C-22
now says, in the ledger's own voice: *the load-bearing empirical base does not exist here, and this
run did not build one.* That is the correct thing for it to say.
C-26 │ REBUILT (I side with the Inquisitor against the Null here): *Jensen, Kelly & Pedersen,
under a Bayesian hierarchical model of factor replication with shrinkage across factors, report that
82.4 percent of factors replicate on the criterion that the posterior z-score equals or exceeds 1.96
— equivalently, that the posterior probability of a non-positive alpha falls below 2.5 percent — and
separately report that the majority of factors work out of sample in a new data set covering 93
countries.* — The two attackers split on this record: the Null conceded it as "accurate to the paper,"
the Inquisitor called it a hard kill. The Inquisitor is right, on the merits of the statistics rather
than on authority. **"Non-zero posterior means" is not the paper's replication criterion and is
near-vacuous as stated**: under a hierarchical model shrinking toward a non-zero grand mean, posterior
means are essentially never exactly zero, so "the majority have non-zero posterior means" would be
true of almost any such model and would carry no information about replication. The paper's criterion
is a threshold on the posterior z-score, which is a substantive test. Second change: the original
welded abstract findings (1) and (3) into one proposition — "the majority … have non-zero posterior
means, holding out of sample across 93 countries" — which asserts that the same majority holds
country by country. The abstract asserts no such conjunction, and I have unwelded them. Provenance
concession: the record dates the source 2021-03-05 but SSRN records last revision 2022-03-07 and the
paper was published in the Journal of Finance 78(5), 2023 (doi 10.1111/jofi.13249); the misstated
criterion is precisely the error an abstract-only citation produces.
C-27 │ HELD AS WRITTEN: Every clause verified verbatim by both attackers independently — the
crossing at September 30 2022, the triggered test, the fair value less costs of disposal model, and
"no impairment of goodwill as at September 30, 2022." The Null does not contest it; the Inquisitor's
PROVENANCE attack concerns the source date field (2022-09-30 is the balance-sheet date; the
statements were approved for issue November 2 2022), which is a metadata correction to make and not a
defect in the claim — though in a ledger that runs on cutoff arithmetic the correction should be made.
The SELECTION attack names a real gap (Lightspeed's Q1 FY2023 statements, which would establish
whether September 30 2022 was itself the first crossing) but names it as ASSERTED, and an unretrieved
document does not wound a claim about a retrieved one. I note without prompting that this record is
the strongest single record in the subset for the opposing case: it documents a ¶12(d) crossing
that produced a completed test and no recognition, which is the missing-cell structure the Null builds
its C-14 rival on. It holds because it is true, not because it helps.
C-28 │ HELD AS WRITTEN: Verified by both attackers — three indicators, correct count, correct
content, correct ordering. The PROVENANCE attack (same PDF and same note as C-46, printed as
"sources: 1" on each) is a real source-independence defect of the ledger's bookkeeping, not of the
claim: the ledger's Ballard base is one filing read twice and should be recorded as such. The
SELECTION attack is also correct and also does not contradict the claim — the same note records the
recognition those indicators produced ($40,277 thousand of corporate goodwill written down to nil on
impairment tests performed at September 30 2024), and that recognition should be entered as its own
record, because it is exactly the crossing-to-recognition endpoint C-14 claimed to be counting and
did not count. A claim is not falsified by a fact it omits that is consistent with it.
C-31 │ REBUILT: *TELUS Digital recognised a US$224 million non-cash goodwill impairment charge for
the quarter ended June 30, 2025; the release attributes the decline in recoverable amount to "higher
weighted average cost of capital, lower perpetual growth rate and lower cash flow forecasts arising
from pricing pressure on margins," and does not discuss impairment indicators of any kind.* — The
final clause "the release names no market-capitalisation trigger" is withdrawn as an argument from
silence in a promotional document: a press release's silence on ¶12(d) is not the issuer's position on
¶12(d). The issuer's actual indicator assessment is in TELUS Digital's Q2 2025 interim consolidated
statements furnished on Form 6-K, which the ledger does not hold and which the record's own falsifier
is nevertheless written against. Provenance: the cited page is Nasdaq's syndicated copy of a Business
Wire release — a redistribution channel, and it should be re-cited to the issuer's own filing. **And I
concede the Null's structural point in full**, because it is the sharpest thing either attacker wrote:
this is the one clean recognition event in the subset, and the issuer attributes it to **discount rates
and forecast revisions** — a common-cause structure in which rates and outlook move market price and
recoverable amount together on different lags — not to a known fact management sat on. Nothing in the
ledger separates common cause from lagged recognition, and I have nothing that does. Restating the
delay explanation more confidently would demonstrate the Null's point rather than defeat it.
C-35 │ HELD AS WRITTEN: Verified by both — "Goodwill impairment 16 —" in the FY2022 comparative
column. The claim asserts exactly one thing, that no goodwill impairment loss appears in the
comparative fiscal year ended March 31 2022, and that is precisely what the statements show. The
Inquisitor's INFERENCE attack lands on the falsifier's phrasing, not the claim's: the claim does
not assert that no crossing occurred or that no test was performed in FY2022, and should not be read
as doing so. The SELECTION attack is correct as a ledger criticism — the same statements carry the
$748,712 thousand FY2023 loss and the fullest trigger narrative in the ledger, and the ledger declined
to read the affirmative half of a document it already holds — but that is now booked at C-15, where I
have moved it.
C-36 │ REBUILT: *BCE's release states non-cash asset impairment charges totalling $2,190 million
for 2024, mainly related to Bell Media's TV and radio properties. The release states no goodwill
component and decomposes the charge no further.* — The amount and attribution are verified to the
digit by both attackers and I do not touch them. What I concede is **admissibility, which is the
Null's L-2 and it is right**: Bell Media's impaired assets span goodwill and indefinite-life
intangibles (broadcast licences are tested under IAS 36 alongside goodwill and are not goodwill),
the release splits neither, and the word "goodwill" does not appear. **In a goodwill-impairment ledger
this record supports no goodwill observation.** The document that would fix it — BCE's 2024 annual
audited consolidated financial statements with the impairment note, CGU identification and the
goodwill-versus-indefinite-life split, filed on SEDAR+ — is absent from the ledger, and I confirmed it
exists. Until it is held, C-36 should be reclassified as context, not as an impairment datum.
C-37 │ REBUILT: *The Canadian Press reported the figure BCE published the same day, rounded from
$2,113 million to $2.11 billion, mainly related to Bell Media properties. The record is wire coverage
of the release at C-41 and carries no evidentiary weight independent of it; it should not count
toward the ledger's source totals.* — The record already said the honest half ("not as independent
verification of it"), and I credit whoever wrote that. But the Inquisitor's point survives the
candour: a record of zero independent evidentiary value was still entered and still counted. The
article's genuinely independent content concerns the CRTC fibre-access decision, the share price and
analyst commentary — none of it touching the impairment. Making the zero-weight explicit in the claim
text is the fix.
C-41 │ REBUILT: *BCE's Q3 2024 results news release states non-cash asset impairment charges
totalling $2,113 million, "mainly related to Bell Media's TV and radio properties," for the quarter
ended September 30, 2024.* — "$2,113 million against Bell Media" asserted a full segment
attribution the release does not make; "mainly related to" leaves an unquantified remainder outside
those properties. As with C-36, the release states no goodwill component and makes no allocation. The
falsifier is written against BCE's Q3 2024 interim financial statements, which are the primary
document and are cited nowhere in the ledger — a falsifier aimed at a document the ledger cannot open
is not operative, and it should be rewritten against the cited release or the interim statements
should be retrieved.
C-45 │ REBUILT (SFAS 142 framing conceded reversed): *Hayn & Hughes report that goodwill
write-offs lag behind the economic impairment of goodwill — measured on the authors' own
fundamentals-based construct — by an average of three to four years, with the delay extending to about
ten years for roughly a third of companies, in a US sample composed mainly of goodwill generated
before SFAS 142; the authors infer generalisability to the post-142 regime from sensitivity tests
on a smaller sample of write-offs made upon adoption.* — The two attackers split; the Inquisitor is
right and I concede the point. The published version's caveat is explicit that most analyses are
conducted on pre-SFAS-142 goodwill and that generalisability is an argued inference, and that
sentence is absent from the 2005 SSRN abstract the ledger cites — so the record's "framed around SFAS
142" reverses the authors' own qualification, and the reversal was **caused by citing the preprint
abstract instead of the 2006 published paper (doi 10.1177/0148558X0602100303). Re-cite it. And I
concede the transport problem, which is the larger issue and which neither the original claim nor its
falsifier touches**: a three-to-four-year lag measured under a US amortisation-era regime with a
two-step reporting-unit test, on a constructed proxy for "economic impairment," is not transportable
to a one-step IAS 36 CGU test with a mandatory annual cycle in Canada 2022–2025 without a bridging
record, and the ledger holds none. The post-142 literature that tests exactly that generalisation —
Ramanna & Watts (2012), Li & Sloan (2017) — is absent. Meanwhile the one interval the ledger actually
measures is Lightspeed's one quarter. **This anchor should not be carried as a benchmark for the
Canadian window; at most it establishes that lagged recognition has been documented somewhere, under
a different regime.**
C-46 │ REBUILT: *Ballard's FY2024 audited consolidated statements record, in the 2023 comparative,
a US$23,991 thousand goodwill impairment relating to the restructuring and effective closure of
Ballard Motive Solutions in the UK, presented in those FY2024 statements within loss from
discontinued operations.* — The amount is exact to the thousand and both attackers confirm it. The
narrowing is on the presentation clause: the source is the comparative column of a filing dated
2025-03-12, fifteen months after the year it describes, and comparatives are restated for
reclassification — so this document establishes how the item is presented in the FY2024 statements
and cannot establish how it was presented as originally reported. The contemporaneous primary,
Ballard's FY2023 audited consolidated financial statements, is not in the ledger. Provenance
concession, same as C-28: this is the same PDF and the same note as C-28, and "sources: 1" printed on
each conceals that Ballard's entire evidentiary base here is one filing read twice. Substantive
consequence, already booked at C-14: Ballard impaired goodwill in both 2023 and 2024, so its
status as a "first-time impairer" in 2024 is indeterminate on the ledger's own records.
WHERE I DEFEATED THE NULL, AND WHERE I DID NOT
Defeated, narrowly, at L-3. The Null says C-11 (HXZ, 65 percent fail) and C-26 (JKP, majority
replicate) support incompatible propositions with nothing in the ledger adjudicating between them, so
whatever prior is imported comes from a live dispute pointing in no direction. Chen & Zimmermann
(FEDS 2021-037) discriminates, and it does so by identifying the source of the disagreement rather
than by restating either side: the two results are computed over differently composed sets. HXZ's
452 includes characteristics that were never significant in their originating papers plus HXZ-created
modifications of originals; restricted to the 161 clearly significant in the originals, 98 percent
clear |t| ≥ 1.96. That is not "the leading explanation said louder" — it is a partition that makes
both numbers simultaneously true and tells you which question each answers. What it does not do is
settle the deeper methodological dispute (t-hurdles with microcap mitigation versus Bayesian shrinkage
across factors), which remains live, and I do not claim it does.
Not defeated at L-1. The Null's naming of crossing is correct and I have no answer. Corus is
tested on market enterprise value (which includes net debt), Lightspeed on **market
capitalisation, and the TELUS comparison invokes owners' equity against market capitalisation**
across two entities. For a levered broadcaster, enterprise-value and market-capitalisation tests can
cross in different quarters or one can cross and the other never. Any lag series mixing them measures
intervals between events of different kinds. Nothing in the ledger fixes a single definition, and I
cannot supply one after the fact without choosing the answer.
Not defeated at L-2. Conceded in full at C-36, C-37 and C-41. Three of twenty records are one
issuer, one event, one primary document, under a term — impairment — that has shifted from goodwill
to non-cash asset impairment generally.
**Not defeated on C-14's retrieval-by-outcome rival, on C-16's minority-float rival, or on C-31's
common-cause rival.** Each needs a specific discriminating fact — a denominator of non-impairing
crossers; disclosed CGU headroom or the take-private consideration against the float price; something
separating rate-and-forecast-driven declines from withheld recognition. None is in the ledger, none
was in this run's retrieval, and none was reachable on my channel this round. I name them rather than
paper over them.
CONCESSION: MOVED — substantially, and at the places that decide the question rather than at the margins. I concede C-14 outright and recommend it be struck, not narrowed: it has no external source, its enumeration is falsified from inside the ledger by C-07's Corus, its "first-time impairer" test is undefined and outcome-determinative at n=3 given Ballard's impairments in both 2023 and 2024, and its falsifier tests the universe while the live risk is the retrieved population. I concede C-21 as a duplicate of C-15 that should be merged and struck. I concede that "thirteen" at C-19 is simply wrong and that no count is a finding of a list the notice calls illustrative. I concede that C-45 reverses the authors' own SFAS 142 caveat because it cites the preprint abstract, and further that a lag measured in a US amortisation-era regime is not transportable into IAS 36 Canada 2022–2025 without a bridging record the ledger does not hold. I concede that C-26 substitutes a near-vacuous criterion for a posterior z-score threshold, siding with the Inquisitor against the Null's concession. I concede that the entire BCE block cannot support a goodwill observation. Most importantly I concede the three rivals the Null named and I could not discriminate: retrieval-by-outcome at C-14, minority-float-versus-optimistic-DCF at C-16, and common-cause-versus-delay at C-31 — and I have named the specific fact that would settle each rather than arguing around them. What survives is narrower and, I think, sturdier: C-08, C-27, C-28 and C-35 hold as written and the attacks on them are ledger-bookkeeping defects, not claim defects; C-04 rebuilds into a correct statement of both limbs of IAS 36 that concedes its own consequence, that a crossing-to-recognition gap is not evidence of deferral; C-07, C-11, C-22, C-36, C-41 and C-46 rebuild narrower and stay true; and C-15 rebuilds into the one thing the ledger actually measured — Lightspeed's crossing at September 30 2022 with a completed test and no impairment, its US$748.7M recognition at December 31 2022, an interval of one quarter. That interval cuts against the thesis the ledger was assembled to support, and booking it is the honest output. Where the evidence stops: there is no denominator, no agreed definition of crossing, no second measured interval, and no adjudication between common cause and delay. On the facts the ledger is largely sound. On the lag thesis it does not reach, and after this round it should stop claiming to.
ROUND 3 — 04-advocate/advocate-r3.md
ADVOCATE — ROUND 3 (LIVE, cutoff 2026-08-22)
Retrieval performed this round against: BCE's Q3 2024 consolidated financial statements (SEC), Ballard's FY2024 statements note 11 and its 2023 Annual Report MD&A, Corus's fiscal 2023 annual MD&A, Lightspeed's Q2 FY2023 interim statements, IAS 36 consolidated text, the Grant Thornton IAS 36 article, CFO Dive, FEDS 2021-037, and the published Journal of Finance (2023) version of Jensen/Kelly/Pedersen.
I concede more than I rebuild in substance. Three of the four records the Null contests I concede in their operative half, and on the fourth (C-15) the evidence I found discriminates against the ledger. The Inquisitor's BCE attack is correct and dissolves the shared conclusion of three records. Where I hold, I hold on verified text and say which attack it survives.
C-04 │ REBUILT: IAS 36 ¶9 requires an entity to assess at the end of each reporting period whether any indication of impairment exists, and ¶12(d) names the carrying amount of the entity's net assets exceeding its market capitalisation among the indications to be considered; where an indication exists the entity shall estimate recoverable amount, but that obligation is qualified by ¶15, which — "[a]part from when the requirements in paragraph 10 apply" — provides that materiality applies in identifying whether recoverable amount needs to be estimated, that an entity "need not re-estimate the asset's recoverable amount if no events have occurred that would eliminate" a previously demonstrated significant excess, and that previous analysis may show recoverable amount insensitive to one or more ¶12 indications; separately and in addition, ¶10 requires goodwill acquired in a business combination and indefinite-life intangibles to be tested annually irrespective of indication, ¶96 permits that annual test at any single date provided it is performed at the same time each year, and ¶99 permits a preceding period's detailed calculation to be carried forward on stated criteria; recognition follows only where recoverable amount falls below carrying amount; an interval between a ¶12(d) crossing and a recognised loss is therefore not dispositive (weakened from "not by itself evidence", per the Null's exact quibble) that a test was deferred — sourced to the IAS 36 consolidated text, all six paragraphs verified verbatim this run, because the Inquisitor's PROVENANCE attack is half right and I concede that half: the Grant Thornton article carries the substance of ¶10, ¶12(d) and ¶96 (I re-fetched it and the Null's verification of that substance is correct against mine) but prints no paragraph number other than ¶96, and says nothing at all about ¶99, so it cannot bear a six-paragraph reading and is demoted to corroboration of three propositions; on the SELECTION attack I do not concede — ¶15 does not satisfy the record's falsifier, which asks for a provision permitting deferral of testing to a later period, whereas ¶15 dispenses with the estimate outright on stated conditions rather than postponing it, and the Inquisitor's own reading therefore strengthens the record's conclusion (if ¶15 can relieve the estimate entirely, an observed interval is weaker evidence of deferral, not stronger), which is why I have written ¶15 into the claim rather than around it.
C-07 │ REBUILT: Corus's fiscal 2023 annual MD&A states verbatim that the Company's share price "has continued to decline meaningfully from August 31, 2021, which resulted in the Company's carrying value being greater than its market enterprise value at August 31, 2022, May 31, 2023, and August 31, 2023" — a carrying-value-against-market-enterprise-value measure including net debt, not IAS 36 ¶12(d)'s net-assets-against-market-capitalisation measure, and the document states nothing about November 30 2022 or February 28 2023; the same document records $690.0 million of fiscal 2023 Television-CGU impairments for the year ended August 31, 2023, decomposed at the year level only — goodwill $295.2M, broadcast licences $219.8M, brands and trade marks $175.0M — and separately splits that total $590.0M to the quarter ended May 31 2023 and $100.0M to the quarter ended August 31 2023 with no component allocation to either quarter; and the same document records that "[f]or the year ended August 31, 2022 … a non-cash goodwill impairment charge of $350.0 million was recorded in the Television CGU", attributed to actual results falling short of previous estimates and a less robust outlook rather than to the enterprise-value crossing — I concede the decomposition outright: both attackers are right that $295.2M + $219.8M + $175.0M = $690.0M is the fiscal-year total and cannot be the composition of a $590.0M quarter, I verified it in the cited PDF, and since goodwill is the only component this ledger needs dated, the record as written mis-dated the single figure it existed to date; I further concede the Null's inversion, because the $350.0M goodwill charge lands in the very quarter ended August 31 2022 that the record names as the first crossing, so the ledger holds a zero-quarter Corus observation it did not report and cannot, on one document that allocates no goodwill to any fiscal-2023 quarter, sustain a three-quarter interval; the one thing the retrieval gives back is that the $350.0M is attributed to operating underperformance and not to the crossing, which is a partial discriminator against reading it as indicator-responsive, and it is not enough to rescue the interval.
C-08 │ REBUILT: In April 2022 Carla Nunes, a Kroll managing director, told CFO Dive "Relative to 2021 I expect impairments to go up — and some of them might be related to impacts from the Russia-Ukraine conflict," in an article framed on Kroll's assessment of more than 8,900 public U.S. companies — the restriction to goodwill impairments and to US public companies is the article's framing and Kroll's data population, not the quoted sentence, and the claim is narrowed to say so; the record survives its stated falsifier (the direction is up, not flat or declining) but I record the Inquisitor's independence point as correct and unrebutted: C-08 relays a Kroll officer forecasting the market Kroll's study franchise serves, and C-22's sole source is Kroll's own publications hub, so these are two records with one interested origin and should not be counted as two bases.
C-11 │ REBUILT: Hou, Xue and Zhang report that 65 percent of the 452 characteristic portfolios in their own data library — built with microcaps mitigated via NYSE breakpoints and value-weighted returns — have absolute t-values below 1.96; separately and on a different construction, Chen and Zimmermann (FEDS 2021-037), whose 319 characteristics begin from HXZ's 240 but who state "[u]sing the original paper's results, we select the stock-weighting, rebalancing frequency, and quantile sort (if applicable)" — i.e. each original paper's own weighting and breakpoints, predominantly equal-weighted, not HXZ's uniform screen — find t-stats above 1.96 for 98 percent of the 161 characteristics in their set that were clearly significant in the original papers; the two figures are two teams' portfolios on overlapping but non-identical universes under different construction rules, and this record asserts no decomposition of the gap between them, with FEDS 2021-037 entered as a second source since the HXZ abstract page cannot carry the Chen–Zimmermann half at all — I concede the record's "restricted to the 161" framing, which presented one denominator filtered to a clean core and was wrong to, and I concede the Null's contest on the merits rather than defeating it: the discriminating number (HXZ's failure rate restricted to originally-significant characteristics under HXZ's own construction) exists in neither paper, and the nearest evidence I could retrieve — Chen and Zimmermann's own §5.2, where simple liquidity adjustments "reduce mean returns by a factor of about 1/3, on average" — points toward the Null's construction rival rather than the denominator-composition explanation the record's framing invited.
C-15 │ REBUILT: Lightspeed's Q3 FY2023 release states that at December 31, 2022 the carrying amount of its net assets exceeded its market capitalization and that this was an impairment trigger, records a US$748.7 million non-cash goodwill impairment charge at that date, and states that "The Company conducts its annual goodwill impairment test every December 31"; the earliest crossing Lightspeed disclosed is September 30 2022, where the Q2 FY2023 interim statements record the same condition with a completed fair-value-less-costs-of-disposal test showing no impairment — so the documented quantity is a one-quarter upper bound on issuer disclosure-to-recognition distance, and not a measurement of responsiveness to the indicator, because (i) the December 31 right endpoint was fixed by a standing annual-test policy under ¶96 and would have arrived on that date had no crossing occurred, and (ii) the left endpoint is the earliest disclosed crossing, not the earliest crossing: Lightspeed's Q1 FY2023 statements disclose no indicator, no test and no market-capitalisation comparison at June 30 2022, and nothing in the ledger dates the crossing itself — I concede the interval sentence and withdraw the term "crossing-to-recognition interval"; the Null's structural objection is correct and the record's own ¶96 observation was already fatal to it, and I concede it having failed to defeat the Null's rival on the one quantity it named: I retrieved the September 30 2022 test's key assumptions from the ledger's own C-27 document (30% discount rate, 3.4 terminal value multiple, 32% revenue growth) together with its sensitivity disclosure that impairment would have arisen at a 34% discount rate, a 2.8 multiple or 28% revenue growth, which is a moderate rather than razor-thin cushion and therefore, on the Null's own stated test, evidence for its "prompt test, real pass, later deterioration" rival — the correction to the Null is only that this number was inside the ledger and it said there was none.
C-16 │ HELD AS WRITTEN: it survives the Inquisitor's SELECTION attack, which does not touch it — the claim reports what TELUS Corporation's 2024 audited statements say about the TELUS digital experience CGU (recoverable amount exceeded carrying value on a fair-value-less-costs-of-disposal DCF, no impairment recognised, auditor-designated critical audit matter) and expressly discloses that no headroom is quantified, all of which the Null verified verbatim including the negative; the attack's TELUS Digital goodwill-versus-market-capitalisation comparison is drawn from a different issuer's filings and a market-data vendor, neither in the ledger, and it is an argument that the ledger under-covers a live ¶12(d)-crossing-without-recognition observation at the subsidiary level — which I accept as a coverage gap worth flagging to the Adjudicator, and which is not a defect in this record's text.
C-19 │ HELD AS WRITTEN: it survives the Inquisitor's SELECTION attack — the Null pulled the notice and confirmed all twelve enumerated items in order with impairment of non-financial assets in position five and the non-exhaustive "Some areas … include" framing intact, so the stated falsifier is squarely absent; the attack establishes that a later and more substantive regulator document (CSA Staff Notice 51-365, 7 November 2024) exists and is not in the ledger, which is a completeness point about the ledger and leaves every word of this record true.
C-21 │ REBUILT: Lightspeed's Q3 FY2023 release states that at December 31, 2022 the carrying amount of its net assets exceeded its market capitalization and that this was an impairment trigger; this record is a second entry of that one sentence from the same document at the same URL as C-15, carries no evidence independent of it, is weighted at zero, and its LOAD-BEARING designation adds nothing — the fuller trigger narrative ("During the three months ended December 31, 2022, there were changes in macroeconomic conditions and the Company's share price and market capitalization decreased…") and the US$748,712 thousand figure are attributed here to Lightspeed's FY2023 audited statements held at C-35, not to the earnings release, correcting the Inquisitor's PROVENANCE attack, which is right that the release gives "($748.7) million" and no such narrative; with that attribution fixed the record is exactly what the Null called it — a duplicate that voids itself — and the ledger's effective load-bearing base is nine, not ten.
C-22 │ CONCEDED: I cannot rebuild this record in a form that bears load, and the honest disposition is to say so — its entire evidentiary content is one run's own negative retrieval result, tiered E2 and pinned to a URL (Kroll's publications hub) that is evidence for nothing about datasets it does not host, and the record names its own exclusions (Audit Analytics, S&P Capital IQ, Compustat/WRDS, SEDAR+ full-text search), which are precisely the four places an issuer-level Canadian impairment-and-market-capitalisation pairing would be built, so the Null's rival that the base exists or is trivially constructible inside them is not merely undefeated but undefeatable from within this ledger; the positive sub-clause fails independently on the Inquisitor's retrieval (the 2014 edition states it is the third annual study and that the inaugural "2012 Study" was launched in February 2013, and both the 2012 and 2013 surveys drew on FEI Canada members, so "published in 2012" and "the 2012 jointly with FEI Canada" are both wrong), and the Null's further point is decisive on its own — that study reports aggregate goodwill impairment among Canadian public companies and is not an indicator-to-recognition pairing at all, so it cannot be "the nearest precedent construction" for an object of a different kind; the only residue I would preserve is a non-load-bearing procedural note that this run did not inherit an external base and did not search the four named sources, which establishes something about the run and nothing about the world.
C-26 │ REBUILT: Jensen, Kelly and Pedersen report that under their Bayesian hierarchical model of factor replication with shrinkage across factors, 82.4 percent of the US factors that were significant in their original papers replicate — 98 of 119 — on the criterion that the posterior z-score equals or exceeds 1.96, equivalently that the posterior probability of a non-positive alpha falls below 2.5 percent; they separately report that the majority of factors work out of sample in a new large data set covering 93 countries, and it is that global data set, not the replication denominator, that comprises 153 factors; the two findings are distinct and this record asserts no conjunction of them — sourced to the published Journal of Finance 78(5) (2023) version, since I concede both PROVENANCE points: the cited 2021 SSRN abstract page contains neither figure (I fetched the NBER working-paper version and it reports 84.9%, 77.3%, 84.0% and 80.7%, not 82.4%, confirming the number is specific to the published revision) and SSRN serves the current revision at a fixed abstract_id so the recorded date pins no text; I also concede the Inquisitor's INFERENCE attack in full — binding 82.4 percent to "a data set of 153 factors" imported the global count into the replication denominator, which is the exact conflation the record's own final sentence disclaims, and 98/119 = 82.35 percent settles it arithmetically.
C-27 │ HELD AS WRITTEN: it survives both attacks on substance and is the strongest record in the subset — Lightspeed's Q2 FY2023 interim statements state the test reason ("The carrying amount of the Company's net assets exceeded the Company's market capitalization as at September 30, 2022"), the fair-value-less-costs-of-disposal model built on discounted cash flows, and the conclusion of no goodwill impairment at that date, every element verified verbatim by the Null and again by me, with the stated falsifier absent; the Inquisitor's PROVENANCE attack lands only on the source metadata, and it is right — the statements were authorised for issue on 2 November 2022 and the record's source line carries the 2022-09-30 balance-sheet date, which I correct without touching the claim; the independence point (this document is the sole basis for C-15's September-30 half and is not on C-15's source list) is likewise correct and is handled in C-15's rebuild rather than here; I add for the Adjudicator that this same note carries the quantitative material the Null said the ledger lacked — a 30% discount rate, 3.4 terminal value multiple and 32% revenue growth, with impairment arising at 34%, 2.8 or 28% respectively.
C-28 │ REBUILT: Ballard's FY2024 goodwill note names the decline in the Corporation's market capitalization in 2024 as one of three indicators of potential impairment, alongside the global corporate restructuring initiated in September 2024 and indicators of slowing hydrogen and fuel cell policy implementation and market adoption; the same note states that the Corporation "updated its goodwill and non-financial asset impairment tests as of September 30, 2024" and that it "recognized goodwill impairment charges of $40,277,000 to write-down corporate goodwill to $nil", so this is an indicator-to-recognition observation at zero quarters, measured off-cycle at an interim date — I verified all four elements in the cited PDF and concede the Inquisitor's SELECTION attack without reservation: the record extracted the indicator sentence and discarded, from the same paragraph, the measurement date and the recognition, which are the two facts a market-capitalisation-indicator-to-recognition series is built from, and it is the only record in this subset that selected against its own thesis by omission rather than by over-reach; on the PROVENANCE attack I note the Ballard PDF is shared with C-46 and that the two records carry genuinely independent facts, so the correct treatment is one document and two observations, not a netting-out.
C-31 │ REBUILT: TELUS Digital's second quarter 2025 results news release states "As at June 30, 2025, we recorded a non-cash goodwill impairment charge of $224 million" and attributes the decline in recoverable amount to "higher weighted average cost of capital, lower perpetual growth rate and lower cash flow forecasts arising from pricing pressure on margins"; the release identifies no impairment indicator and gives no account of what prompted the test at that date — a negative about the release only, not about the issuer's disclosure, since IAS 36 indicator disclosure lives in the interim financial statements, which the ledger does not hold and which the record's own falsifier invokes — conceding the Inquisitor's PROVENANCE attack, whose point that a syndicated press release cannot establish silence in a financial statement is correct, and recording that the stated falsifier is therefore not executable from within the ledger.
C-35 │ REBUILT: Lightspeed's FY2023 audited statements show nil goodwill impairment in the comparative column for the year ended March 31, 2022, which establishes only that no goodwill impairment loss was recognised in the year ended March 31, 2022 — not that none was recognised before April 1, 2022, since the note reconciles fiscal 2022 from an opening balance of 971,939 at 1 April 2021 into which any earlier impairment is already netted and would not separately appear — and it establishes nothing about whether a crossing occurred or a test was performed in that year, the statements carrying no comparative disclosure of FY2022 testing or indicators; I concede the Inquisitor's INFERENCE attack entirely, and note that it is the sharper of the two readings of this record, since the Null verified the same note and endorsed the over-broad wording: a record written to police what a comparative column can establish had over-read its own comparative by every year preceding fiscal 2022.
C-36 │ REBUILT: BCE's 2024 fourth-quarter and full-year results release states non-cash asset impairment charges totalling $2,190 million for 2024, "mainly related to Bell Media's TV and radio properties to reflect a further decline in demand and spending in the traditional advertising market," and the release itself states no goodwill component and decomposes the charge no further — but the conclusion that this supports no goodwill observation is withdrawn: $2,113 million of that full-year total is the Q3 2024 charge, which BCE's Q3 2024 consolidated financial statements decompose to include $1,132 million of goodwill impairment in Bell Media, 54 percent of the charge and its largest single component, so the goodwill observation exists and the record's silence was an artefact of citing the press release and stopping; the assertion that this is one disclosure event with C-41 is also withdrawn — C-41 is the Q3 release of 7 November 2024 stating $2,113 million for the quarter ended 30 September 2024 and C-36 is the full-year release of 6 February 2025 stating $2,190 million for the year, two periods four months apart with an unexplained $77 million increment that no record in the subset addresses; I verified the $1,132 million decomposition in the SEC-filed statements and the absence of any goodwill reference or separate Q4 impairment figure in the full-year release, and I concede both attacks.
C-37 │ REBUILT: The Canadian Press reported the figure BCE published the same day, rounded from $2,113 million to $2.11 billion, mainly related to Bell Media properties; the record is wire coverage of the release at C-41, carries no evidentiary weight independent of it and should not count toward source totals, and its genuinely independent content concerns the CRTC fibre-access decision, the share price closing at $38.94 and Scotiabank analyst commentary, none of which touches the impairment; the article describes the charge as "asset impairment" and the word goodwill appears nowhere in the press chain — a statement about the press chain only, since BCE's Q3 2024 consolidated financial statements, filed the same day under the same accession number, state a $1,132 million goodwill impairment within Bell Media as part of the same $2,113 million — conceding the Inquisitor's PROVENANCE attack, which is right that "no goodwill component is stated anywhere in the reporting" asserted a universal negative over the issuer's disclosure record on the strength of a newswire summary, and right that the Null's charitable reading of that phrase was more generous than the words allowed.
C-41 │ REBUILT: BCE's Q3 2024 results news release states non-cash asset impairment charges totalling $2,113 million, "mainly related to Bell Media's TV and radio properties," for the quarter ended September 30, 2024, with no goodwill component and no allocation in the release itself; BCE's Q3 2024 consolidated financial statements, filed the same day as EX-99.2 to the same 6-K, resolve both gaps: goodwill $1,132 million, indefinite-life intangible assets (broadcast licences and brands) $627 million, programme and feature film rights $144 million, property plant and equipment $85 million, software $85 million, finite-life intangibles $10 million and prepaid expenses $7 million, against CGU groups named as the English and French TV services and radio markets, attributed to a further decline in advertising demand and spending — so the record does support a goodwill observation and the "unquantified remainder" it declared is quantified; I verified the decomposition and concede the SELECTION attack, which is correct that the record's own falsifier pointed at a free, indexed document under the accession number the record already cites, and that "the record supports no goodwill observation" was a statement about the ledger's retrieval rather than about BCE.
C-45 │ REBUILT: Hayn and Hughes report that goodwill write-offs lag behind the economic impairment of goodwill — measured on the authors' own fundamentals-based construct, not a regulatory or accounting quantity — by an average of three to four years, with the delay extending to about ten years for roughly a third of companies examined; it is a mean and not a median, and the ten-year tail forces the remaining two-thirds to average roughly a quarter of a year, so a median well below three years is not merely compatible but implied; the sample-composition and SFAS 142 sensitivity-test clauses are removed as unsupported by the cited artifact — both attackers independently found the SSRN abstract page carries the two lag figures and nothing else, and I accept that — and the record's conclusion stands as a limitation rather than an assertion: it is not transportable to a one-step IAS 36 CGU test with a mandatory annual cycle in Canada 2022–2025 without a bridging record this ledger does not hold, which is a statement about the ledger and not about the literature, where Li and Sloan, "Has goodwill accounting gone bad?", Review of Accounting Studies 22(2) (2017), tests impairment timeliness under the post-SFAS-142 impairment-only regime and is the standard citation the ledger is missing.
C-46 │ REBUILT: Ballard's FY2024 audited consolidated statements record, in the 2023 comparative, a US$23,991 thousand goodwill impairment "related to the restructuring and effective closure of operations of Ballard Motive Solutions in the U.K.", presented within loss from discontinued operations; the caveat that this filing cannot establish presentation as originally reported is withdrawn — Ballard's 2023 Annual Report MD&A §5.2 states that "[d]uring the fourth quarter of 2023, we completed a restructuring of operations at Ballard Motive Solutions in the U.K. and effectively closed the operation" and that the historic operating results for both 2023 and 2022 "have been removed from continuing operating results and are instead presented separately … as loss from discontinued operations", so the discontinued-operations classification was adopted in Q4 2023 as originally reported and was not created by FY2024 restatement; the record's substantive point survives untouched and is the reason the correction matters rather than the reason it does not — a closure-driven write-off inside discontinued operations is not an indicator-triggered CGU test outcome and does not belong in a market-capitalisation-indicator-to-recognition series — and I concede the Inquisitor's SELECTION attack, which is right that the record's central caveat was a limit on retrieval dressed as a limit on evidence.
WHAT THE ADJUDICATOR SHOULD TAKE FROM THIS
Two conclusions this round's retrieval forces, neither of which the ledger held going in:
1. The ledger has two zero-interval observations it did not report. Corus recognised $350.0 million of Television-CGU goodwill impairment in the year ended August 31 2022 — the same period-end as the first crossing C-07 names — and Ballard updated its tests as of September 30 2024 and wrote corporate goodwill down to nil by $40,277 thousand in the same note C-28 quotes for indicators. Both were inside documents the ledger already cites. The subset as written reported an interval at Lightspeed and an interval at Corus and reported neither zero.
2. BCE yields a goodwill observation of $1,132 million. Three records (C-36, C-37, C-41) agreed that the largest Canadian impairment event in the subset supported no goodwill observation. It does, in a statement filed under the accession number one of those records already cites.
The corrected direction of the evidence is therefore the opposite of the direction the subset's constructed quantities pointed. That is the finding, and it is not one I can argue against.
CONCESSION: MOVED — I entered expecting to narrow four or five records and hold the rest, and the retrieval moved me on far more than that. I concede C-22 outright: a load-bearing negative whose search space excluded SEDAR+, Compustat/WRDS, Capital IQ and Audit Analytics cannot be rebuilt into anything that bears load, and its one checkable positive sub-clause is wrong on three counts. I concede the operative half of all three of the Null's remaining contests and defeat none of them: C-07's component allocation is arithmetically impossible and the $350.0M fiscal-2022 charge inverts its interval; C-11's "restricted to the 161" framing collapses once Chen and Zimmermann's own construction rule is read, and the only discriminating evidence I could retrieve — their §5.2 finding that liquidity adjustments cut mean returns by about a third — leans toward the Null's construction rival, not away from it; and on C-15 I found the September 30 2022 sensitivity disclosure the Null said was absent, inside the ledger's own C-27 document, and it is a moderate cushion, which on the Null's own stated test is evidence for its rival — my only correction to the Null there is that the number existed, and it did not help me. I concede the Inquisitor's BCE attack across C-36, C-37 and C-41 and withdraw their shared conclusion, and the C-36/C-41 merger with it. I concede C-28's omission, C-35's over-reading, C-46's dissolved caveat, C-31's mis-scoped negative, C-26's imported denominator, C-45's body-of-paper clauses, C-21's misattribution, and half of C-04's provenance. I hold three records unamended — C-16, C-19 and C-27 — each because the attack against it establishes a gap in the ledger's coverage or a defect in its metadata rather than an error in its text, and I hold one inference against the Inquisitor: IAS 36 ¶15 relieves an estimate rather than deferring a test, so it does not satisfy C-04's falsifier and in fact strengthens C-04's conclusion, which is why I wrote it into the claim. What survives is a subset that is reliable where it reports what a document says and unreliable wherever it built a quantity across documents — and where the quantities are rebuilt correctly, they point the other way.
VERDICT MAP
Final consolidated map after round 3, 05-verdict/map-FINAL-r3.md. Verdict tokens carry (round2)
or (round3) where the verdict or the claim text moved in that round. The tripwire did not fire, so
no verdict is marked (tripwire).
> Note for auditors: the headings # 1 · CONFIDENCE ARITHMETIC, # 2 · THE CRUX and
> # 3 · TRIPWIRE appearing below are the Adjudicator's own three required sections under §8,
> reproduced verbatim inside this item. They are not §12 items 1, 2 and 3. Nothing in the agent's
> output was renumbered, because C2 forbids editing it.
ADJUDICATION — CONSOLIDATED VERDICT MAP (FINAL, AFTER ROUND 3)
run: LIVE · evidence cutoff 2026-08-22 · ledger = 48 claims (C-01..C-48)
ANACHRONISM vector disabled for this run; no attack was filed under it in any of the three rounds, so none was rejected on that ground.
Round 3 jurisdiction: the 19 claims that re-entered as NARROWED or CONTESTED after round 2
(C-04 C-07 C-08 C-11 C-15 C-16 C-19 C-21 C-22 C-26 C-27 C-28 C-31 C-35 C-36 C-37 C-41 C-45 C-46).
The other 29 are frozen by construction and are carried forward verbatim from map-final.md.
Round-3 attacks scored: Inquisitor 27 · Null 4 (C-07, C-11, C-15, C-22) = 31.
Round 3 accepted 25 · rejected 6 (see rejected-attacks-r3.md).
Every round-3 Inquisitor attack was filed evidence_status: RETRIEVED, so the ASSERTED bar was not
reached in this round and no attack was rejected on it.
Format: C-nn │ VERDICT │ tier │ load │ claim text (replacement text where NARROWED) │ reason
(round3) on the verdict token marks a claim whose verdict or text changed this round.
C-01 │ HELD │ E1 │ SUPPORTING │ Reuters reported IOSCO naming 'too little, too late' goodwill impairment a priority issue for securities regulators. │ Single PROVENANCE attack (third-hand wire, primary absent from ledger). The claim asserts what the wire reported, and the attack's own evidence confirms that attribution; the Advocate then retrieved the primary (IOSCO FR13/23, December 2023), which carries the phrase in IOSCO's own words. Citation hygiene, not a defeater.
C-02 │ KILLED │ E2 │ LOAD-BEARING │ Majority non-recognition within the indicator year, as reported by ESMA plus André Filip Paugam, places the median indicator-to-recognition lag above four quarters. │ Falsified as a non-sequitur: an indicator in Q4 followed by a charge in Q1 is a one-quarter lag counted as annual non-recognition, so an annual non-recognition rate places no lower bound on a quarterly median. Both attackers verified that neither cited source reports any recognition-timing distribution; the Advocate conceded outright and his own measurement runs the other way. See graveyard-FINAL.md.
C-03 │ HELD │ E0 │ LOAD-BEARING │ Corus disclosed that its carrying value was greater than its market enterprise value at August 31, 2022. │ Only attack was duplication with C-07 off one MD&A clause — a bookkeeping defect that instructs weighting, not truth. The Null verified the disclosure independently against Corus's FY2022 year-end release and expressly did not contest it.
C-04 │ NARROWED (round3) │ E2 │ LOAD-BEARING │ IAS 36 ¶9 requires an entity to assess at the end of each reporting period whether any indication of impairment exists and, where one does, to estimate the recoverable amount of the asset or CGU; ¶12(d) names the carrying amount of the entity's net assets exceeding its market capitalisation as such an indicator. That estimation obligation is qualified by ¶15, which — "[a]part from when the requirements in paragraph 10 apply" — provides that materiality applies in identifying whether recoverable amount needs to be estimated, that an entity need not re-estimate where previous calculations showed recoverable amount significantly greater than carrying amount and no events have occurred that would eliminate that difference, and that previous analysis may show recoverable amount insensitive to one or more ¶12 indications; ¶15's opening words place goodwill and indefinite-life intangibles outside that relief. Separately and in addition, ¶10 requires goodwill acquired in a business combination and indefinite-life intangibles to be tested annually irrespective of whether any indication exists, ¶96 permitting that annual test at any single date provided it is performed at the same time each year, and ¶99 permitting a preceding period's detailed calculation to be carried forward on stated criteria. Recognition follows only where recoverable amount falls below carrying amount. An interval between a ¶12(d) crossing and a recognised loss is therefore not dispositive that a test was deferred. The paragraph numbering, and the ¶15 and ¶99 text, are verified against the IAS 36 consolidated standard text retrieved this run; the ledger's sole cited source — a Grant Thornton advisory article — carries the substance of ¶10, ¶12(d) and ¶96, prints no paragraph number other than ¶96 (and calls ¶96 an anti-abuse provision), and says nothing about ¶15 or ¶99. │ Text moved twice more. The PROVENANCE attack is accepted in the half the Advocate concedes and the Null independently confirms: the article carries the substance of three of the paragraphs but the numbering of one, so it cannot bear a six-paragraph reading of the standard, and the reading now rests on the consolidated text both the Inquisitor and the Advocate retrieved verbatim this run. The SELECTION attack is accepted on its substantive half — ¶15 is material text a five-paragraph reading omitted, and it is now written into the claim — and rejected on its conclusion: ¶15 dispenses with an estimate on stated conditions rather than postponing testing to a later period, so it does not satisfy a falsifier that asks for a provision permitting deferral, and its opening carve-out excludes ¶10 assets, which is to say the whole subject matter of a goodwill ledger. Its direction also runs the record's way, not against it: a provision that can relieve the estimate outright makes an observed interval weaker evidence of deferral, not stronger. The Null's exact quibble is adopted — "not by itself evidence" becomes "not dispositive." The Null did not contest this record at all. Tier stands at E2 because the ledger's own source line still points at the advisory article; the standard's text is corroboration retrieved outside the ledger.
C-05 │ HELD │ E0 │ SUPPORTING │ Grant Thornton lists carrying amount of net assets exceeding market capitalisation as an external indicator of potential impairment under IAS 36. │ Attacked only as a duplicate of C-20 at a different tier — ledger hygiene. The indicator sentence was verified verbatim by both attackers and the Advocate.
C-06 │ HELD │ E0 │ LOAD-BEARING │ Lightspeed Commerce recognized a US$748.7 million goodwill impairment charge in the three months ended December 31 2022. │ Attacked only for eighth-order duplication across the Lightspeed cluster. The charge is triangulated four ways — SEC release, Q3 interim statements, FY23 audited statements, and the ledger's own equity roll-forward (US$3,304.4M → US$2,539.5M) — and the Null concedes it explicitly.
C-07 │ NARROWED (round3) │ E0 │ LOAD-BEARING │ Corus's fiscal 2023 annual MD&A states that the Company's share price "has continued to decline meaningfully from August 31, 2021, which resulted in the Company's carrying value being greater than its market enterprise value at August 31, 2022, May 31, 2023, and August 31, 2023." The measure is carrying value against market enterprise value, which includes net debt — not net assets against market capitalisation — and the document does not state the condition's status at November 30 2022 or February 28 2023. The same document records $690.0M of Television-CGU impairments for the year ended August 31 2023, decomposed at the year level only — goodwill $295.2M, broadcast licences $219.8M, brands and trade marks $175.0M — and separately splits that year total into $590.0M in the quarter ended May 31 2023 and $100.0M in the quarter ended August 31 2023, with no component allocation to either quarter; the document therefore does not date the goodwill component to any quarter, and no Corus fiscal-2023 crossing-to-recognition interval can be read off it. The same document also records that for the year ended August 31 2022 a non-cash goodwill impairment charge of $350.0M was recorded in the Television CGU, attributed to actual results falling short of previous estimates and to a less robust outlook rather than to the enterprise-value crossing — a recognition falling at the same period-end, August 31 2022, as the first crossing the MD&A names. │ Text moved on both round-3 attacks, and the round-2 text is corrected on arithmetic that survived a full round unchecked. $295.2M + $219.8M + $175.0M = $690.0M, which is the fiscal-year total; presenting those three as the composition of the $590.0M quarter was impossible on its face, and it mis-dated the single figure — the goodwill component — that this record existed to date. Both attackers found it independently in the cited PDF and the Advocate conceded it outright after re-verifying. The Null's second point is accepted with it: the record named a crossing at August 31 2022 and then reported recognition at May 31 2023 while the $350.0M goodwill charge sitting at the named crossing's own period-end went unmentioned. That charge is now booked in the text. The Null's rival — Corus recognised at the crossing, with no interval — is not defeated; the one thing retrieval gives back against it is that the $350.0M is attributed to operating underperformance rather than to the crossing, a partial discriminator the Advocate correctly declined to call sufficient. Because the narrowed text asserts no interval, the rival no longer contests anything the claim says.
C-08 │ NARROWED (round3) │ E1 │ SUPPORTING │ In April 2022 Carla Nunes, a Kroll managing director, told CFO Dive "Relative to 2021 I expect impairments to go up — and some of them might be related to impacts from the Russia-Ukraine conflict." The quoted sentence names neither goodwill nor a population; the restriction to goodwill impairments among US public companies is the article's framing and Kroll's data population — an assessment of more than 8,900 public US companies — and not Nunes's words. It remains a dated ex ante forecast of direction only, carrying nothing about IFRS issuers, outturns, or lag. │ Text moved on the accepted INFERENCE attack, which quotes Nunes verbatim and shows the two restrictions the record placed in her mouth are the headline's and the ledger's. The Advocate concedes and supplies the article's own framing, which does carry the population, so the restriction survives as the article's rather than the speaker's. The PROVENANCE attack — that C-08 and C-22 are one interested Kroll origin behind two URLs — is rejected for the second time: it is source concentration and a weighting instruction, not circularity, and the two records are different documents about different objects. The Advocate accepts its weighting consequence and so do I: these are not two independent bases.
C-09 │ HELD │ E0 │ SUPPORTING │ ESMA found 43 percent of 235 sampled European issuers carried market capitalisation below equity book value at 31 December 2011. │ Attacked only for non-independence with C-02 and C-29 — a weighting instruction. The Null retrieved the primary and matched both figures exactly (235 entities across 23 jurisdictions; 43% below equity).
C-10 │ HELD │ E0 │ LOAD-BEARING │ André Filip Paugam find only 20 to 25 percent of firms exhibiting economic impairment indications recognised an accounting impairment in the same period. │ Attacked only as one abstract split into a result record and a method record (with C-24) — weighting. The Null verified the sentence exactly against the abstract.
C-11 │ NARROWED (round3) │ E0 │ SUPPORTING │ Hou, Xue and Zhang report that 65 percent of the 452 characteristic portfolios in their own data library — built with microcaps mitigated via NYSE breakpoints and value-weighted returns — have absolute t-values below 1.96. Separately, and on a different construction, Chen & Zimmermann (FEDS 2021-037), whose 319 characteristics begin from HXZ's 240 but who state that "[u]sing the original paper's results, we select the stock-weighting, rebalancing frequency, and quantile sort (if applicable)" — each original paper's own weighting and breakpoints, predominantly equal-weighted — find t-stats above 1.96 for 98 percent of the 161 characteristics in their own set that were clearly significant in the original papers. The 161 is a subset of Chen & Zimmermann's 319, not a restriction of HXZ's 452; the two figures are two teams' portfolios on overlapping but non-identical universes under different construction rules, and this record asserts no decomposition of the gap between them. The HXZ abstract page cannot carry the Chen & Zimmermann half; FEDS 2021-037 is required as a second source and the ledger enters none. The record still cannot serve as a general reliability discount on published findings and does not transfer to descriptive recognition frequencies read off filings. │ Text moved for the second consecutive round, and this time the round-2 narrowing is itself corrected. Round 2 wrote "restricted to the 161," which presented one denominator filtered to a clean core; the INFERENCE attack shows Chen & Zimmermann do not rebuild HXZ's portfolios at all and use precisely the construction HXZ's screen is designed to reject, so 65 % and 98 % are not one population measured two ways and the gap is not attributable to the 452→161 restriction. The PROVENANCE attack is accepted rather than treated as hygiene — unlike round 2's, it shows the cited page cannot carry the record's second sentence at all and that a two-paper record reports "sources: 1." The Null's contest is accepted with them: its construction rival (the NYSE-breakpoint/value-weight screen, not denominator composition, drives the gap) is undefeated, and the Advocate conceded it on the merits and retrieved evidence leaning the Null's way (C&Z §5.2: liquidity adjustments cut mean returns by about a third). The discriminating number — HXZ's failure rate restricted to originally-significant characteristics under HXZ's own construction — exists in neither paper. Because the narrowed text asserts no decomposition, no live disagreement survives it.
C-12 │ KILLED │ E2 │ LOAD-BEARING │ Lightspeed's disclosed lag from market-capitalization trigger to goodwill impairment recognition is zero quarters. │ Falsified by the ledger's own primary source: Lightspeed's FY23-Q2 goodwill note states the carrying amount of net assets exceeded market capitalization as at September 30 2022 and that this triggered an impairment test. The disclosed lag is one quarter, not zero. The ledger cited that PDF three times without ever quoting the sentence. Advocate conceded outright. See graveyard-FINAL.md.
C-13 │ HELD │ E0 │ LOAD-BEARING │ Grant Thornton's IFRS team states identifying an impairment indicator does not automatically trigger recognition of an impairment loss. │ Attacked for the ledger drawing mutually inconsistent records off one page — correct, but the inconsistency is C-43's, not C-13's. Verified verbatim by both attackers and the Advocate; it is the record the Null calls its central exhibit and the record that narrows C-04.
C-14 │ KILLED (round2) │ E3 │ SUPPORTING │ Where the in-window population of first-time impairers is small, single-issuer reclassification moves the reported median; on this run's retrieval the population of ledger issuers with a determinable crossing-to-recognition pair is three (Lightspeed, Ballard, TELUS Digital), of which two are reclassifiable on definitional grounds alone. │ Verdict changed. Round 1 narrowed this record only because the Advocate had established its antecedent by count; round 2 removes exactly that. The enumeration is falsified from inside the ledger — Corus is a fourth issuer with both endpoints in a single document the ledger already holds (C-07's source names the crossings and records $295.2M of goodwill within $590.0M of Q3 fiscal 2023 charges). Its governing term "first-time impairer" is nowhere defined and is outcome-determinative at this n: Ballard impaired goodwill in 2023 (C-46) and again in 2024 (C-28's note, C-44's $40,277 thousand), so whether Ballard is a 2024 first-timer turns on a choice the ledger never states. The record has no source of any kind and its stated falsifier tests the universe (over one hundred issuers) while the live risk is the retrieved population, which no stated falsifier can catch. The Null's retrieval-by-outcome rival stands undefeated and the Advocate conceded outright, recommending the record be struck rather than narrowed. See graveyard-FINAL.md.
C-15 │ NARROWED (round3) │ E0 │ LOAD-BEARING │ Lightspeed's Q3 FY2023 release states that at December 31, 2022 the carrying amount of its net assets exceeded its market capitalization and that this was an impairment trigger, records a US$748.7 million non-cash goodwill impairment charge at that date, and states that "The Company conducts its annual goodwill impairment test every December 31," so under IAS 36 ¶96 that test would have occurred irrespective of the trigger. The earliest crossing Lightspeed disclosed is September 30 2022, where the Q2 FY2023 interim statements — C-27's source, which is not listed on this record — record the same condition with a completed fair-value-less-costs-of-disposal test showing no impairment. The documented quantity is therefore a one-quarter upper bound on issuer disclosure-to-recognition distance and not a measurement of responsiveness to the indicator: the right endpoint was fixed by a standing annual-test policy and would have arrived on that date had no crossing occurred, and the left endpoint is the earliest disclosed crossing, not the earliest crossing — Lightspeed's Q1 FY2023 statements disclose no indicator, no test and no market-capitalisation comparison at June 30 2022, and nothing in the ledger dates the crossing itself. │ Text moved, and the sentence round 2 called "the only interval in the ledger with both endpoints documented" is withdrawn. All three round-3 attacks are accepted and they converge. The PROVENANCE attack shows the record's headline quantity rests on a document it does not cite: the release contains no reference to September 30 2022, to a prior test, or to a fair-value-less-costs-of-disposal conclusion. The INFERENCE attack retrieves Lightspeed's Q1 FY2023 statements, which disclose nothing at June 30 2022, and shows the Q2 note dates the decline across six months rather than dating the crossing. The Null's structural objection is decisive and was already implied by the record's own ¶96 observation: an interval whose right endpoint is fixed by a standing annual calendar cannot measure responsiveness to an indicator that arrived after the endpoint was set. The Advocate conceded and then, on the Null's own stated discriminator, went further — he retrieved the September 30 2022 sensitivity disclosure from inside the ledger's own C-27 document and found a moderate rather than razor-thin cushion, which is evidence for the Null's "prompt test, real pass, later deterioration" rival. Because the interval sentence is withdrawn from the claim rather than left standing, no live contest remains; what the ledger loses is its only measured responsiveness observation.
C-16 │ NARROWED │ E0 │ SUPPORTING │ TELUS Corporation's 2024 audited consolidated statements record that management determined the recoverable amount of the TELUS digital experience CGU exceeded its carrying value at the measurement date and that no impairment was recognized, on a fair value less costs of disposal calculation using discounted cash flow projections — a judgement the auditor designated a critical audit matter; the document quantifies no headroom. │ Text unchanged; the single round-3 attack rejected. The Null verified every clause verbatim including the negative and the absence of any headroom figure in the CAM or Note 18(d), and did not contest. The SELECTION attack is the strongest of its kind filed in this run — it finally extracts the subsidiary-level figures (TELUS Digital goodwill US$1,926M at December 31 2024 against TELUS International (Cda) market capitalisation of US$1.08bn) rather than merely naming an absent document — but the figures come from a different issuer's release and a market-data vendor, neither in the ledger, and they contradict no word of this record, which reports the parent's audited conclusion about a parent-level CGU. Recorded because it matters elsewhere and not here: the attack establishes a genuine coverage gap — the ledger holds no subsidiary-level ¶12(d)-crossing-without-recognition observation, which is the observation type most load-bearing for the question the ledger was built to answer, and the Advocate flags it as such rather than defending against it.
C-17 │ KILLED │ E3 │ LOAD-BEARING │ Threshold claims stating a median exceeds a bound far beneath the reference class central estimate resolve true more often than thresholds set near that estimate. │ LOAD-BEARING on zero external evidence, and its central term has no referent here: the candidate reference classes (Hayn & Hughes, ESMA, André/Filip/Paugam) are measured in three mutually non-convertible constructs, so there is no "reference class central estimate" for a threshold to be placed relative to. The tabulation its own falsifier presupposes is named nowhere. Advocate conceded with nothing offered in defence. See graveyard-FINAL.md.
C-18 │ HELD │ E0 │ LOAD-BEARING │ Lightspeed had 150,315,764 subordinate voting shares issued outstanding at September 30, 2022. │ Attacked only for three records off one PDF — weighting. The Advocate closed the Null's caveat that this is not a total share count (the FY23-Q2 share capital note shows no other class outstanding at that date) and performed the implied per-share arithmetic.
C-19 │ NARROWED │ E0 │ SUPPORTING │ CSA Staff Notice 51-364 names impairment of non-financial assets among a non-exhaustive list of areas that may be impacted by the current economic environment ("Some areas that may be impacted by the current economic environment include …"), alongside known trends/events/uncertainties, liquidity and capital resources, debt covenants, risk factor disclosure, going concern, events after the reporting period, significant judgement and measurement uncertainties, expected credit losses, financial instrument risk disclosure, non-GAAP and other financial measures, and material change reporting. │ Text unchanged; the single round-3 attack rejected. The Null pulled the notice and confirmed all twelve items in order with impairment of non-financial assets in position five and the "Some areas … include" framing intact, so the stated falsifier is squarely absent and round 2's correction of the count holds. The SELECTION attack retrieves a real and more substantive successor document — CSA Staff Notice 51-365 of 7 November 2024, pre-cutoff, reporting actual review findings that issuers give boilerplate impairment disclosure and do not sufficiently explain the events and circumstances leading to an impairment — but it contradicts nothing in this record and, if anything, runs the ledger's way. It is a completeness point about what the ledger cites, on the same ground the round-1 and round-2 SELECTION attacks naming absent-but-consistent documents were rejected.
C-20 │ HELD │ E1 │ LOAD-BEARING │ Grant Thornton lists carrying amount of net assets exceeding market capitalisation among the IAS 36 external indicators requiring an impairment review. │ Attacked as a duplicate of C-05 at a different tier — hygiene. Verified verbatim by both attackers. The falsifier's appeal to IAS 36 paragraph 12, a text the ledger never sources, is a genuine citation gap that does not touch the claim's truth.
C-21 │ NARROWED (round3) │ E0 │ LOAD-BEARING │ Lightspeed's Q3 FY2023 release states that at December 31, 2022 the carrying amount of its net assets exceeded its market capitalization and that this was an impairment trigger. This record is a second entry of the same sentence in the same document at the same URL as C-15 and carries no evidence independent of it: it is weighted as zero additional support, and its LOAD-BEARING designation adds nothing to the ledger's load-bearing base. The fuller trigger narrative ("During the three months ended December 31, 2022, there were changes in macroeconomic conditions and the Company's share price and market capitalization decreased…") and the US$748,712 thousand figure belong to Lightspeed's FY2023 audited statements held at C-35, not to the earnings release, which gives "($748.7) million" and carries no such narrative. │ Text moved on the accepted PROVENANCE attack. Round 2 added a closing sentence to a record whose whole function is to declare itself weightless, and mis-sourced it: the narrative and the exact-thousands figure are in the FY2023 audited statements, which the record locates at C-35 but never enters as its own source. The attribution is corrected in the text; the Advocate supplied the correction and the Null, which conceded the record, called contesting a self-voided duplicate "padding" while noting that the LOAD-BEARING tag still counts one record toward the base. That count is handled in the arithmetic below, not by changing the record's filed load.
C-22 │ NARROWED (round3) │ E2 │ LOAD-BEARING │ Retrieval in this run — across Kroll's goodwill-impairment publications hub, IOSCO, Reuters and issuer filings — did not surface any dataset pairing S&P/TSX Composite non-financial issuers' first goodwill impairments with net-assets-above-market-capitalization crossings at quarter-end granularity for 2022–2025. The named search space excludes Audit Analytics, S&P Capital IQ, Compustat/WRDS extracts and SEDAR+ full-text search, which are the four venues in which such a pairing would be constructed, so the record establishes what this run's retrieval did not reach and nothing about what exists; it is a procedural note about the run and carries no load. The claim that Duff & Phelps/Kroll's Canadian Goodwill Impairment Study is "the nearest precedent construction for this population" is withdrawn: that series reports aggregate goodwill impairment among Canadian public companies and is not an indicator-to-recognition pairing at all, so it is a different object rather than a nearer one. Its cadence as retrieved is three annual editions, the inaugural "2012 Study" launched in February 2013 and the last in 2014, with both the 2012 and 2013 surveys drawing on FEI Canada / CFERF participation — so "published in 2012" and "the 2012 jointly with FEI Canada" were both wrong. │ Text moved and the record's positive content is stripped out. The PROVENANCE attack is accepted: the sole source is this run's own unlogged retrieval pinned to a URL that hosts nothing about the datasets in question, and the record's one checkable factual assertion is wrong on retrieval in both particulars. The Null's contest is accepted on the load question, and the Advocate conceded it outright — "I cannot rebuild this record in a form that bears load." A negative whose search space excludes SEDAR+, Compustat/WRDS, Capital IQ and Audit Analytics cannot bear weight against the object's existence, and the Null's rival that the base exists or is trivially constructible inside them is not merely undefeated but undefeatable from within this ledger. The Null's further point is decisive on the positive half independently: the Kroll series measures a different quantity, so it cannot be the nearest construction of an object of another kind. Not killed — the retrieval-bounded negative is true and uncontested by all three parties, and the record's stated falsifier (a pre-cutoff study reporting that pairing) did not fire, because nobody produced one. What is gone is its load, and it goes the same way C-21's did: voided in the text rather than removed from the ledger.
C-23 │ HELD │ E0 │ LOAD-BEARING │ Lightspeed Commerce recognized a non-cash goodwill impairment charge of US$748.7 million for the quarter ended December 31, 2022. │ Attacked only as the third rendering of one charge (with C-06 and C-42) — weighting. Amount and period conceded by the Null and verified by the Advocate.
C-24 │ HELD │ E0 │ SUPPORTING │ André Filip Paugam measure economic impairment indications using market-to-book below one, equity market value minus book value less than goodwill, negative EBITDA. │ Attacked only for sharing an abstract page with C-10 — weighting. All three proxies verified exact against the abstract by the Null and the Advocate.
C-25 │ HELD │ E0 │ LOAD-BEARING │ Corus recorded a non-cash goodwill impairment charge of $350.0 million in the Television cash generating unit in fiscal 2022. │ The single-document attack fails on the record: the Null corroborated the charge outside the ledger's Corus PDF against the fiscal 2022 Q4 release. One of the few records in this ledger with genuine source independence.
C-26 │ NARROWED (round3) │ E0 │ SUPPORTING │ Jensen, Kelly & Pedersen report that, under a Bayesian hierarchical model of factor replication with shrinkage across factors, 82.4 percent of the US factors that were significant in their original papers replicate — 98 of 119 — on the criterion that the posterior z-score equals or exceeds 1.96, equivalently that the posterior probability of a non-positive alpha falls below 2.5 percent. They separately report that the majority of factors work out of sample in a new large data set covering 93 countries, and it is that global data set, not the replication denominator, that comprises 153 factors. The two findings are distinct and this record asserts no conjunction of them. The figures belong to the published Journal of Finance 78(5) (2023) version; the cited 2021 SSRN abstract page carries neither, the NBER working-paper version reports 84.9 / 77.3 / 84.0 / 80.7 percent rather than 82.4, and SSRN serves the current revision at a fixed abstract_id, so the recorded date pins no text. │ Text moved for the second consecutive round, and round 2's own narrowing is corrected. The INFERENCE attack shows the record bound 82.4 percent to "a data set of 153 factors," importing the global out-of-sample factor count into the replication denominator — the exact conflation the record's final sentence disclaims. The attackers split (the Null verified 82.4 percent and a 153-factor data set in the published paper and did not contest), and the Advocate broke the tie for the Inquisitor with arithmetic that settles it: 98/119 = 82.35 percent. The PROVENANCE attack is accepted on the same footing round 2 accepted its predecessor — the cited artifact carries neither figure, and the Advocate's retrieval of the NBER version, which reports different numbers entirely, shows the 82.4 is specific to the published revision. The source is corrected to the published Journal of Finance version in the text.
C-27 │ NARROWED (round3) │ E0 │ LOAD-BEARING │ Lightspeed performed a goodwill impairment test as at September 30, 2022 because the carrying amount of its net assets exceeded its market capitalization, using a fair value less costs of disposal model, and the test demonstrated no impairment of goodwill at that date. The same note discloses the test's key assumptions — a 30% discount rate, a 3.4 terminal value multiple and 32% revenue growth — and states that impairment would have arisen at a 34% discount rate, a 2.8 multiple or 28% revenue growth: a moderate rather than razor-thin cushion at that date. │ Verdict unchanged and both round-3 attack limbs rejected, but the text is moved to book material the record's own primary source carries and the ledger never entered. The Null calls this "the cleanest thing in the subset" and every clause was verified verbatim for the third consecutive round. The PROVENANCE attack's date point (2022-09-30 is the balance-sheet date; the statements were authorised 2 November 2022) is metadata and was rejected on identical grounds at C-32 in round 1 and at this record in round 2; its independence point is real and is booked where it belongs, in C-15's narrowed text, not here. The addition is the Advocate's: the sensitivity disclosure is the exact quantity the Null said the ledger lacked, it was inside this record's own document all along, and it is the discriminator on which C-15's rival now turns — which is why it is booked rather than left in an attack file. This remains the ledger's one documented instance of a ¶12(d) crossing producing a completed test and no recognition.
C-28 │ NARROWED │ E0 │ LOAD-BEARING │ Ballard's 2024 goodwill note names the decline in the Corporation's market capitalization as one of three indicators of potential impairment, alongside the global corporate restructuring initiated in September 2024 and indicators of slowing hydrogen and fuel cell policy implementation and market adoption. │ Text unchanged; both round-3 attacks rejected, both for the second time. The SELECTION attack repeats round 2's false premise: it says the measurement date and the US$40,277 thousand recognition appear nowhere, when they are booked at C-44 (HELD, E0, LOAD-BEARING) — "on testing performed as at September 30, 2024" — and the attack is again reading only the re-entered subset rather than the ledger. The PROVENANCE attack (same PDF and note as C-46) is the weighting instruction rejected at C-44 in round 1 and at this record in round 2; the Null and the Advocate both confirm the two records carry independent facts. The Advocate conceded the SELECTION attack and drew the consequence that Ballard is a zero-quarter observation measured off-cycle — a reading round 1's narrowing already anticipated and which the ledger supports across C-28 and C-44 jointly. Round 1's narrowing stands unaltered: the September 2024 restructuring is a sufficient independent trigger arising in the same quarter as the test, so this record cannot on its own date a market-capitalisation-indicator-to-recognition interval.
C-29 │ HELD │ E0 │ LOAD-BEARING │ Among ESMA-sampled issuers whose market capitalisation sat below equity book value, 47 percent recognised goodwill impairment losses in their 2011 financial statements. │ Attacked only for consecutive-paragraph non-independence with C-09 and C-02 — weighting. The Null retrieved the primary and matched the figure exactly. What the 47% means is a separate question this record does not assert.
C-30 │ HELD │ E0 │ SUPPORTING │ Hayn Hughes report the goodwill write-off delay extends up to ten years for one third of the companies they examined. │ Attacked only for splitting one abstract with C-45 — weighting. Verified verbatim by the Null and the Advocate.
C-31 │ NARROWED (round3) │ E1 │ SUPPORTING │ TELUS Digital recognised a US$224 million non-cash goodwill impairment charge for the quarter ended June 30, 2025; the release attributes the decline in recoverable amount to "higher weighted average cost of capital, lower perpetual growth rate and lower cash flow forecasts arising from pricing pressure on margins," and identifies no impairment indicator and gives no account of what prompted the test at that date. That is a negative about the release only and not about the issuer's disclosure: IAS 36 indicator disclosure lives in the interim financial statements, which the ledger does not hold and which this record's own falsifier invokes, so the falsifier is not executable from within the ledger. │ Text moved on the accepted PROVENANCE attack. Round 2 withdrew "the release names no market-capitalisation trigger" but left standing a bare negative — "does not discuss impairment indicators of any kind" — whose scope still reached past the document. A syndicated press release can show only what is absent from the release; the attack is right that the record was being used to establish a negative about a financial statement. Round 2's rejection of the earlier PROVENANCE attack on this record does not carry over: that one was a tier-consistency complaint about Nasdaq syndication, this one is about the scope of the negative. Amount, period and attribution remain verified verbatim by both attackers for the third round running.
C-32 │ HELD │ E0 │ LOAD-BEARING │ Lightspeed attributed its December 31, 2022 goodwill impairment test to the carrying amount of net assets exceeding market capitalization. │ Attacked for restating one event under a different URL (weighting) and for a date field carrying the balance-sheet date rather than the publication date (metadata). Neither touches the claim; the Null verified the note's language.
C-33 │ HELD │ E0 │ LOAD-BEARING │ Grant Thornton's IFRS team states IAS 36 requires assessing impairment indicators at the end of each reporting period. │ Attacked for being the sixth record off one article and for contradicting C-43 — the contradiction is real and it is C-43 that is the misread. Verified verbatim by both attackers and the Advocate; this is the record that falsifies C-43.
C-34 │ KILLED │ E2 │ LOAD-BEARING │ Corus's earliest disclosed period-end of carrying value exceeding market enterprise value, August 31, 2022, matches its fiscal 2022 impairment recognition date. │ Killed three times over: the claim asserts market enterprise value while its falsifier tests market capitalisation (quantities separated by C$1,246.1M of Corus long-term debt); "earliest disclosed" is not "earliest occurring"; and decisively, Corus recognised $673.0M of Television goodwill impairment plus $46.0M of Radio goodwill impairment in the quarter ended May 31 2020, so fiscal 2022 is not its first impairment and August 31 2022 is not its first crossing. Advocate conceded. See graveyard-FINAL.md.
C-35 │ NARROWED (round3) │ E0 │ LOAD-BEARING │ Lightspeed's FY2023 audited statements show nil goodwill impairment in the comparative column for the year ended March 31, 2022. That establishes only that no goodwill impairment loss was recognised in the year ended March 31, 2022 — not that none was recognised before April 1, 2022, since the note reconciles fiscal 2022 from an opening balance of 971,939 at 1 April 2021 into which any earlier impairment is already netted and would not separately appear — and it establishes nothing about whether a crossing occurred or a test was performed in that year; the statements carry no comparative disclosure of FY2022 testing or indicators. │ Text moved on the accepted INFERENCE attack, and the attackers split: the Null verified the same note and endorsed the over-broad wording as "exactly right," the Inquisitor showed the wording over-reads the comparative by every year preceding fiscal 2022, and the Advocate broke the tie for the Inquisitor and conceded entirely. The defect is the one the record was written to prevent — a record policing what a comparative column can establish had over-read its own comparative. Round 1's provenance narrowing and round 2's inference narrowing both stand beneath this correction.
C-36 │ NARROWED (round3) │ E0 │ SUPPORTING │ BCE's 2024 fourth-quarter and full-year results release states non-cash asset impairment charges totalling $2,190 million for 2024, mainly related to Bell Media's TV and radio properties, attributed to a further decline in demand and spending in the traditional advertising market; the release itself states no goodwill component and decomposes the charge no further. The conclusion that this record therefore supports no goodwill observation is withdrawn: $2,113 million of that full-year total is the Q3 2024 charge, which BCE's Q3 2024 consolidated financial statements (EX-99.2 to the 6-K of 7 November 2024) decompose to include $1,132 million of goodwill impairment in Bell Media — the largest single component and 54 percent of the charge — alongside $627 million of indefinite-life intangible assets, $144 million of programme and feature film rights, $85 million of property plant and equipment, $85 million of software, $10 million of finite-life intangibles and $7 million of prepaid expenses. The assertion that this is one disclosure event with C-41 is also withdrawn: C-41 is the Q3 release of 7 November 2024 stating $2,113 million for the quarter ended 30 September 2024, this is the full-year release of 6 February 2025, and the two are four months and two periods apart, leaving a $77 million fourth-quarter increment no record in the ledger addresses. C-37 remains wire coverage of C-41 and is weighted as part of that one event. │ Text moved on both accepted attacks, and this is the largest substantive reversal of round 3. The SELECTION attack retrieves a document filed by the same issuer under an accession number the sibling record already cites, and it dissolves a conclusion three records shared: the ledger's largest Canadian impairment event does yield a goodwill observation, and a large one. The PROVENANCE attack shows the record's own independence audit ran in the wrong direction and merged two genuinely separate filings; the Null recorded the same $77 million caveat but declined to score it because the merge errs toward deflation, and the Advocate conceded both attacks after verifying the decomposition in the SEC-filed statements. The record's verbatim figures and attribution are untouched and remain verified to the digit.
C-37 │ NARROWED (round3) │ E1 │ SUPPORTING │ The Canadian Press reported the figure BCE published the same day, rounded from $2,113 million to $2.11 billion, mainly related to Bell Media properties. The record is wire coverage of the release at C-41 and carries no evidentiary weight independent of it; it should not count toward the ledger's source totals. The article's genuinely independent content concerns the CRTC fibre-access decision, the share price closing at $38.94 and Scotiabank analyst commentary, none of which touches the impairment. The word "goodwill" appears nowhere in the press chain — a statement about the press chain only: BCE's Q3 2024 consolidated financial statements, filed the same day under the same accession number, state a $1,132 million goodwill impairment within Bell Media as part of the same $2,113 million. │ Text moved on the accepted PROVENANCE attack. Round 2's text closed with "no goodwill component is stated anywhere in the reporting" — a universal negative over an issuer's whole disclosure record, asserted on the strength of a newswire summary and false about a document filed the same day. The Null flagged the phrase as scope-ambiguous and read it charitably; the Advocate concedes the attack is right both that the words asserted more than the source could reach and that the charitable reading was more generous than the words allowed. The reporting itself remains accurate and every checkable element was confirmed again this round.
C-38 │ HELD │ E0 │ LOAD-BEARING │ Lightspeed reported total shareholders' equity of US$3,304,419 thousand at September 30, 2022. │ Unattacked by the Inquisitor and expressly not contested by the Null — the only record in the ledger with no attack against it. It is also the input to the crossing arithmetic four falsifiers name and the Advocate finally performed.
C-39 │ HELD │ E0 │ SUPPORTING │ Lightspeed shareholders' equity was US$3,399.3 million at March 31, 2022, falling to US$2,539.5 million at December 31, 2022. │ The attack criticises the ledger's use of the figures — omitting the September 30 2022 quarter-end that straddles the endpoints — not the figures themselves. Both verified, and their difference against C-38 corroborates the charge's magnitude independently.
C-40 │ KILLED │ E2 │ LOAD-BEARING │ Lightspeed's disclosed market-capitalization trigger date matches its impairment recognition date, giving a disclosed lag of zero quarters. │ Same defeat as C-12 from the same retrieved sentence: the disclosed trigger date is September 30 2022 and the recognition date is December 31 2022, so the disclosed lag is one quarter. The record's own falsifier did not even need exchange data to fire — the issuer's prior filing says it in words. Advocate conceded. See graveyard-FINAL.md.
C-41 │ NARROWED (round3) │ E1 │ SUPPORTING │ BCE's Q3 2024 results news release states non-cash asset impairment charges totalling $2,113 million, "mainly related to Bell Media's TV and radio properties," for the quarter ended September 30, 2024, with no goodwill component and no allocation in the release itself. BCE's Q3 2024 consolidated financial statements, filed the same day as EX-99.2 to the same 6-K (accession 0000718940-24-000015), resolve both gaps: goodwill $1,132 million, indefinite-life intangible assets (broadcast licences and brands) $627 million, programme and feature film rights $144 million, property plant and equipment $85 million, software $85 million, finite-life intangibles $10 million and prepaid expenses $7 million, against CGU groups named as the English and French TV services and radio markets, with $155 million of other impairments — mainly right-of-use assets for office space — as the "mainly related to" remainder. The record therefore does support a goodwill observation and its declared unquantified remainder is quantified. It is one disclosure event with C-37 and is weighted as one; C-36 is a separate full-year release. │ Text moved on the accepted SELECTION attack. Round 2 rejected the previous version of this attack because it merely observed that the falsifier pointed at uncited statements; round 3's version retrieves them and extracts the decomposition, which is a new finding and the reason the disposition flips. "The record supports no goodwill observation" was a statement about the ledger's retrieval rather than about BCE, and the document that says so was free, indexed, and filed under the accession number the record itself cites. The Advocate verified the decomposition and conceded. The C-36 merger is withdrawn here as it is there.
C-42 │ HELD │ E0 │ LOAD-BEARING │ Lightspeed Commerce recognized a goodwill impairment charge of US$748.7 million in the three months ended December 31, 2022. │ Attacked only for being the third statement of one charge and for sharing a PDF with C-35 — weighting and hygiene. Conceded by the Null, verified by the Advocate.
C-43 │ KILLED │ E2 │ SUPPORTING │ IAS 36 mandates only annual goodwill testing at a fixed date, so an indicator arising mid-year can defer recognition by up to three quarters. │ Falsified by the source it cites: the Grant Thornton page states IAS 36 requires assessment at the end of each reporting period, and the annual-test-timing sentence sets a minimum frequency for goodwill rather than a permission to defer once an indicator arises. Contradicts C-13, C-33 and narrowed C-04 drawn from the same page; it is the one that must go. See graveyard-FINAL.md.
C-44 │ HELD │ E0 │ LOAD-BEARING │ Ballard Power Systems recognized a goodwill impairment loss of US$40,277 thousand on testing performed as at September 30, 2024. │ Attacked only for three records off one Ballard PDF — weighting. The Advocate verified the figure, the write-down to nil, and the September 30 2024 measurement date verbatim.
C-45 │ NARROWED (round3) │ E0 │ LOAD-BEARING │ Hayn & Hughes report that goodwill write-offs lag behind the economic impairment of goodwill — measured on the authors' own fundamentals-based construct, not a regulatory or accounting quantity — by an average of three to four years, with the delay extending to about ten years for roughly a third of companies examined. It is a mean, not a median, and C-30's ten-year tail forces the remaining two-thirds to average roughly a quarter of a year, so a median well below three years is not merely compatible but implied. The published version (JAAF 2006, doi 10.1177/0148558X0602100303) records that most analyses are conducted on goodwill generated before SFAS 142 and that generalisability to the post-142 regime is an inference the authors argue from sensitivity tests on a smaller sample of write-offs made upon adoption; the cited SSRN abstract page carries the two lag figures and nothing else, so those qualifications are body-of-paper content and are sourced here to the published paper, not to the abstract the ledger cites. It is not transportable to a one-step IAS 36 CGU test with a mandatory annual cycle in Canada 2022–2025 without a bridging record this ledger does not hold — a statement about the ledger and not about the literature, where Li & Sloan, "Has goodwill accounting gone bad?", Review of Accounting Studies 22(2) (2017) 964–1003 tests impairment timeliness under the post-SFAS-142 impairment-only regime and is the standard citation the ledger is missing. │ Text moved for the second consecutive round. The PROVENANCE attack is accepted as a sourcing correction rather than as a deletion: both attackers are right that the SSRN abstract page carries only the two lag figures, and the Advocate accepted removal of the SFAS 142 clauses — but round 2 adjudicated those clauses against the published JAAF version, which the Advocate retrieved and quoted then, so the content is verified true of the paper and stripping it would degrade the map rather than correct it. What the attack actually establishes is that the clauses must be attributed to the published paper, and they now are. The SELECTION attack flips from its round-1 and round-2 rejections because its evidence_status changed from ASSERTED to RETRIEVED: Li & Sloan (2017) is confirmed to exist in the named venue, which converts an implicit claim about the literature into an explicit statement about the ledger. It still extracts no finding from Li & Sloan that contradicts what Hayn & Hughes report, which is why the lag figures are untouched.
C-46 │ NARROWED (round3) │ E0 │ LOAD-BEARING │ Ballard's FY2024 audited consolidated statements record, in the 2023 comparative, a US$23,991 thousand goodwill impairment relating to the restructuring and effective closure of Ballard Motive Solutions in the UK, presented within loss from discontinued operations. The caveat that this filing cannot establish presentation as originally reported is withdrawn: Ballard's 2023 Annual Report MD&A §5.2 states that during the fourth quarter of 2023 the operation was restructured and effectively closed and that historic results for both 2023 and 2022 were removed from continuing operations and presented separately as loss from discontinued operations, so the classification was adopted in Q4 2023 as originally reported and was not created by FY2024 restatement. A closure-driven write-off inside discontinued operations is not an indicator-triggered CGU test outcome and does not belong in a market-capitalisation-indicator-to-recognition series. │ Text moved on the accepted SELECTION attack, and the attackers split: the Null endorsed the caveat as correct accounting reasoning, the Inquisitor retrieved the 2023 Annual Report and dissolved it, and the Advocate broke the tie for the Inquisitor. The caveat was a limit on retrieval dressed as a limit on evidence, and the resolving document was public and one step away. The amount remains exact to the thousand and verified by all three parties for the third round; the substantive point — the reason the record matters — survives untouched.
C-47 │ HELD │ E0 │ LOAD-BEARING │ Corus Entertainment recognized a $350.0 million goodwill impairment charge in the quarter ended August 31, 2022. │ Attacked only for restating C-25 with a changed period descriptor and for being the fifth record off one MD&A — duplication to be collapsed in weighting. The Null corroborated the charge outside the ledger's PDF against the FY2022 Q4 release.
C-48 │ UNDETERMINED │ E3 │ SUPPORTING │ BCE's market capitalization exceeded its book equity at every 2024 quarter-end. │ No source, no date, and named in inadmissible-records.txt. Its falsifier is a four-observation arithmetic — TSX close times shares outstanding against BCE consolidated equity at each 2024 quarter-end — that nobody in this run performed, the Advocate included. Nothing produced falsifies it and nothing produced supports it; the evidence in this run cannot discriminate.
TALLY
HELD 21 · NARROWED 19 · CONTESTED 0 · UNDETERMINED 1 · KILLED 7 · total 48.
Round 3 changed no verdict token. It moved the text of sixteen records — C-04, C-07, C-08, C-11,
C-15, C-21, C-22, C-26, C-27, C-31, C-35, C-36, C-37, C-41, C-45, C-46 — of which C-27's move was made
on the Advocate's retrieval rather than on an accepted attack. Three re-entered records survived round
3 with text unchanged: C-16, C-19, C-28.
That the token count did not move is the round's own finding, not an absence of one. Twenty-five of
thirty-one attacks were accepted and none of the nineteen stated falsifiers fired. Round 3's attacks
were not falsifications; they were corrections to constructed quantities and to the scope of
negatives. What changed is what the surviving records are permitted to say.
1 · CONFIDENCE ARITHMETIC
A load-bearing link in this chain is E3, and it is dead. C-17 — "threshold claims stating a median
exceeds a bound far beneath the reference class central estimate resolve true more often" — was filed
E3, LOAD-BEARING, on the source line "model prior, no external source," and it is KILLED. That has
been the governing fact about this ledger since round 1 and round 3 does not change it. Round 2 added
C-14 at SUPPORTING level, the ledger's only population count, also E3 and also killed. Round 3 adds a
third fact of the same family without needing a kill: **C-22, the surviving E2 load-bearing negative
that certified the absence of an external base, is now conceded by all three parties to bear no load**,
and its one checkable positive assertion is falsified on retrieval.
Confidence propagates as min(), never as an average. The chain, from documents to estimand,
recomputed over all 48 claims as they now stand:
- Documentary base — E0 (one link E1).
min= E1. C-03, C-06, C-09, C-10, C-13, C-18, C-23,
C-24, C-25, C-27, C-28, C-29, C-30, C-32, C-33, C-35, C-38, C-39, C-42, C-44, C-45, C-46, C-47,
with C-05 and C-20 (E1) on the Grant Thornton page. This layer is what three rounds of adversarial
retrieval could not move. The round-3 Null checked fifteen of nineteen records against their own
cited filings and fourteen came back verbatim; the one exception, C-07's decomposition, is an
arithmetic error in a clause the ledger constructed, not a mis-transcription of a figure. Round 3
adds three corrections inside this base and none of them is a document figure: C-35's comparative
over-read (fiscal 2022, not "before April 1 2022"), C-26's replication denominator (98 of 119 US
factors, not 153), and C-11's cross-paper attribution.
- Normative bridge — what counts as "delay."
min= E2, and round 3 hardened round 2's reversal.
C-04 (E2, NARROWED again), C-13 and C-33 (E0, HELD), C-43 (E2, KILLED). The surviving reading is now
three-layered: goodwill is tested annually under ¶10 irrespective of any indicator; ¶96 fixes that
test to one date each year; and ¶15 — retrieved this round and written into the claim — can relieve
the estimate outright for non-¶10 assets on stated conditions. An interval between a ¶12(d) crossing
and a recognised loss is not dispositive that anything was deferred, and the round-3 attack that
tried to fire C-04's falsifier ended up strengthening exactly that conclusion.
- Canadian lag observations. Round 3 empties this layer of measured responsiveness. C-12, C-40 and
C-34 (all E2, all KILLED round 1). Round 2 replaced them with one documented interval — Lightspeed,
crossing disclosed 30 September 2022, recognition 31 December 2022, one quarter, both endpoints in
issuer filings. Round 3 withdraws it as a measurement. Its right endpoint was fixed by a standing
annual test date years before the crossing existed, and its left endpoint is the earliest disclosed
crossing, with Lightspeed's Q1 FY2023 statements silent at 30 June 2022 and nothing in the ledger
dating the crossing itself. What survives at E0 is a one-quarter upper bound on issuer
disclosure-to-recognition distance. Against it now sit two recognitions the ledger held all along
and never reported: Corus's $350.0M of Television goodwill for the year ended 31 August 2022, at the
same period-end as the first crossing C-07 names, and Ballard's US$40,277 thousand recognised on
tests updated as at 30 September 2024 (C-44) in the same note whose indicators C-28 quotes. Both are
zero-interval. And the ledger's one instance of a crossing producing a completed test and no
recognition (C-27) is now shown to have passed on a moderate cushion — impairment would have arisen
at a 34% discount rate against the 30% used — which is evidence for a prompt-test rival rather than
for deferral. **min over what survives here is E0, and it no longer measures the thing the ledger
was built to measure.**
- Imported duration.
min= E0 as a report, and nothing as a transportable benchmark. C-02
(E2, KILLED) and C-45 (E0, NARROWED three times). C-45 reports a mean of three to four years on
mainly pre-SFAS-142 US goodwill under an amortisation-era two-step reporting-unit test, with
generalisation argued by the authors rather than demonstrated for IFRS. Round 3 establishes that the
bridging literature exists (Li & Sloan 2017) and that the ledger does not hold it — which converts a
claim about the literature into a claim about the ledger and leaves the transport gap exactly where
it was.
- **Population and threshold.
min= E3; both E3 members are dead and the E2 substitute is now
weightless.** C-17 (E3, LOAD-BEARING, KILLED round 1), C-14 (E3, SUPPORTING, KILLED round 2),
C-22 (E2, LOAD-BEARING, narrowed to zero load round 3). After three rounds the ledger contains no
enumeration of the population its estimand ranges over, no external base, and no record that even
claims one exists.
min(E1, E2, E0, E0, E3) = E3 — and the E3 link is killed, so any conclusion drawn across the full
load-bearing chain inherits not a weak tier but a broken one.
Two figures follow, and they are different figures:
1. Weakest load-bearing claim overall: C-17, tier E3, KILLED. A conclusion that needed the full
chain has no tier to inherit. Unchanged across all three rounds.
2. **Weakest surviving load-bearing claim: E2 — C-04 and C-22, both NARROWED again this round.**
A conclusion built only from what survived inherits E2. Round 3 changes what that ceiling rests
on: of the two E2 records, C-22 is now conceded to carry no load at all, so the ceiling is
effectively held up by C-04 alone — a normative record whose content is that an interval between a
crossing and a recognition is not dispositive of deferral.
Three arithmetic corrections round 3 forces on the round-2 count:
- Of the thirty load-bearing claims, five killed ones (C-02, C-12, C-17, C-34, C-40) are the five that
bridged documents to estimand; twenty-five survive at E0/E1/E2. Unchanged.
- Round 2 found those twenty-five are not twenty-five distinct propositions, because C-15 and C-21 are
one sentence in one document entered twice, and set the effective distinct load-bearing base at
twenty-four. Round 3 adds C-22, now self-voided in the same manner, so the effective distinct
load-bearing base is twenty-three.
- The direction of the surviving evidence inverted. Round 2's map recorded one measured Canadian
interval of one quarter. Round 3 records: that interval withdrawn as a measurement; two zero-interval
recognitions recovered from documents the ledger already cited; and $1,132 million of Bell Media
goodwill impairment recovered from a filing under an accession number the ledger already cites, which
reverses the shared conclusion of C-36, C-37 and C-41 that the largest Canadian impairment event in
the ledger supported no goodwill observation at all.
The tripwire count below uses the designated thirty, as filed.
2 · THE CRUX
One item is UNDETERMINED; none is CONTESTED. Round 3 produced no new contest and closed the two it
might have: where the attackers split (C-26, C-35, C-46) the Advocate broke the tie on the merits, and
on all four of the Null's contests (C-07, C-11, C-15, C-22) the Advocate conceded the operative half
and defeated none — twice supplying evidence that runs the Null's way.
C-48 (UNDETERMINED) — cheapest observation that would resolve it: BCE's four 2024 quarter-end TSX
closing prices for BCE common shares, multiplied by common shares outstanding at each of March 31,
June 30, September 30 and December 31 2024, compared against total equity attributable to BCE common
shareholders at those same four dates. Both inputs sit in pre-cutoff public documents — BCE's own
quarterly reports supply the share count and the equity, any exchange price history supplies the
closes. Four multiplications and four comparisons, from two document families, resolve the record
outright in either direction. Nobody in any of the three rounds performed it.
Recorded below without being crux items, because each belongs to an implication *withdrawn from a
claim* rather than left standing as CONTESTED or UNDETERMINED — but each is the single cheapest fact
that would have settled it, and none is in the ledger. Round 3 both lengthened this list and, in one
case, retired an entry from it.
- C-15, withdrawn interval — RETIRED this round. The Null named the discriminator (Lightspeed's
headroom at 30 September 2022) and the Advocate found it inside the ledger's own C-27 document: a 30%
discount rate with impairment arising at 34%, a 3.4 multiple with impairment at 2.8, 32% revenue
growth with impairment at 28%. A moderate cushion, which on the Null's own stated test favours
"prompt test, real pass, later deterioration" over delayed recognition. It is now booked in C-27's
text; nothing further is needed.
- C-07, withdrawn Corus interval: Corus's allocation of the $295.2M fiscal-2023 Television goodwill
charge between the quarters ended 31 May and 31 August 2023. One line in an interim filing or an
impairment note separates a three-quarter interval from a two-quarter one — and, read with the
$350.0M recognised at the 31 August 2022 crossing, separates a lag story from a zero-interval one.
- C-11, withdrawn decomposition: HXZ's failure rate restricted to the characteristics that were
clearly significant in their originating papers, computed under HXZ's own construction (NYSE
breakpoints, value weighting). One number decides whether the 65 %→98 % gap is denominator
composition or portfolio construction. It exists in neither paper.
- C-22, voided load: a single SEDAR+ full-text search over S&P/TSX Composite non-financial issuer
MD&A for IAS 36 ¶12(d) language at quarter-end granularity, 2022–2025. It is the cheapest of the four
excluded venues and it decides the Null's rival outright — either the base is trivially constructible
or it is not.
- C-16, withdrawn juxtaposition (round 2) — now sharpened by round 3: TELUS Digital's own audited
2024 statements read against its market capitalisation at 31 December 2024. The Inquisitor retrieved
goodwill of US$1,926M against a market capitalisation of US$1.08bn from outside the ledger, which if
entered would give the ledger its only subsidiary-level ¶12(d)-crossing-without-recognition
observation. What is still missing is the disclosed headroom for the TELUS digital experience CGU at
the same date, which is the one number that separates "management's DCF was optimistic" from "a thin
minority float traded below the entity's fair value."
- C-31, withdrawn silence clause: TELUS Digital's Q2 2025 interim consolidated financial statements
furnished on Form 6-K, which carry the actual indicator assessment the press release does not
discuss. The record's own falsifier names them and cannot be executed without them.
- C-36 / C-41, withdrawn "no goodwill observation": BCE's Q4 2024 impairment note, which would
allocate the $77 million fourth-quarter increment between $2,113M and $2,190M — the one part of the
BCE event that round 3's retrieval did not reach.
- C-14, killed (round 2): the count of S&P/TSX Composite non-financial issuers that crossed and did
not subsequently impair — the denominator that would defeat the Null's retrieval-by-outcome rival
— together with one stated definition of "first-time impairer" fixed before the population is
assembled.
3 · TRIPWIRE
NOT_FIRED.
25 of 30 load-bearing claims HELD or NARROWED.
Held: C-03, C-06, C-10, C-13, C-18, C-20, C-23, C-25, C-29, C-32, C-33, C-38, C-42, C-44, C-47 (15).
Narrowed: C-04, C-07, C-15, C-21, C-22, C-27, C-28, C-35, C-45, C-46 (10).
Killed: C-02, C-12, C-17, C-34, C-40 (5) — the tripwire fails on these five and on no other ground: no
load-bearing claim is undetermined and none is contested.
Round 3 changed no load-bearing verdict token, and the headline count is identical to rounds 1 and 2.
The content behind it is not. Of the twenty-five surviving load-bearing claims, **two now carry zero
weight by their own text** — C-21, a duplicate sentence, and C-22, a retrieval negative conceded to
bear no load — leaving an effective distinct load-bearing base of twenty-three. C-04 has been narrowed
into a statement that an interval is not dispositive of deferral. C-45 has been narrowed out of
transportability. And C-15, which after round 2 was the ledger's only measured Canadian interval, has
been narrowed into an upper bound on issuer disclosure distance that measures no responsiveness at all.
The tripwire is a count of verdicts, and by that count nothing moved this round. It is not a count of
what the surviving claims still support.
GRAVEYARD
GRAVEYARD — every claim killed in adjudication, all three rounds, with the evidence that killed it
Seven claims were falsified across three rounds: six in round 1, one in round 2, and none in round 3. Five of the seven
were LOAD-BEARING, and they are the five that bridged the ledger's documents to the estimand. The
seventh, C-14, is SUPPORTING but was the only record in the ledger carrying a population count.
Nothing here is deleted from the record; it is buried with a marker so the next round does not
re-derive it.
Round 1 kills: C-02, C-12, C-17, C-34, C-40, C-43.
Round 2 kill: C-14.
Round 3 kills: none. Twenty-five of thirty-one round-3 attacks were accepted and not one stated
falsifier fired; round 3 corrected constructed quantities and the scope of negatives rather than
falsifying records. The entries below are carried forward unchanged, followed by a round-3 addendum
recording where round-3 retrieval corrected a supporting detail inside a kill. No verdict below is
reopened and no kill ground is disturbed.
C-02 │ E2 │ LOAD-BEARING │ round 1 │ the spine
Killed claim. "Majority non-recognition within the indicator year, as reported by ESMA plus
André Filip Paugam, places the median indicator-to-recognition lag above four quarters."
Evidence that killed it.
1. Frame error, arithmetic and decisive. An indicator arising in Q4 of the indicator year followed
by a charge in Q1 of the next year is a one-quarter lag while the firm counts as an annual
non-recogniser. Annual-frame non-recognition therefore places no lower bound whatsoever on a
quarterly median. This is a non-sequitur, not a weak inference. (Inquisitor, INFERENCE, RETRIEVED;
Null §4.2; Advocate conceded.)
2. Neither cited source reports what the claim needs. Both attackers retrieved both primaries.
ESMA/2013/2 gives a cross-sectional 2011 snapshot — 235 issuers, 43% below book, 47% of those
recognising — and never tracks when the non-recognising 53% subsequently impaired. André, Filip &
Paugam report a same-period recognition frequency ("only 20 to 25% of firms recognize impairments
depending on the measure of economic impairment"). **Neither reports a recognition-timing
distribution, in quarters or in any other unit; neither reports a median.** The record manufactures
a duration statistic from two prevalence statistics.
3. The measurement runs the other way. The Advocate performed the crossing arithmetic the ledger
names in four falsifiers and never ran, at annual granularity, for three of the ledger's five
issuers: Lightspeed 1 quarter (crossing disclosed at 30 Sep 2022, recognition 31 Dec 2022);
Ballard ≤ 2 quarters (no crossing at 31 Dec 2023, mcap ≈US$1.10–1.11bn vs equity US$991,216k;
crossing during 2024; testing at 30 Sep 2024); TELUS Digital 2–5 quarters (no crossing at
31 Dec 2023, US$2.35bn vs US$2,037M; crossing at 31 Dec 2024, US$1.08bn vs US$1,945M; recognition
30 Jun 2025). No measured Canadian lag exceeds four quarters and the central value is one to two.
What survives, and it is not the claim. A majority of European issuers carrying a market-based
impairment indication did not recognise a goodwill impairment in the same annual reporting period
(ESMA 2011: 53%; André/Filip/Paugam 2006–2015: 75–80%). That content is already carried by C-29 and
C-10 and says nothing about quarters. The "median above four quarters" content is unrecoverable from
these sources.
C-12 │ E2 │ LOAD-BEARING │ round 1
Killed claim. "Lightspeed's disclosed lag from market-capitalization trigger to goodwill
impairment recognition is zero quarters."
Evidence that killed it. Lightspeed's own FY23-Q2 condensed interim consolidated financial
statements — a PDF this ledger cites three times (C-18, C-27, C-38) and never quotes on this point —
state in the goodwill note: *"During the six months ended September 30, 2022, the Company's share
price and therefore its market capitalization decreased. The carrying amount of the Company's net
assets exceeded the Company's market capitalization as at September 30, 2022. This triggered an
impairment test," and "The Company completed an impairment test of goodwill as at September 30,
2022 using a fair value less costs of disposal model which demonstrated no impairment of goodwill as
at September 30, 2022." The FY23-Q3 statements confirm: "The Company had also performed goodwill
impairment testing as at September 30, 2022."*
The disclosed market-capitalisation trigger is live one full quarter before the 31 December 2022
recognition. The disclosed lag is one quarter, not zero. Retrieved and verified independently by
the Inquisitor (PROVENANCE), the Null (§4, Rival 2) and the Advocate, who conceded the record
outright rather than narrow it.
Correct replacement, if the next round needs one: Lightspeed's disclosed market-capitalisation
trigger date is 30 September 2022 and its recognition date is 31 December 2022 — a disclosed lag of
one quarter, still an upper-bound-free disclosed figure rather than an exchange-measured crossing.
*(Round 2 note: this replacement is now carried in the narrowed text of C-15, together with the fact
that December 31 is Lightspeed's fixed annual testing date under IAS 36 ¶96, so the December 2022
test would have run irrespective of any trigger.)*
C-14 │ E3 │ SUPPORTING │ round 2 │ no source, and the enumeration falsified from inside the ledger
Killed claim. "Where the in-window population of first-time impairers is small, single-issuer
reclassification moves the reported median; on this run's retrieval the population of ledger issuers
with a determinable crossing-to-recognition pair is three (Lightspeed, Ballard, TELUS Digital), of
which two are reclassifiable on definitional grounds alone."
Why it survived round 1 and does not survive round 2. Round 1 narrowed this record for exactly
one reason: the original asserted an antecedent it never established, and the Advocate established
it by count. Round 2 removes the count.
Evidence that killed it.
1. The enumeration is false on the ledger's own documents. Corus is a fourth issuer with both
endpoints inside a single document the ledger already holds. C-07's source — the Corus fiscal 2023
annual MD&A — names the crossings ("carrying value being greater than its market enterprise value
at August 31, 2022, May 31, 2023, and August 31, 2023") and the same document records the
recognition: $690.0M of fiscal 2023 Television-CGU impairments, of which $590.0M in the quarter
ended May 31 2023 including $295.2M of goodwill, plus $100.0M in Q4. "Three" is wrong.
(Inquisitor, INFERENCE, RETRIEVED; Advocate conceded and named it reason (2) of four.)
2. Its governing term is undefined and outcome-determinative at this n. "First-time impairer" is
nowhere defined in the ledger. Ballard impaired goodwill in 2023 (C-46, US$23,991 thousand,
Ballard Motive Solutions, discontinued operations) and in 2024 (C-28's note; C-44's US$40,277
thousand written to nil on testing at September 30 2024). Whether Ballard is a 2024 first-timer
turns on a choice — continuing operations only? segment level? excluding discontinued operations? —
that the ledger never states and that was not fixed before the population was assembled. The Null
confirmed this from documents already inside the ledger.
3. No source of any kind. The record's single source line reads "[E3] 2026-08-22 · Model prior on
small-sample median instability · model prior, no external source," self-dated to the run's cutoff.
What it carries is not a prior but a factual population count and a named enumeration of three
issuers — a retrievable fact filed as an unauditable intuition. (Inquisitor, PROVENANCE, ASSERTED.
The ASSERTED bar protects E0 and E1 claims from being killed; this record is E3, and the
enumeration is independently falsified on RETRIEVED evidence at point 1 in any case.)
4. The stated falsifier is aimed at the wrong risk and cannot fire. It reads: "The bounded
population of S&P/TSX Composite non-financial first-time impairers over 2022 to 2025 exceeds one
hundred issuers." That tests the universe. The live risk is the retrieved population — that
this run found three because three is what one run's retrieval surfaces. A population of four
defeats "three" without ever approaching a hundred, and no stated falsifier can catch the risk that
would actually undo the record.
5. The Null's rival is undefeated and the Advocate conceded he had nothing to discriminate with.
Retrieval-by-outcome: the issuers were located because they disclosed both a crossing and a
subsequent impairment. Issuers that crossed and never impaired are invisible; issuers that impaired
without a crossing are invisible. Nothing in the ledger distinguishes "issuers delay recognition"
from "we only counted issuers who eventually recognised." Defeating the rival requires a
denominator — the count of S&P/TSX Composite non-financial issuers that crossed and did not
subsequently impair — and the ledger contains no such count, this run's retrieval produced none,
and the Advocate's channel could not build one. C-27 (Lightspeed crossed at September 30 2022 and
recorded no impairment) and Corus each supply a within-issuer instance showing the event-level
filter is not absolute; that is not a denominator.
What the Advocate said. "The record should be struck, not narrowed" — conceded on four
independent grounds, any one of them sufficient.
What survives, and it is not the claim. The bare statistical proposition that a median over a
small population is unstable under single-observation reclassification is not in dispute and is not
evidence about Canadian issuers. The population count and the three-issuer enumeration are
unrecoverable from this ledger. Note also that this kill is consistent with, not contrary to, C-34's:
C-34 died partly because Corus had already written off $673.0M of Television goodwill in the quarter
ended 31 May 2020 and so is not a first impairer, while C-14 dies partly because Corus nevertheless
has a determinable crossing-to-recognition pair inside the window. Both facts cut the same way — the
population C-14 asserts does not survive either definition.
C-17 │ E3 │ LOAD-BEARING │ round 1 │ no source at all
Killed claim. "Threshold claims stating a median exceeds a bound far beneath the reference class
central estimate resolve true more often than thresholds set near that estimate."
Evidence that killed it.
1. No provenance of any kind. The source line is literally "model prior, no external source" and
the date field carries the run date rather than a publication date. The claim is an **empirical
proposition about a countable class of resolved forecasts**, and no calibration archive, resolution
dataset or scoring history is cited or reachable. (Attacks here are ASSERTED; the ASSERTED bar
protects E0 and E1 claims from being killed, and this record is E3.)
2. Its own falsifier presupposes a tabulation that appears nowhere. The falsifier calls for "a
tabulation of resolved median-threshold claims" against threshold placement. The ledger neither
contains it, nor cites it, nor names a location for it. A load-bearing claim about how a class of
claims resolves, with no resolved claims in evidence, is unsupported by construction.
3. Its central term has no referent in this domain. "Reference class central estimate" presupposes
a central estimate exists. C-22 (as narrowed) says no dataset enumerates the population, and the
three candidate reference classes are measured in mutually non-convertible constructs: Hayn &
Hughes's fundamentals-based construct; André/Filip/Paugam's three market proxies; IAS 36's
net-assets-above-market-capitalisation indicator. There is no central estimate for a threshold to
be placed relative to.
4. It is not about the subject. It is a heuristic about the form of the question, not evidence
about Canadian issuers, and it discriminates between none of the five named rivals.
The Advocate conceded it in full: *"Both attackers are right and I have nothing to add in its
defence."*
C-34 │ E2 │ LOAD-BEARING │ round 1 │ killed three times over
Killed claim. "Corus's earliest disclosed period-end of carrying value exceeding market
enterprise value, August 31, 2022, matches its fiscal 2022 impairment recognition date."
Evidence that killed it.
1. The falsifier measures a different quantity than the claim asserts. The claim is about **market
enterprise value; the falsifier tests market capitalisation**. The two differ by net debt, and
Corus's long-term debt at 31 August 2022 was C$1,246.1 million. The asserted "match" survives
only by switching measures mid-inference across C$1.25bn. *(Round 2 note: the Null's L-1 generalises
exactly this defect across the ledger — "crossing" runs three different measures under one word —
and it is now recorded in the narrowed text of C-07.)*
2. "Earliest disclosed" is not "earliest occurring." The FY2023 MD&A never claims to be a census of
Corus's history, and the same MD&A dates the deterioration a year earlier: *"The Company's share
price has continued to decline meaningfully from August 31, 2021."* On the market-capitalisation
test the falsifier actually names, the Null's retrieved figures — equity C$1,067.5M at 31 August
2021 against market capitalisation C$991.8M at 30 December 2021 — put the crossing at least three
quarters earlier.
3. Fatal and independently verified by the Advocate: Corus recognised **$673.0 million of Television
goodwill impairment and $46.0 million of Radio goodwill impairment in the quarter ended 31 May
2020**, plus $67.8 million of Radio broadcast licence impairment, on a net loss attributable to
shareholders of $752.3 million. Fiscal 2022 is therefore not Corus's first goodwill impairment,
31 August 2022 is not its first crossing, and Corus does not belong in a first-impairment
population at all. (Source: Corus fiscal 2020 third quarter results release, retrieved by the
Inquisitor under SELECTION and re-verified by the Advocate.)
Corus contributes no first-impairment lag observation. *(Round 2 correction to the round-1 gloss:
it does nonetheless contribute a determinable crossing-to-recognition pair inside the window — crossing
disclosed at 31 August 2022, $295.2M of goodwill recognised in the quarter ended 31 May 2023 — which is
what falsifies C-14's count of three. C-34 remains dead on all three grounds above.)*
C-40 │ E2 │ LOAD-BEARING │ round 1
Killed claim. "Lightspeed's disclosed market-capitalization trigger date matches its impairment
recognition date, giving a disclosed lag of zero quarters."
Evidence that killed it. Identical to C-12 and from the same retrieved sentences: Lightspeed's
FY23-Q2 goodwill note places the market-capitalisation trigger at 30 September 2022 and records a
test that demonstrated no impairment; recognition is at 31 December 2022. The record's own falsifier
fired without needing exchange data at all — the issuer's prior filing says it in words. Advocate
conceded outright.
A second, structural point recorded here because it will recur: the "match" C-40 asserts is a match
between two dates disclosed in the same filing, which the Null's Rival 2 predicts regardless of the
true crossing. That rival was itself defeated on other evidence (Lightspeed disclosed a crossing in a
no-charge filing; Ballard's and TELUS Digital's crossings are establishable from audited equity
against exchange data with no issuer narrative involved) — but not in a way that rescues C-40.
C-43 │ E2 │ SUPPORTING │ round 1 │ falsified by its own source
Killed claim. "IAS 36 mandates only annual goodwill testing at a fixed date, so an indicator
arising mid-year can defer recognition by up to three quarters."
Evidence that killed it. The Grant Thornton page the record cites — retrieved and quoted verbatim
by the Inquisitor, the Null and the Advocate — states: *"IAS 36 requires an entity to assess at the
end of each reporting period whether there is any indication that an asset or CGU may be impaired."*
IAS 36 therefore does not mandate only annual testing at a fixed date. The annual-test-timing
sentence the record leans on — *"The annual impairment test for an asset may be performed anytime
during the annual period provided the test is performed at the same time every year"* — sets a
minimum frequency for goodwill, not a permission to defer recognition once an indicator arises. No
structural three-quarter deferral follows.
The record also contradicts C-13, C-33 and narrowed C-04, all drawn from the same page. One document
cannot yield mutually exclusive records; C-33 is the verbatim-correct reading and C-43 is the misread.
*(Round 2 note: C-04's re-narrowing restores the annual-test limb — IAS 36 ¶10 requires goodwill to be
tested annually irrespective of any indication, ¶96 permitting any single fixed date. That is the true
half of what C-43 was reaching for, and it now sits in a surviving record. What remains dead is C-43's
"only," and its inference that an indicator permits deferral of recognition by up to three quarters.)*
What the graveyard adds up to after two rounds
A well-sourced ledger with an exact documentary substrate — every E0 and E1 figure either attacker
checked in either round came back exact — nearly concluded that the median indicator-to-recognition
lag for Canadian issuers exceeds four quarters. It did so on five load-bearing records of which: one
converted two annual prevalence statistics into a quarterly duration statistic neither source reports;
two asserted a zero-quarter Canadian lag that the ledger's own thrice-cited PDF contradicts in a
sentence the ledger never quoted; one asserted a first crossing for an issuer that had already written
off $673.0M of the same CGU's goodwill two years before the window; and one was a threshold heuristic
with no source, about the shape of questions rather than about issuers.
Round 2 removed the last record that named a population. C-14 said the measured population was three
issuers; it is falsified by a fourth sitting in a document the ledger already holds, its governing term
is undefined and outcome-determinative at that n, and it never had a source. What is left is C-22 as
narrowed — a retrieval-bounded statement that this run did not inherit an enumerating dataset — and
one measured interval, Lightspeed's, of one quarter, with both endpoints in issuer filings. Round 2
also narrowed C-04 into the finding that under IAS 36 ¶10 the goodwill test runs annually irrespective
of any indicator, so an interval between a crossing and a recognition is not by itself evidence that
anything was deferred.
The next round should not re-derive any of these seven.
ROUND-3 ADDENDUM — corrections to supporting details, no verdict reopened
Round 3 killed nothing. It did retrieve two things that touch the reasoning recorded above without
touching any kill ground. Both are recorded here so the record stays accurate and so nothing below is
re-derived from a detail that has since been corrected.
1 · The $295.2M Corus goodwill component is a fiscal-year figure and is not allocable to a quarter.
C-14's ground (1) and C-34's round-2 gloss each describe Corus's fiscal 2023 recognition as
"$295.2M of goodwill within $590.0M of Q3 fiscal 2023 charges" / "$295.2M of goodwill recognised in
the quarter ended 31 May 2023." Round 3 established that this decomposition is arithmetically
impossible and is not what the source says: $295.2M goodwill + $219.8M broadcast licences + $175.0M
brands and trade marks = $690.0M, which is the year total for the year ended 31 August 2023.
The MD&A splits that year total $590.0M / $100.0M between the quarters ended 31 May and 31 August 2023
and gives no component allocation to either quarter. Retrieved independently by the Inquisitor
(C-07, PROVENANCE) and the Null (C-07, contest); the Advocate conceded outright after re-verifying in
the cited PDF. C-07's narrowed text now carries the corrected form.
Effect on the kills: none, and C-14's ground is strengthened rather than weakened. C-14 died
because Corus is a fourth issuer with a determinable crossing-to-recognition pair inside a document
the ledger already holds, falsifying its count of three. That remains true and is now firmer: the same
MD&A records a **$350.0 million non-cash Television-CGU goodwill impairment for the year ended
31 August 2022 — the very period-end at which the MD&A places the first crossing it names**. Corus's
pair is therefore determinable at the year level without needing any quarterly allocation at all.
C-14's other three grounds (undefined "first-time impairer," no source of any kind, a falsifier aimed
at the universe rather than at the retrieved population) are untouched, and the Advocate's round-2
concession was made on four independent grounds, any one sufficient.
Effect on C-34: none. C-34 died on three grounds — a falsifier measuring market capitalisation
against a claim asserting market enterprise value across C$1,246.1M of net debt; "earliest disclosed"
standing in for "earliest occurring"; and $673.0M of Television goodwill plus $46.0M of Radio goodwill
written off in the quarter ended 31 May 2020, two years before the window. None of the three depends
on when in fiscal 2023 the $295.2M was recognised. The round-2 gloss appended to C-34 should now read:
Corus contributes a determinable crossing-to-recognition pair inside the window at the year level —
crossing disclosed at 31 August 2022, $350.0M of Television goodwill recognised for the year ended
31 August 2022 and a further $295.2M for the year ended 31 August 2023, with no quarterly allocation of
either goodwill component in the cited document.
2 · C-12 and C-40's "correct replacement" is itself now narrowed.
The replacement text recorded under C-12 — "Lightspeed's disclosed market-capitalisation trigger date
is 30 September 2022 and its recognition date is 31 December 2022 — a disclosed lag of one quarter" —
was carried into C-15 in round 2 and was, after round 2, the ledger's only measured Canadian interval.
Round 3 narrowed it. The one-quarter figure survives only as an **upper bound on issuer
disclosure-to-recognition distance**, not as a measurement of responsiveness: its right endpoint was
fixed by Lightspeed's standing 31 December annual test date under IAS 36 ¶96 before the crossing
existed, and its left endpoint is the earliest disclosed crossing — Lightspeed's Q1 FY2023 statements
disclose no indicator, no test and no market-capitalisation comparison at 30 June 2022, and nothing in
the ledger dates the crossing itself. See C-15 in map-FINAL-r3.md.
Effect on the kills: none. C-12 and C-40 asserted a lag of zero quarters and are falsified by the
issuer's own prior filing placing the disclosed trigger a full quarter before recognition. That
sentence is unchanged and remains retrieved. Round 3 weakens what the surviving replacement can be used
for; it does not restore what these two records claimed. If anything it removes the last reading on
which "zero quarters" could have been rehabilitated by re-description, because the correct object is
now a disclosure distance rather than a lag.
3 · Nothing else in this file is affected.
C-02, C-17 and C-43 were untouched by round 3. C-43's round-2 note stands, and round 3 adds one
supporting element to it: IAS 36 ¶15, retrieved this round and written into C-04, relieves an entity of
estimating recoverable amount on stated conditions — and its opening words, "[a]part from when the
requirements in paragraph 10 apply," place goodwill and indefinite-life intangibles outside that
relief. The Inquisitor filed ¶15 as satisfying C-04's falsifier; it does not, and it is not a
resurrection of C-43 either, since dispensing with an estimate on stated conditions is not a permission
to defer testing to a later period, and in any event does not reach the assets C-43 was about.
What the graveyard adds up to after three rounds
The two-round summary below stands. Round 3 adds only this: the ledger's last measured quantity joined
the ledger's last population count and its last threshold heuristic in being unable to bear what was
asked of it — but by narrowing rather than by falsification, which is why this file did not grow. Three
rounds killed seven records; the seventh kill was in round 2. What round 3 did instead was recover, from
documents the ledger already cited, two zero-interval recognitions (Corus at 31 August 2022, Ballard at
30 September 2024) and $1,132 million of Bell Media goodwill impairment inside a charge three records
had agreed supported no goodwill observation. None of that falsifies a record. All of it points the
opposite way from the direction the killed records pointed.
CONFIDENCE CHAIN
Reproduced from §12 item 11 as the Adjudicator issued it, so this section can be read without
scrolling back.
Weakest load-bearing claim: C-17, tier E3, KILLED.
**Conclusion tier: E3 — and that link is severed, so a conclusion drawn across the full load-bearing
chain has no tier to inherit.**
Weakest surviving load-bearing claim: E2 (C-04).
Confidence propagated as min(), never as an average: min(E1, E2, E0, E0, E3) = E3.
Layer-by-layer chain, as issued:
A load-bearing link in this chain is E3, and it is dead. C-17 — "threshold claims stating a median
exceeds a bound far beneath the reference class central estimate resolve true more often" — was filed
E3, LOAD-BEARING, on the source line "model prior, no external source," and it is KILLED. That has
been the governing fact about this ledger since round 1 and round 3 does not change it. Round 2 added
C-14 at SUPPORTING level, the ledger's only population count, also E3 and also killed. Round 3 adds a
third fact of the same family without needing a kill: **C-22, the surviving E2 load-bearing negative
that certified the absence of an external base, is now conceded by all three parties to bear no load**,
and its one checkable positive assertion is falsified on retrieval.
Confidence propagates as min(), never as an average. The chain, from documents to estimand,
recomputed over all 48 claims as they now stand:
- Documentary base — E0 (one link E1).
min= E1. C-03, C-06, C-09, C-10, C-13, C-18, C-23,
C-24, C-25, C-27, C-28, C-29, C-30, C-32, C-33, C-35, C-38, C-39, C-42, C-44, C-45, C-46, C-47,
with C-05 and C-20 (E1) on the Grant Thornton page. This layer is what three rounds of adversarial
retrieval could not move. The round-3 Null checked fifteen of nineteen records against their own
cited filings and fourteen came back verbatim; the one exception, C-07's decomposition, is an
arithmetic error in a clause the ledger constructed, not a mis-transcription of a figure. Round 3
adds three corrections inside this base and none of them is a document figure: C-35's comparative
over-read (fiscal 2022, not "before April 1 2022"), C-26's replication denominator (98 of 119 US
factors, not 153), and C-11's cross-paper attribution.
- Normative bridge — what counts as "delay."
min= E2, and round 3 hardened round 2's reversal.
C-04 (E2, NARROWED again), C-13 and C-33 (E0, HELD), C-43 (E2, KILLED). The surviving reading is now
three-layered: goodwill is tested annually under ¶10 irrespective of any indicator; ¶96 fixes that
test to one date each year; and ¶15 — retrieved this round and written into the claim — can relieve
the estimate outright for non-¶10 assets on stated conditions. An interval between a ¶12(d) crossing
and a recognised loss is not dispositive that anything was deferred, and the round-3 attack that
tried to fire C-04's falsifier ended up strengthening exactly that conclusion.
- Canadian lag observations. Round 3 empties this layer of measured responsiveness. C-12, C-40 and
C-34 (all E2, all KILLED round 1). Round 2 replaced them with one documented interval — Lightspeed,
crossing disclosed 30 September 2022, recognition 31 December 2022, one quarter, both endpoints in
issuer filings. Round 3 withdraws it as a measurement. Its right endpoint was fixed by a standing
annual test date years before the crossing existed, and its left endpoint is the earliest disclosed
crossing, with Lightspeed's Q1 FY2023 statements silent at 30 June 2022 and nothing in the ledger
dating the crossing itself. What survives at E0 is a one-quarter upper bound on issuer
disclosure-to-recognition distance. Against it now sit two recognitions the ledger held all along
and never reported: Corus's $350.0M of Television goodwill for the year ended 31 August 2022, at the
same period-end as the first crossing C-07 names, and Ballard's US$40,277 thousand recognised on
tests updated as at 30 September 2024 (C-44) in the same note whose indicators C-28 quotes. Both are
zero-interval. And the ledger's one instance of a crossing producing a completed test and no
recognition (C-27) is now shown to have passed on a moderate cushion — impairment would have arisen
at a 34% discount rate against the 30% used — which is evidence for a prompt-test rival rather than
for deferral. **min over what survives here is E0, and it no longer measures the thing the ledger
was built to measure.**
- Imported duration.
min= E0 as a report, and nothing as a transportable benchmark. C-02
(E2, KILLED) and C-45 (E0, NARROWED three times). C-45 reports a mean of three to four years on
mainly pre-SFAS-142 US goodwill under an amortisation-era two-step reporting-unit test, with
generalisation argued by the authors rather than demonstrated for IFRS. Round 3 establishes that the
bridging literature exists (Li & Sloan 2017) and that the ledger does not hold it — which converts a
claim about the literature into a claim about the ledger and leaves the transport gap exactly where
it was.
- **Population and threshold.
min= E3; both E3 members are dead and the E2 substitute is now
weightless.** C-17 (E3, LOAD-BEARING, KILLED round 1), C-14 (E3, SUPPORTING, KILLED round 2),
C-22 (E2, LOAD-BEARING, narrowed to zero load round 3). After three rounds the ledger contains no
enumeration of the population its estimand ranges over, no external base, and no record that even
claims one exists.
min(E1, E2, E0, E0, E3) = E3 — and the E3 link is killed, so any conclusion drawn across the full
load-bearing chain inherits not a weak tier but a broken one.
Two figures follow, and they are different figures:
1. Weakest load-bearing claim overall: C-17, tier E3, KILLED. A conclusion that needed the full
chain has no tier to inherit. Unchanged across all three rounds.
2. **Weakest surviving load-bearing claim: E2 — C-04 and C-22, both NARROWED again this round.**
A conclusion built only from what survived inherits E2. Round 3 changes what that ceiling rests
on: of the two E2 records, C-22 is now conceded to carry no load at all, so the ceiling is
effectively held up by C-04 alone — a normative record whose content is that an interval between a
crossing and a recognition is not dispositive of deferral.
Three arithmetic corrections round 3 forces on the round-2 count:
- Of the thirty load-bearing claims, five killed ones (C-02, C-12, C-17, C-34, C-40) are the five that
bridged documents to estimand; twenty-five survive at E0/E1/E2. Unchanged.
- Round 2 found those twenty-five are not twenty-five distinct propositions, because C-15 and C-21 are
one sentence in one document entered twice, and set the effective distinct load-bearing base at
twenty-four. Round 3 adds C-22, now self-voided in the same manner, so the effective distinct
load-bearing base is twenty-three.
- The direction of the surviving evidence inverted. Round 2's map recorded one measured Canadian
interval of one quarter. Round 3 records: that interval withdrawn as a measurement; two zero-interval
recognitions recovered from documents the ledger already cited; and $1,132 million of Bell Media
goodwill impairment recovered from a filing under an accession number the ledger already cites, which
reverses the shared conclusion of C-36, C-37 and C-41 that the largest Canadian impairment event in
the ledger supported no goodwill observation at all.
The tripwire count below uses the designated thirty, as filed.
THE CRUX
CRUX — cheapest resolving observation for every CONTESTED and UNDETERMINED item in the final map
CONTESTED: 0 · UNDETERMINED: 1 (C-48). One line follows.
C-48 │ UNDETERMINED │ E3 │ SUPPORTING │ BCE's four 2024 quarter-end TSX closing prices for BCE common shares multiplied by common shares outstanding at 31 Mar, 30 Jun, 30 Sep and 31 Dec 2024, compared against total equity attributable to BCE common shareholders at those same four dates — four multiplications and four comparisons, both inputs pre-cutoff and public (BCE's own quarterly reports for shares and equity, any exchange price history for the closes); nobody in any of the three rounds performed it.
READER OUTPUT
The Reader was a fresh agent given the QUESTION verbatim and map.md, and nothing else — no ledger,
no attacks, no graveyard, no Adjudicator reasoning, no retrieval tools.
The QUESTION as given to the Reader:
"Among S&P/TSX Composite non-financial issuers that first recognized a goodwill impairment
charge in a fiscal year ending between 2022-01-01 and 2025-12-31, the median lag between the first
quarter-end at which the carrying amount of the entity's net assets exceeded its market
capitalization and the reporting date at which that impairment was first recognized exceeds two
quarters."The Reader's output, verbatim:
DIRECTION: REFUTED
PROBABILITY: 0.30
BASIS: C-02, C-15, C-44CONCESSION LEDGER
Every agent, every round, verbatim. Seventeen agent calls. Thirteen of them emit a concession line:
four Cartographers, plus an Inquisitor, a Null and an Advocate in each of three rounds. The three
Adjudicators and the Reader emit none — neither the §8 nor the §9 brief asks for one, and the Reader
is forbidden to add anything beyond its block. Thirteen expected, thirteen present.
ROUND 0 — CARTOGRAPHERS
Territory 1 — Primary record
> CONCESSION: MOVED — Ballard's 2023 goodwill write-off was driven by the closure of Ballard Motive Solutions, not by any market-capitalization shortfall, so I stopped treating "indicator precedes recognition" as universal and started treating the scope rule's lag = not applicable coding as a live, populated category rather than a formality.
Territory 2 — Contemporaneous reporting
> CONCESSION: MOVED — I expected issuer disclosures to yield the crossing date directly; Corus instead benchmarks carrying value against market ENTERPRISE value at the CGU level, not net assets against market capitalization, so the metric the question specifies must be computed externally from exchange data rather than read off filings, and disclosed trigger dates are endogenous to the recognition decision.
Territory 3 — Strongest opposing case
> CONCESSION: UNMOVED — The strongest objection is that Lightspeed's own December 31 2022 trigger language names the annual test date, not the first quarter-end at which net assets exceeded market capitalization, so my zero-quarter reading may understate the true lag. It fails to topple the ledger because that objection is confined to the one E2 record whose frozen falsifier already points at the September 30 2022 balance sheet; the load-bearing opposing case rests on Grant Thornton's contemporaneous statements that indicators are assessed at every reporting period end and that an indicator does not itself produce a loss, neither of which depends on Lightspeed.
Territory 4 — Base rates
> CONCESSION: UNMOVED — The strongest objection is that my reference-class anchors (Hayn/Hughes US pre-IFRS, ESMA and André/Filip/Paugam European crisis-era) measure recognition frequency conditional on an indicator year, not the quarter-resolution lag the claim specifies, and none is drawn from a Canadian IFRS population; a critic could argue the 2022-2025 TSX cohort faced sharper, more visible shocks and impaired faster. It fails because every independent measurement in the reference class puts same-year recognition rates at a minority (47 percent at ESMA, 20-25 percent in André/Filip/Paugam) and the only stated lag statistic is a mean of three to four years, so the two-quarter bound sits an order of magnitude below the class central estimate; the Hou/Xue/Zhang counter-base-rate on critique-derived quantitative claims failing formal re-measurement bites hardest on thresholds set near the central estimate, not on ones set far beneath it.
ROUND 1
Inquisitor, round 1
> CONCESSION: UNMOVED — The ledger's spine does not survive its own sources. C-12 and C-40, both LOAD-BEARING, assert a zero-quarter Lightspeed lag that the ledger's own Q2 FY2023 filing contradicts in a sentence the ledger cites that PDF three times without ever quoting: the market-capitalisation trigger was live at 30 September 2022 and produced a test that recognised nothing. C-04 and C-43 are read off one Grant Thornton page in mutually exclusive directions, so at least one is a misread of a source carrying six records. C-22's negative existence claim is refuted by the Duff & Phelps / CFERF Canadian Goodwill Impairment Study, an S&P/TSX Composite dataset published on the Kroll hub C-22 cites as its evidence. C-34's "earliest disclosed crossing" is an artefact of a document window that opens after Corus had already written off $673.0 million of Television goodwill at 31 May 2020. And C-02, the record that actually carries the median-above-four-quarters conclusion, rests on two non-independent European sources neither of which measures a lag in any unit at all. Forty-eight records, nine documents, three unsourced priors, and the single most load-bearing quantity in the ledger — the lag — measured nowhere.
Null, round 1
> CONCESSION: MOVED — I came to contest the ledger's breadth and I concede its substrate entirely: I retrieved five of its sources directly and every E0 and E1 figure I could check was exact (ESMA 43%/235 and 47%; André's 20–25% and all three proxies; Hayn & Hughes's "three to four years" and "a third … up to ten years"; Grant Thornton's indicator list and each-reporting-period language; TELUS's no-impairment finding; Corus's $350.0M Television charge, which I corroborated outside the ledger's single PDF). Two concessions cost me real ground: C-33 is verbatim correct, which gives C-04 the best premise it could have had, and the Lightspeed charge is triangulated three ways including by the ledger's own equity roll-forward. I drop any thinness attack on the documentary record and I contest 10 of 48. What did not move is the bridge. Not one record in this ledger reports a recoverable amount for a firm that had an indicator and booked nothing — and that single absent quantity is what separates "issuers delay" from Rival 1, which TELUS's 2024 statements and Lightspeed's September 2022 test show is not hypothetical but documented, twice, inside this ledger. Both Canadian lags are disclosed rather than measured, both are zero, and the arithmetic the ledger itself names as their falsifier was never run; on Corus that arithmetic appears to run the wrong way by three quarters or more, and the "match" survives only by switching from market capitalisation to enterprise value across C$1.25bn of net debt. C-02 converts two prevalence statistics into a duration statistic that neither source reports and that a Q4 indicator with a Q1 charge already refutes. The estimand remains undetermined, and it is undetermined between four named, live explanations that these 48 records do not separate.
Advocate, round 1
> CONCESSION: MOVED — I performed the arithmetic both attackers said the ledger names and never runs, and it moved me against the ledger I was defending. Lightspeed's own FY23-Q2 goodwill note, cited three times in this ledger and never quoted, says the market-capitalisation crossing was live at 30 September 2022 and produced a test that recognised nothing; C-12 and C-40 are false as written and I concede them outright rather than narrow them. C-02, the record carrying the whole median-above-four-quarters conclusion, converts two annual prevalence statistics into a quarterly duration statistic that neither source reports and that a Q4 indicator with a Q1 charge already refutes — conceded. C-43 is falsified by the page it cites; C-34 dies three times over, decisively on my own retrieval that Corus wrote off $673.0M of Television goodwill in the quarter ended 31 May 2020, which removes it from any first-impairment population; C-17 is load-bearing on no source and about the shape of questions rather than about issuers; C-48 was never checked by anyone. Seven concessions, five of them LOAD-BEARING, including the spine. What I did rebuild, I rebuilt narrower: C-22 survives its refutation only because the Canadian Goodwill Impairment Study stopped at a 2014 edition covering 2007–2013 with no quarter-end or first-impairment construct, and C-14 survives only because I could finally count the population it presupposed — three. Against the Null I win two rivals and only two. Rival 2 falls to evidence, not assertion: Lightspeed disclosed a crossing in a filing that recorded no charge, and Ballard's and TELUS Digital's crossings are establishable from audited equity against exchange data with no issuer narrative involved — none of those lags is zero. Rival 1's exculpatory form falls to ESMA's and IOSCO's own finding that shielding is "among the most significant causes leading to the non-recognition (or non-timely recognition) of goodwill impairment," which makes lawful non-recognition the mechanism of lateness rather than an alternative to it. Rivals 3 and 5 I concede, and Rival 4 I split. But the number that came out of the measurement is the one that matters: one quarter for Lightspeed, at most two for Ballard, two to five for TELUS Digital. The ledger asked whether the median exceeds four quarters, and the best Canadian evidence I could assemble says it does not.
Adjudicator, round 1 — the Adjudicator brief in §8 requires no concession line; its output is the verdict map itself. Verdicts issued: HELD=21 NARROWED=20 CONTESTED=0 UNDETERMINED=1 KILLED=6.
ROUND 2
Inquisitor, round 2
> CONCESSION: MOVED — I expected the issuer-filing records to fail on existence or wording and they did not. C-07, C-08, C-27, C-28, C-31, C-36, C-41 and C-46 survived verbatim checks against their cited documents, and C-11's numbers match its abstract exactly. The ledger is disciplined at the sentence level, and several records hedge in ways that pre-empt the obvious attack. What it is not disciplined about is the boundary of a document: the recurring failure across C-07, C-15, C-21, C-28 and C-35 is citing a filing for a negative or a date while leaving the recognition sitting in the same filing unbooked — Lightspeed's $748.7M, Ballard's $40,277 thousand, Corus's $295.2M of goodwill — which is what makes C-14's three-issuer population collapse against the ledger's own C-07. Three hard kills stand on retrieval: C-19 miscounts twelve as thirteen on a list that is expressly non-exhaustive, C-45 reverses the authors' own SFAS 142 caveat because it cites the preprint abstract rather than the published paper, and C-26 substitutes "non-zero posterior means" for a posterior z-score threshold of 1.96, turning an 82.4 percent finding into a near-vacuous one. C-14 and C-22 are the structural weak points: both are self-sourced, one openly carries no source at all, and both convert a single run's retrieval into a population fact.
Null, round 2
> CONCESSION: MOVED — I expected the breadth to be illusory and it partly is: every record carries a single source, C-15 and C-21 are one sentence entered twice as two LOAD-BEARING records, C-28 and C-46 are one PDF, and C-37 and C-41 are one event. But I could not break the facts. Thirteen records I retrieved against verified against primary documents, several to the digit and the comma — C-27, C-31, C-45 and C-07 verbatim including their negative riders — and C-22's falsifier held under my own attempt to kill it. This is a ledger of accurate claims, and I concede 14 of 20 without reservation. I moved further than that: C-19 aside, the records I contest are mostly not wrong, they are insufficient, and I have said so rather than pretending otherwise. What I am not moved on is the verdict. The thesis needs a lag; the ledger measures one interval, Lightspeed's, and it is one quarter against an imported anchor of three to four years. Its own C-31 attributes the only clean recognition to WACC and forecast revisions and expressly names no market-cap trigger — common cause, not delayed recognition, and nothing here separates them. Its own C-22 says the required dataset does not exist; its own C-14 says the substitute is three issuers, tier E3, no external source, two reclassifiable. C-04 omits the annual-test rule that dissolves the framing, and C-16's contradiction evaporates once a minority float in a consolidated subsidiary stops being read as the fair value of a CGU. On the facts: moved. On the question: undetermined, and the ledger says so itself in three places.
Advocate, round 2
> CONCESSION: MOVED — substantially, and at the places that decide the question rather than at the margins. I concede C-14 outright and recommend it be struck, not narrowed: it has no external source, its enumeration is falsified from inside the ledger by C-07's Corus, its "first-time impairer" test is undefined and outcome-determinative at n=3 given Ballard's impairments in both 2023 and 2024, and its falsifier tests the universe while the live risk is the retrieved population. I concede C-21 as a duplicate of C-15 that should be merged and struck. I concede that "thirteen" at C-19 is simply wrong and that no count is a finding of a list the notice calls illustrative. I concede that C-45 reverses the authors' own SFAS 142 caveat because it cites the preprint abstract, and further that a lag measured in a US amortisation-era regime is not transportable into IAS 36 Canada 2022–2025 without a bridging record the ledger does not hold. I concede that C-26 substitutes a near-vacuous criterion for a posterior z-score threshold, siding with the Inquisitor against the Null's concession. I concede that the entire BCE block cannot support a goodwill observation. Most importantly I concede the three rivals the Null named and I could not discriminate: retrieval-by-outcome at C-14, minority-float-versus-optimistic-DCF at C-16, and common-cause-versus-delay at C-31 — and I have named the specific fact that would settle each rather than arguing around them. What survives is narrower and, I think, sturdier: C-08, C-27, C-28 and C-35 hold as written and the attacks on them are ledger-bookkeeping defects, not claim defects; C-04 rebuilds into a correct statement of both limbs of IAS 36 that concedes its own consequence, that a crossing-to-recognition gap is not evidence of deferral; C-07, C-11, C-22, C-36, C-41 and C-46 rebuild narrower and stay true; and C-15 rebuilds into the one thing the ledger actually measured — Lightspeed's crossing at September 30 2022 with a completed test and no impairment, its US$748.7M recognition at December 31 2022, an interval of one quarter. That interval cuts against the thesis the ledger was assembled to support, and booking it is the honest output. Where the evidence stops: there is no denominator, no agreed definition of crossing, no second measured interval, and no adjudication between common cause and delay. On the facts the ledger is largely sound. On the lag thesis it does not reach, and after this round it should stop claiming to.
Adjudicator, round 2 — no concession required by brief. Verdicts after round 2: HELD=21 NARROWED=19 CONTESTED=0 UNDETERMINED=1 KILLED=7. Verdict-token changes over the 20 re-entered claims: 1 (C-14, NARROWED → KILLED).
ROUND 3
Inquisitor, round 3
> CONCESSION: UNMOVED — the narrowing that let this subset survive round two was applied to claim text, not to source lists, and the two came apart. Records were rewritten until they said only what a careful reader would concede, and then left pointing at whatever source they had entered before the rewrite: C-11 and C-26 and C-45 now state findings their SSRN abstract pages have never contained, C-15 and C-21 state findings their earnings release does not contain, and C-04 states five paragraph numbers its consultancy article never prints. Meanwhile the hedges themselves became the attack surface — C-46, C-37, C-41 and C-36 each declare a limit on the evidence that is in fact a limit on the retrieval, and each is dissolved by a document filed by the same issuer, often on the same day, most decisively BCE's Q3 2024 statements, which put $1,132 million of goodwill inside a charge three records agree supports no goodwill observation. C-07 fails on arithmetic that survived a round of attack unchecked: three components summing to the year total, presented as the composition of the quarter. What a narrowing left behind here is a ledger that reasons well about what its sources would show if anyone opened them.
Null, round 3
> CONCESSION: MOVED — I came to argue correlation and left unable to. shared-sources.txt is a host-level artifact, not a common origin: three SSRN abstract_ids are three papers, not one, and no dataset, author or account stands behind this subset. The Lightspeed duplication was disclosed and voided by the ledger before I reached it, and the Ballard PDF carries two independent facts I verified separately. Fifteen of nineteen records survived direct retrieval against their own primary sources, four of them — C-19's twelve-item list, C-27's test disclosure, C-37's article, C-45's abstract — word for word, and eight records volunteer limitations I had prepared to impose on them. On thinness I am also moved: single-sourcing is not a defect for a claim of the form "filing F states S," and I withdraw it as a general argument, retaining it only where a claim exceeds its document. I hold four. C-07 I hold hardest, and on new evidence rather than on doubt: the components are misallocated by arithmetic the record cannot survive, and Corus recognised $350M of Television goodwill in the very quarter of the first crossing the record names — a fact that does not weaken the lag conclusion so much as invert it. C-15 and C-22 I hold on structure, not on error: an interval cannot measure responsiveness when a standing annual test date fixes its right endpoint, and a negative cannot bear load when its search space excluded SEDAR+, Compustat/WRDS, Capital IQ and Audit Analytics. C-11 I hold because 65% and 98% are not two views of one denominator and the ledger has no number that decomposes the gap. Undetermined, on these four, is the finding — not the fallback.
Advocate, round 3
> CONCESSION: MOVED — I entered expecting to narrow four or five records and hold the rest, and the retrieval moved me on far more than that. I concede C-22 outright: a load-bearing negative whose search space excluded SEDAR+, Compustat/WRDS, Capital IQ and Audit Analytics cannot be rebuilt into anything that bears load, and its one checkable positive sub-clause is wrong on three counts. I concede the operative half of all three of the Null's remaining contests and defeat none of them: C-07's component allocation is arithmetically impossible and the $350.0M fiscal-2022 charge inverts its interval; C-11's "restricted to the 161" framing collapses once Chen and Zimmermann's own construction rule is read, and the only discriminating evidence I could retrieve — their §5.2 finding that liquidity adjustments cut mean returns by about a third — leans toward the Null's construction rival, not away from it; and on C-15 I found the September 30 2022 sensitivity disclosure the Null said was absent, inside the ledger's own C-27 document, and it is a moderate cushion, which on the Null's own stated test is evidence for its rival — my only correction to the Null there is that the number existed, and it did not help me. I concede the Inquisitor's BCE attack across C-36, C-37 and C-41 and withdraw their shared conclusion, and the C-36/C-41 merger with it. I concede C-28's omission, C-35's over-reading, C-46's dissolved caveat, C-31's mis-scoped negative, C-26's imported denominator, C-45's body-of-paper clauses, C-21's misattribution, and half of C-04's provenance. I hold three records unamended — C-16, C-19 and C-27 — each because the attack against it establishes a gap in the ledger's coverage or a defect in its metadata rather than an error in its text, and I hold one inference against the Inquisitor: IAS 36 ¶15 relieves an estimate rather than deferring a test, so it does not satisfy C-04's falsifier and in fact strengthens C-04's conclusion, which is why I wrote it into the claim. What survives is a subset that is reliable where it reports what a document says and unreliable wherever it built a quantity across documents — and where the quantities are rebuilt correctly, they point the other way.
Adjudicator, round 3 — no concession required by brief. Verdicts after round 3: HELD=21 NARROWED=19 CONTESTED=0 UNDETERMINED=1 KILLED=7. Verdict-token changes over the 19 re-entered claims: 0.
Reader — the §9 brief forbids caveats and additions; no concession line is emitted by design.
CONTROL ARM
CONTROL_ARM: NO
The input block sets CONTROL_ARM: NO with the stated reason "unscoreable without a key; skip". No
control-arm agent was spawned, no control-arm Reader block was produced, and retrieval_calls_control
is n/a. Measurement 1 of §15.3 — Crucible versus control arm on the same claim, the measurement the
architecture exists to settle — is therefore not available from this run.
HALT RECORD
| Round | Re-entered subset | Subset size | new_landed_attacks | verdict_changes (subset only) |
|---|---|---|---|---|
| 1 | initial ledger, nothing re-entered | 48 | 40 | n/a — initial assignment, no prior map |
| 2 | round-1 NARROWED + CONTESTED | 20 | 44 | 1 |
| 3 | round-2 NARROWED + CONTESTED | 19 | 25 | 0 |
halt_reason: ZERO_VERDICT_CHANGES. Round 3 produced 25 accepted attacks and moved no verdict
token, which is the §10 halt condition exactly: attacks that land without changing a verdict are cost
without effect. Round 3 was also MAX_ROUNDS, so the hard cap and the halt condition coincide; the
halt condition is recorded as the operative reason because it was satisfied on its own terms.
Round 2's single verdict change was C-14 (SUPPORTING, E3), NARROWED → KILLED. No load-bearing verdict
token moved in either round 2 or round 3, though claim text moved in 17 records in round 2 and in 16
in round 3 — which is why verdict_changes is computed over the re-entered subset and on the verdict
token, not on the map as a whole.
Re-entry followed §10 strictly: HELD, KILLED and UNDETERMINED items did not re-enter in any round,
and each round's Inquisitor and Null were fresh agents in fresh contexts, given the subset in stripped
format with its source lists and a shared-sources.txt mechanically restricted to the subset.
Tripwire
NOT_FIRED in all three rounds. 25 of 30 load-bearing claims HELD or NARROWED; five
load-bearing claims were KILLED (C-02, C-12, C-17, C-34, C-40), so the condition "every load-bearing
claim is verdicted HELD or NARROWED" was never met. No tripwire pass was run, no verdicts are marked
(tripwire), and tripwire_verdicts_moved is 0. The count was identical in all three rounds.
REJECTED ATTACKS
attacks_scored_zero: 57 = the line count of this section's source file
(05-verdict/rejected-attacks-ALL.md, the three per-round files concatenated in order). Verbatim:
C-01 │ PROVENANCE │ rejected: the claim asserts what the Reuters wire reported, which the attack's own evidence confirms, and the Advocate retrieved IOSCO FR13/23 carrying the phrase in IOSCO's own words — citation hygiene, not a defeater.
C-02 │ PROVENANCE │ rejected: non-independence of ESMA/2013/2 and the André/Filip/Paugam panel is a weighting instruction; it does not falsify the claim's content, which fails on the INFERENCE ground instead.
C-02 │ SELECTION │ rejected: names Ramanna & Watts and Glaum/Landsman/Wyrwa as absent but produces no finding that contradicts the claim; unpersuasive as an independent defeater.
C-03 │ PROVENANCE │ rejected: duplication with C-07 off one MD&A clause is a bookkeeping defect, and the Null verified the disclosure independently against Corus's FY2022 year-end release.
C-05 │ PROVENANCE │ rejected: duplication with C-20 and the tier split are ledger hygiene; the indicator sentence was verified verbatim by both attackers and the Advocate.
C-06 │ PROVENANCE │ rejected: eighth-order duplication is a weighting instruction; the US$748.7M charge is triangulated by release, interim statements, audited statements and the ledger's own equity roll-forward.
C-08 │ PROVENANCE │ rejected: trade-press provenance and shared Kroll lineage with C-22 go to weight, not to whether CFO Dive reported the quoted forecast.
C-09 │ PROVENANCE │ rejected: consecutive-paragraph non-independence with C-02 and C-29 is a weighting instruction; the 235-entity sample and 43% figure were verified exact against the primary.
C-10 │ PROVENANCE │ rejected: splitting one abstract into a result record and a method record is a weighting defect; the 20-to-25% sentence was verified exact.
C-11 │ PROVENANCE │ rejected: that Jensen/Kelly/Pedersen was written to answer Hou/Xue/Zhang bears on how the pair is used, not on whether C-11 reports its own paper accurately.
C-12 │ INFERENCE │ rejected: its written-out weaker claim merely restates C-32, and the record is falsified outright on the PROVENANCE ground, so no narrowing survives to adopt.
C-13 │ PROVENANCE │ rejected: the mutual inconsistency across the Grant Thornton records is C-43's misread, not C-13's; C-13 is the faithful reading and was verified verbatim by all three parties.
C-14 │ PROVENANCE │ rejected: "no external source" restates the record's declared E3 model-prior tier rather than defeating it, and the Advocate supplied the population count the record lacked.
C-14 │ NULL │ rejected: the Advocate established the antecedent by count — three determinable crossing-to-recognition pairs, two of them reclassifiable — breaking the C-14/C-22 self-block, so the record narrows rather than fails.
C-18 │ PROVENANCE │ rejected: three line items from one filing is a weighting instruction; the Advocate closed the share-class caveat and showed the figure does support the market-capitalisation computation.
C-20 │ PROVENANCE │ rejected: duplication of C-05 is hygiene, and the falsifier's appeal to an uncited IAS 36 paragraph 12 is a citation gap that does not touch the claim's truth.
C-21 │ PROVENANCE │ rejected: near-duplication of C-15 is a weighting defect; the narrowing that lands on this record comes from the accepted C-15 INFERENCE attack, not from this one.
C-22 │ NULL │ rejected: the narrowed record drops the universal negative and scopes to 2022-2025 quarter-end granularity; the Null asserts no counter-dataset exists and the one produced stops at a 2014 edition covering 2007-2013.
C-23 │ PROVENANCE │ rejected: third rendering of one charge is a weighting instruction; the amount and period are conceded by the Null and verified by the Advocate.
C-24 │ PROVENANCE │ rejected: sharing an abstract page with C-10 is a weighting defect; all three economic-impairment proxies were verified exact.
C-25 │ PROVENANCE │ rejected: the single-document objection fails on the record — the Null corroborated the $350.0M Television charge outside the ledger's Corus PDF against the FY2022 Q4 release.
C-29 │ PROVENANCE │ rejected: non-independence with C-09 and C-02 is a weighting instruction; the 47% figure was retrieved and matched exactly against ESMA/2013/2.
C-30 │ PROVENANCE │ rejected: splitting one abstract with C-45 is a weighting defect; the ten-year-tail sentence was verified verbatim by the Null and the Advocate.
C-31 │ PROVENANCE │ rejected: republication rather than the primary filing is a tier-consistency complaint; amount, period and attribution were confirmed against the release itself.
C-32 │ PROVENANCE │ rejected: restating one event under a different URL is weighting and the balance-sheet date field is metadata; the Q3 FY23 note's language was verified by the Null.
C-33 │ PROVENANCE │ rejected: the contradiction the attack identifies belongs to C-43; C-33 is the verbatim-correct reading and is the record that falsifies C-43.
C-39 │ PROVENANCE │ rejected: the attack targets the ledger's use of the equity figures rather than the figures, which were verified and whose difference corroborates the charge's magnitude.
C-42 │ PROVENANCE │ rejected: third statement of one charge from the audited-statement side, and a shared PDF with C-35 — weighting and hygiene, not a defeater.
C-44 │ PROVENANCE │ rejected: three records off one Ballard PDF is a weighting instruction; the $40,277k figure and the September 30 2024 measurement date were verified verbatim.
C-45 │ SELECTION │ rejected: names Li & Sloan and the IASB Exposure Draft as missing but produces no finding contradicting what Hayn & Hughes report; the transport defect is already carried by the accepted INFERENCE and PROVENANCE attacks.
C-47 │ PROVENANCE │ rejected: restatement of C-25 with a changed period descriptor is a duplication defect; the charge was corroborated outside the ledger's PDF against the FY2022 Q4 release.
C-04 │ PROVENANCE │ rejected: resting a statement of a standard's text on an E2 advisory summary is citation hygiene; the Inquisitor's own retrieval of ¶9, ¶10, ¶12(d), ¶96 and ¶99 confirms the content, so it goes to weight, not truth.
C-08 │ PROVENANCE │ rejected: shared Kroll lineage with C-22 is source concentration and a weighting instruction; Kroll being the origin of a Kroll managing director's forecast is not circularity.
C-08 │ SELECTION │ rejected: the outcome the attack says is unbooked confirms rather than contradicts the forecast — Kroll's own H1 2022 trends publication reports impairments up — so it produces no finding against the claim.
C-11 │ PROVENANCE │ rejected: the SSRN landing page is non-primary and revisable, but both attackers matched every element of the claim verbatim against the cited 2018-10-31 version; hygiene, not a defeater.
C-11 │ NULL │ rejected: the Advocate produced the adjudicating record the attack says the ledger lacks — Chen & Zimmermann partition HXZ's library so both results are simultaneously true — and round 1 already narrowed C-11 out of any general-reliability use.
C-14 │ SELECTION │ rejected: the Duff & Phelps/Kroll Canadian series stops at a 2014 edition and yields no 2022–2025 population, so it contradicts nothing in the claim; the enumeration falls on the INFERENCE ground instead.
C-15 │ PROVENANCE │ rejected: the C-15/C-21 duplication defect attaches to the second entry of the pair, not the first; C-15 is the record that carries the proposition and the narrowing lands on C-21.
C-16 │ SELECTION │ rejected: names TELUS Digital's FY2024 Form 20-F as absent and confirms it exists, but extracts no finding from it that contradicts the claim.
C-26 │ NULL │ rejected: same ground as C-11 — Chen & Zimmermann discriminates by partitioning the two sets, and round 1 already netted the C-11/C-26 replication calibration to zero.
C-27 │ PROVENANCE │ rejected: a source date field carrying the balance-sheet date rather than the 2 November 2022 board approval date is metadata, on the same ground the identical attack on C-32 was rejected in round 1.
C-27 │ SELECTION │ rejected: ASSERTED, and an unretrieved Q1 FY2023 filing produces no finding against a claim both attackers verified verbatim, clause by clause, against the retrieved Q2 FY2023 statements.
C-28 │ PROVENANCE │ rejected: sharing the Ballard PDF with C-46 is a weighting instruction, on the same ground the three-records-off-one-Ballard-PDF attack on C-44 was rejected in round 1.
C-28 │ SELECTION │ rejected: the premise fails against the full ledger — Ballard's 2024 amount and 30 September 2024 test date are booked at C-44, which is HELD; the attack read only the 20-claim round-2 subset.
C-31 │ PROVENANCE │ rejected: Nasdaq's syndication of a Business Wire release is a tier-consistency complaint already rejected on this record in round 1; amount, period and attribution were verified against the release itself.
C-31 │ SELECTION │ rejected: confirms that TELUS Digital's Q2 2025 interim statements exist but extracts no finding from them; the substance is carried by the accepted INFERENCE attack that withdrew the silence clause.
C-35 │ SELECTION │ rejected: a claim is not falsified by a consistent fact it omits, and the omitted affirmative half — the $748,712 thousand FY2023 loss and the trigger narrative — is now booked at C-15 and C-21.
C-36 │ SELECTION │ rejected: names BCE's 2024 annual audited statements as absent without extracting the goodwill-versus-indefinite-life split from them; the admissibility consequence is already carried by the accepted INFERENCE and NULL attacks.
C-41 │ SELECTION │ rejected: that the falsifier points at uncited Q3 2024 interim statements was already booked in round 1's narrowing of this record; the attack adds no new finding.
C-45 │ SELECTION │ rejected: ASSERTED, names Ramanna & Watts and Li & Sloan as absent without producing a finding that contradicts what Hayn & Hughes report — the same ground on which the round-1 SELECTION attack on C-45 was rejected.
C-46 │ PROVENANCE │ rejected: identity of PDF and note with C-28 is a weighting instruction; the accepted narrowing on this record comes from the comparative-column provenance attack, not this one.
C-08 │ PROVENANCE │ rejected: one Kroll origin behind two URLs is source concentration and a weighting instruction, rejected on this record in round 2 and unchanged by re-filing; C-08 and C-22 are different documents about different objects, and the Advocate's acceptance of the weighting point is not an error of truth.
C-16 │ SELECTION │ rejected: the strongest SELECTION attack of the run and still a coverage gap, not a defect — TELUS Digital's US$1,926M goodwill and US$1.08bn market capitalisation come from another issuer's release and a market-data vendor, neither in the ledger, and contradict no word of a claim reporting the parent's audited conclusion about a parent-level CGU.
C-19 │ SELECTION │ rejected: CSA Staff Notice 51-365 is real, pre-cutoff and more substantive, but its findings contradict nothing in this record and if anything run the ledger's way; naming a stronger absent document is a completeness point about what the ledger cites.
C-27 │ PROVENANCE │ rejected: the source date carrying the 2022-09-30 balance-sheet date rather than the 2 November 2022 authorisation date is metadata, rejected on identical grounds at C-32 in round 1 and at this record in round 2; the attack's independence limb is accepted and booked in C-15's narrowed text, not here.
C-28 │ SELECTION │ rejected: the premise fails against the full ledger for the second consecutive round — Ballard's US$40,277 thousand and its 30 September 2024 test date are booked at C-44 (HELD), and the attack is again reading only the re-entered subset rather than the 48-claim map.
C-28 │ PROVENANCE │ rejected: sharing the Ballard PDF and note with C-46 is the weighting instruction rejected at C-44 in round 1 and at this record in round 2; the Null and the Advocate both confirm the two records carry genuinely independent facts.KEYMAP
01-ledger/keymap.json. Convener-only during the run; held outside the working tree from the moment
strip.py wrote it until this section was assembled, so it was absent from the filesystem during
every Round 1, 2 and 3 phase and during all three Adjudicator phases. Appended last, for scoring only.
{
"C-01": {
"claim": "Reuters reported IOSCO naming 'too little, too late' goodwill impairment a priority issue for securities regulators.",
"territories": [
"2"
],
"load": "SUPPORTING",
"tier": "E1",
"sources": [
{
"source": "Reuters wire report republished by Investing.com",
"source_id": "https://investing.com/news/stock-market-news/global-watchdog-seeks-action-on-too-little-too-late-corporate-goodwill-writedown-3111225",
"source_id_norm": "investing.com/news/stock-market-news/global-watchdog-seeks-action-on-too-little-too-late-corporate-goodwill-writedown-3111225",
"source_date": "2023-06-22",
"tier": "E1"
}
]
},
"C-02": {
"claim": "Majority non-recognition within the indicator year, as reported by ESMA plus Andr\u00e9 Filip Paugam, places the median indicator-to-recognition lag above four quarters.",
"territories": [
"4"
],
"load": "LOAD-BEARING",
"tier": "E2",
"sources": [
{
"source": "Inference from ESMA/2013/2 and Andr\u00e9, Filip & Paugam abstract figures retrieved this run",
"source_id": "https://www.esma.europa.eu/sites/default/files/library/2015/11/2013-02.pdf",
"source_id_norm": "www.esma.europa.eu/sites/default/files/library/2015/11/2013-02.pdf",
"source_date": "2013-01-07",
"tier": "E2"
}
]
},
"C-03": {
"claim": "Corus disclosed that its carrying value was greater than its market enterprise value at August 31, 2022.",
"territories": [
"1"
],
"load": "LOAD-BEARING",
"tier": "E0",
"sources": [
{
"source": "Corus Entertainment Inc. annual MD&A, year ended August 31, 2023, impairment discussion",
"source_id": "https://assets.corusent.com/wp-content/uploads/2023/10/Corus-Entertainment-Annual-MDA-and-Financial-Statements-Oct-30-2023-vf.pdf",
"source_id_norm": "assets.corusent.com/wp-content/uploads/2023/10/corus-entertainment-annual-mda-and-financial-statements-oct-30-2023-vf.pdf",
"source_date": "2023-10-30",
"tier": "E0"
}
]
},
"C-04": {
"claim": "Mandatory quarter-end indicator assessment under IAS 36 makes a lag beyond one quarter a departure from required timing.",
"territories": [
"3"
],
"load": "LOAD-BEARING",
"tier": "E2",
"sources": [
{
"source": "Inference from Grant Thornton IAS 36 guidance",
"source_id": "https://www.grantthornton.global/en/insights/articles/IFRS-ias-36/ifrs-ias-36-If-and-when-to-undertake-an-impairment-review/",
"source_id_norm": "www.grantthornton.global/en/insights/articles/ifrs-ias-36/ifrs-ias-36-if-and-when-to-undertake-an-impairment-review",
"source_date": "2021-08-03",
"tier": "E2"
}
]
},
"C-05": {
"claim": "Grant Thornton lists carrying amount of net assets exceeding market capitalisation as an external indicator of potential impairment under IAS 36.",
"territories": [
"3"
],
"load": "SUPPORTING",
"tier": "E0",
"sources": [
{
"source": "Grant Thornton International, 'IAS 36 - If and when to undertake an impairment review'",
"source_id": "https://www.grantthornton.global/en/insights/articles/IFRS-ias-36/ifrs-ias-36-If-and-when-to-undertake-an-impairment-review/",
"source_id_norm": "www.grantthornton.global/en/insights/articles/ifrs-ias-36/ifrs-ias-36-if-and-when-to-undertake-an-impairment-review",
"source_date": "2021-08-03",
"tier": "E0"
}
]
},
"C-06": {
"claim": "Lightspeed Commerce recognized a US$748.7 million goodwill impairment charge in the three months ended December 31 2022.",
"territories": [
"3"
],
"load": "LOAD-BEARING",
"tier": "E0",
"sources": [
{
"source": "Lightspeed Commerce Inc. third quarter fiscal 2023 earnings release filed with SEC",
"source_id": "https://www.sec.gov/Archives/edgar/data/1823306/000182330623000008/earningsreleasefy23q3.htm",
"source_id_norm": "www.sec.gov/archives/edgar/data/1823306/000182330623000008/earningsreleasefy23q3.htm",
"source_date": "2023-02-02",
"tier": "E0"
}
]
},
"C-07": {
"claim": "Corus disclosed carrying value greater than market enterprise value at August 31, 2022, May 31, 2023, August 31, 2023.",
"territories": [
"2"
],
"load": "LOAD-BEARING",
"tier": "E0",
"sources": [
{
"source": "Corus Entertainment annual MD&A and audited consolidated financial statements, fiscal 2023",
"source_id": "https://assets.corusent.com/wp-content/uploads/2023/10/Corus-Entertainment-Annual-MDA-and-Financial-Statements-Oct-30-2023-vf.pdf",
"source_id_norm": "assets.corusent.com/wp-content/uploads/2023/10/corus-entertainment-annual-mda-and-financial-statements-oct-30-2023-vf.pdf",
"source_date": "2023-10-30",
"tier": "E0"
}
]
},
"C-08": {
"claim": "CFO Dive reported Kroll managing director Carla Nunes expected 2022 goodwill impairments to rise relative to 2021.",
"territories": [
"3"
],
"load": "SUPPORTING",
"tier": "E1",
"sources": [
{
"source": "Maura Webber Sadovi, CFO Dive",
"source_id": "https://www.cfodive.com/news/goodwill-impairment-expected-rise-reverse-course-2022-kroll/621614/",
"source_id_norm": "www.cfodive.com/news/goodwill-impairment-expected-rise-reverse-course-2022-kroll/621614",
"source_date": "2022-04-06",
"tier": "E1"
}
]
},
"C-09": {
"claim": "ESMA found 43 percent of 235 sampled European issuers carried market capitalisation below equity book value at 31 December 2011.",
"territories": [
"4"
],
"load": "SUPPORTING",
"tier": "E0",
"sources": [
{
"source": "ESMA/2013/2, Review of practices related to goodwill impairment",
"source_id": "https://www.esma.europa.eu/sites/default/files/library/2015/11/2013-02.pdf",
"source_id_norm": "www.esma.europa.eu/sites/default/files/library/2015/11/2013-02.pdf",
"source_date": "2013-01-07",
"tier": "E0"
}
]
},
"C-10": {
"claim": "Andr\u00e9 Filip Paugam find only 20 to 25 percent of firms exhibiting economic impairment indications recognised an accounting impairment in the same period.",
"territories": [
"4"
],
"load": "LOAD-BEARING",
"tier": "E0",
"sources": [
{
"source": "Andr\u00e9, Filip & Paugam, Examining the Patterns of Goodwill Impairments in Europe and the US, SSRN abstract",
"source_id": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2867340",
"source_id_norm": "papers.ssrn.com/sol3/papers.cfm",
"source_date": "2017-03-20",
"tier": "E0"
}
]
},
"C-11": {
"claim": "Hou Xue Zhang report 65 percent of 452 catalogued anomalies fail to clear an absolute t-value of 1.96 under their replication protocol.",
"territories": [
"4"
],
"load": "SUPPORTING",
"tier": "E0",
"sources": [
{
"source": "Hou, Xue & Zhang, Replicating Anomalies, SSRN abstract page",
"source_id": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3275496",
"source_id_norm": "papers.ssrn.com/sol3/papers.cfm",
"source_date": "2018-10-31",
"tier": "E0"
}
]
},
"C-12": {
"claim": "Lightspeed's disclosed lag from market-capitalization trigger to goodwill impairment recognition is zero quarters.",
"territories": [
"3"
],
"load": "LOAD-BEARING",
"tier": "E2",
"sources": [
{
"source": "Inference from Lightspeed Commerce Q3 FY2023 earnings release",
"source_id": "https://www.sec.gov/Archives/edgar/data/1823306/000182330623000008/earningsreleasefy23q3.htm",
"source_id_norm": "www.sec.gov/archives/edgar/data/1823306/000182330623000008/earningsreleasefy23q3.htm",
"source_date": "2023-02-02",
"tier": "E2"
}
]
},
"C-13": {
"claim": "Grant Thornton's IFRS team states identifying an impairment indicator does not automatically trigger recognition of an impairment loss.",
"territories": [
"3"
],
"load": "LOAD-BEARING",
"tier": "E0",
"sources": [
{
"source": "Grant Thornton International, 'IAS 36 - If and when to undertake an impairment review'",
"source_id": "https://www.grantthornton.global/en/insights/articles/IFRS-ias-36/ifrs-ias-36-If-and-when-to-undertake-an-impairment-review/",
"source_id_norm": "www.grantthornton.global/en/insights/articles/ifrs-ias-36/ifrs-ias-36-if-and-when-to-undertake-an-impairment-review",
"source_date": "2021-08-03",
"tier": "E0"
}
]
},
"C-14": {
"claim": "Medians computed over bounded single-index populations across four fiscal years rest on few observations, so single-issuer reclassification can move the reported statistic.",
"territories": [
"4"
],
"load": "SUPPORTING",
"tier": "E3",
"sources": [
{
"source": "Model prior on small-sample median instability",
"source_id": "model prior, no external source",
"source_id_norm": "model prior, no external source",
"source_date": "2026-08-22",
"tier": "E3"
}
]
},
"C-15": {
"claim": "Lightspeed stated its net assets exceeded its market capitalization as at December 31, 2022, which was an impairment trigger.",
"territories": [
"2"
],
"load": "LOAD-BEARING",
"tier": "E0",
"sources": [
{
"source": "Lightspeed Commerce Q3 fiscal 2023 earnings release filed on SEC EDGAR",
"source_id": "https://www.sec.gov/Archives/edgar/data/1823306/000182330623000008/earningsreleasefy23q3.htm",
"source_id_norm": "www.sec.gov/archives/edgar/data/1823306/000182330623000008/earningsreleasefy23q3.htm",
"source_date": "2023-02-02",
"tier": "E0"
}
]
},
"C-16": {
"claim": "TELUS Corporation recognized no goodwill impairment for the TELUS digital experience cash-generating unit in 2024.",
"territories": [
"1"
],
"load": "SUPPORTING",
"tier": "E0",
"sources": [
{
"source": "TELUS Corporation audited consolidated financial statements, year ended December 31, 2024, filed as Form 40-F exhibit 99.4",
"source_id": "https://www.sec.gov/Archives/edgar/data/868675/000110465925012533/tu-20241231xex99d4.htm",
"source_id_norm": "www.sec.gov/archives/edgar/data/868675/000110465925012533/tu-20241231xex99d4.htm",
"source_date": "2025-02-13",
"tier": "E0"
}
]
},
"C-17": {
"claim": "Threshold claims stating a median exceeds a bound far beneath the reference class central estimate resolve true more often than thresholds set near that estimate.",
"territories": [
"4"
],
"load": "LOAD-BEARING",
"tier": "E3",
"sources": [
{
"source": "Model prior on threshold placement relative to reference class central tendency",
"source_id": "model prior, no external source",
"source_id_norm": "model prior, no external source",
"source_date": "2026-08-22",
"tier": "E3"
}
]
},
"C-18": {
"claim": "Lightspeed had 150,315,764 subordinate voting shares issued outstanding at September 30, 2022.",
"territories": [
"1"
],
"load": "LOAD-BEARING",
"tier": "E0",
"sources": [
{
"source": "Lightspeed Commerce Inc. condensed interim consolidated financial statements, share capital note, September 30, 2022",
"source_id": "https://s28.q4cdn.com/517092977/files/doc_financials/2023/q2/Financial-Statements-FY23-Q2-Final.pdf",
"source_id_norm": "s28.q4cdn.com/517092977/files/doc_financials/2023/q2/financial-statements-fy23-q2-final.pdf",
"source_date": "2022-09-30",
"tier": "E0"
}
]
},
"C-19": {
"claim": "CSA Staff Notice 51-364 named impairment of non-financial assets as an area affected by the current economic environment.",
"territories": [
"3"
],
"load": "SUPPORTING",
"tier": "E0",
"sources": [
{
"source": "Canadian Securities Administrators Staff Notice 51-364",
"source_id": "https://www.osc.ca/sites/default/files/2022-11/csa_20221103_51-364_continuous-disclosure-review.pdf",
"source_id_norm": "www.osc.ca/sites/default/files/2022-11/csa_20221103_51-364_continuous-disclosure-review.pdf",
"source_date": "2022-11-03",
"tier": "E0"
}
]
},
"C-20": {
"claim": "Grant Thornton lists carrying amount of net assets exceeding market capitalisation among the IAS 36 external indicators requiring an impairment review.",
"territories": [
"2"
],
"load": "LOAD-BEARING",
"tier": "E1",
"sources": [
{
"source": "Grant Thornton International, IFRS IAS 36 guidance on if and when to undertake an impairment review",
"source_id": "https://www.grantthornton.global/en/insights/articles/IFRS-ias-36/ifrs-ias-36-If-and-when-to-undertake-an-impairment-review/",
"source_id_norm": "www.grantthornton.global/en/insights/articles/ifrs-ias-36/ifrs-ias-36-if-and-when-to-undertake-an-impairment-review",
"source_date": "2021-08-03",
"tier": "E1"
}
]
},
"C-21": {
"claim": "Lightspeed Commerce stated its net assets exceeded its market capitalization at December 31 2022, which was an impairment trigger.",
"territories": [
"3"
],
"load": "LOAD-BEARING",
"tier": "E0",
"sources": [
{
"source": "Lightspeed Commerce Inc. third quarter fiscal 2023 earnings release filed with SEC",
"source_id": "https://www.sec.gov/Archives/edgar/data/1823306/000182330623000008/earningsreleasefy23q3.htm",
"source_id_norm": "www.sec.gov/archives/edgar/data/1823306/000182330623000008/earningsreleasefy23q3.htm",
"source_date": "2023-02-02",
"tier": "E0"
}
]
},
"C-22": {
"claim": "No published dataset enumerates S&P/TSX Composite non-financial first goodwill impairments paired with net-assets-above-market-capitalization crossing quarter-ends.",
"territories": [
"2"
],
"load": "LOAD-BEARING",
"tier": "E2",
"sources": [
{
"source": "Inference from retrieval across Kroll goodwill impairment studies, IOSCO, Reuters, issuer filings in this run",
"source_id": "https://www.kroll.com/en/insights/goodwill-impairment",
"source_id_norm": "www.kroll.com/en/insights/goodwill-impairment",
"source_date": "2026-08-22",
"tier": "E2"
}
]
},
"C-23": {
"claim": "Lightspeed Commerce recognized a non-cash goodwill impairment charge of US$748.7 million for the quarter ended December 31, 2022.",
"territories": [
"2"
],
"load": "LOAD-BEARING",
"tier": "E0",
"sources": [
{
"source": "Lightspeed Commerce Q3 fiscal 2023 earnings release filed on SEC EDGAR",
"source_id": "https://www.sec.gov/Archives/edgar/data/1823306/000182330623000008/earningsreleasefy23q3.htm",
"source_id_norm": "www.sec.gov/archives/edgar/data/1823306/000182330623000008/earningsreleasefy23q3.htm",
"source_date": "2023-02-02",
"tier": "E0"
}
]
},
"C-24": {
"claim": "Andr\u00e9 Filip Paugam measure economic impairment indications using market-to-book below one, equity market value minus book value less than goodwill, negative EBITDA.",
"territories": [
"4"
],
"load": "SUPPORTING",
"tier": "E0",
"sources": [
{
"source": "Andr\u00e9, Filip & Paugam, Examining the Patterns of Goodwill Impairments in Europe and the US, SSRN abstract",
"source_id": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2867340",
"source_id_norm": "papers.ssrn.com/sol3/papers.cfm",
"source_date": "2017-03-20",
"tier": "E0"
}
]
},
"C-25": {
"claim": "Corus recorded a non-cash goodwill impairment charge of $350.0 million in the Television cash generating unit in fiscal 2022.",
"territories": [
"2"
],
"load": "LOAD-BEARING",
"tier": "E0",
"sources": [
{
"source": "Corus Entertainment annual MD&A and audited consolidated financial statements, fiscal 2023",
"source_id": "https://assets.corusent.com/wp-content/uploads/2023/10/Corus-Entertainment-Annual-MDA-and-Financial-Statements-Oct-30-2023-vf.pdf",
"source_id_norm": "assets.corusent.com/wp-content/uploads/2023/10/corus-entertainment-annual-mda-and-financial-statements-oct-30-2023-vf.pdf",
"source_date": "2023-10-30",
"tier": "E0"
}
]
},
"C-26": {
"claim": "Jensen Kelly Pedersen conclude the majority of asset pricing factors can be replicated, contradicting broad claims that published quantitative findings collapse under re-measurement.",
"territories": [
"4"
],
"load": "SUPPORTING",
"tier": "E0",
"sources": [
{
"source": "Jensen, Kelly & Pedersen, Is There a Replication Crisis in Finance?, SSRN abstract page",
"source_id": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3774514",
"source_id_norm": "papers.ssrn.com/sol3/papers.cfm",
"source_date": "2021-03-05",
"tier": "E0"
}
]
},
"C-27": {
"claim": "Lightspeed's goodwill impairment test as at September 30, 2022 resulted in no impairment of goodwill.",
"territories": [
"1"
],
"load": "LOAD-BEARING",
"tier": "E0",
"sources": [
{
"source": "Lightspeed Commerce Inc. condensed interim consolidated financial statements, three and six months ended September 30, 2022",
"source_id": "https://s28.q4cdn.com/517092977/files/doc_financials/2023/q2/Financial-Statements-FY23-Q2-Final.pdf",
"source_id_norm": "s28.q4cdn.com/517092977/files/doc_financials/2023/q2/financial-statements-fy23-q2-final.pdf",
"source_date": "2022-09-30",
"tier": "E0"
}
]
},
"C-28": {
"claim": "Ballard's 2024 goodwill impairment note names the decline in the corporation's market capitalization as an impairment indicator.",
"territories": [
"1"
],
"load": "LOAD-BEARING",
"tier": "E0",
"sources": [
{
"source": "Ballard Power Systems Inc. audited consolidated financial statements, year ended December 31, 2024, impairment indicators disclosure",
"source_id": "https://www.ballard.com/wp-content/uploads/2025/03/Q4-24-Financial-Statements-FINAL.pdf",
"source_id_norm": "www.ballard.com/wp-content/uploads/2025/03/q4-24-financial-statements-final.pdf",
"source_date": "2025-03-12",
"tier": "E0"
}
]
},
"C-29": {
"claim": "Among ESMA-sampled issuers whose market capitalisation sat below equity book value, 47 percent recognised goodwill impairment losses in their 2011 financial statements.",
"territories": [
"4"
],
"load": "LOAD-BEARING",
"tier": "E0",
"sources": [
{
"source": "ESMA/2013/2, Review of practices related to goodwill impairment",
"source_id": "https://www.esma.europa.eu/sites/default/files/library/2015/11/2013-02.pdf",
"source_id_norm": "www.esma.europa.eu/sites/default/files/library/2015/11/2013-02.pdf",
"source_date": "2013-01-07",
"tier": "E0"
}
]
},
"C-30": {
"claim": "Hayn Hughes report the goodwill write-off delay extends up to ten years for one third of the companies they examined.",
"territories": [
"4"
],
"load": "SUPPORTING",
"tier": "E0",
"sources": [
{
"source": "Hayn & Hughes, Leading Indicators of Goodwill Impairment, SSRN abstract page",
"source_id": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=850705",
"source_id_norm": "papers.ssrn.com/sol3/papers.cfm",
"source_date": "2005-11-21",
"tier": "E0"
}
]
},
"C-31": {
"claim": "TELUS Digital recognized a US$224 million non-cash goodwill impairment charge for the quarter ended June 30, 2025.",
"territories": [
"2"
],
"load": "SUPPORTING",
"tier": "E1",
"sources": [
{
"source": "TELUS Digital second quarter 2025 results news release carried by Nasdaq",
"source_id": "https://www.nasdaq.com/press-release/telus-digital-reports-second-quarter-2025-results-incremental-improvement-revenue",
"source_id_norm": "www.nasdaq.com/press-release/telus-digital-reports-second-quarter-2025-results-incremental-improvement-revenue",
"source_date": "2025-08-01",
"tier": "E1"
}
]
},
"C-32": {
"claim": "Lightspeed attributed its December 31, 2022 goodwill impairment test to the carrying amount of net assets exceeding market capitalization.",
"territories": [
"1"
],
"load": "LOAD-BEARING",
"tier": "E0",
"sources": [
{
"source": "Lightspeed Commerce Inc. condensed interim consolidated financial statements, three and nine months ended December 31, 2022",
"source_id": "https://s28.q4cdn.com/517092977/files/doc_financials/2023/q3/Financial-Statements-FY23-Q3-Final.pdf",
"source_id_norm": "s28.q4cdn.com/517092977/files/doc_financials/2023/q3/financial-statements-fy23-q3-final.pdf",
"source_date": "2022-12-31",
"tier": "E0"
}
]
},
"C-33": {
"claim": "Grant Thornton's IFRS team states IAS 36 requires assessing impairment indicators at the end of each reporting period.",
"territories": [
"3"
],
"load": "LOAD-BEARING",
"tier": "E0",
"sources": [
{
"source": "Grant Thornton International, 'IAS 36 - If and when to undertake an impairment review'",
"source_id": "https://www.grantthornton.global/en/insights/articles/IFRS-ias-36/ifrs-ias-36-If-and-when-to-undertake-an-impairment-review/",
"source_id_norm": "www.grantthornton.global/en/insights/articles/ifrs-ias-36/ifrs-ias-36-if-and-when-to-undertake-an-impairment-review",
"source_date": "2021-08-03",
"tier": "E0"
}
]
},
"C-34": {
"claim": "Corus's earliest disclosed period-end of carrying value exceeding market enterprise value, August 31, 2022, matches its fiscal 2022 impairment recognition date.",
"territories": [
"2"
],
"load": "LOAD-BEARING",
"tier": "E2",
"sources": [
{
"source": "Inference from the Corus fiscal 2023 annual MD&A record in this ledger",
"source_id": "https://assets.corusent.com/wp-content/uploads/2023/10/Corus-Entertainment-Annual-MDA-and-Financial-Statements-Oct-30-2023-vf.pdf",
"source_id_norm": "assets.corusent.com/wp-content/uploads/2023/10/corus-entertainment-annual-mda-and-financial-statements-oct-30-2023-vf.pdf",
"source_date": "2023-10-30",
"tier": "E2"
}
]
},
"C-35": {
"claim": "Lightspeed Commerce recognized no goodwill impairment loss in the fiscal year ended March 31, 2022.",
"territories": [
"1"
],
"load": "LOAD-BEARING",
"tier": "E0",
"sources": [
{
"source": "Lightspeed Commerce Inc. audited consolidated financial statements, fiscal year ended March 31, 2023, comparative period",
"source_id": "https://s28.q4cdn.com/517092977/files/doc_financials/2023/q4/Financial-Statements-FY23-Q4-Final.pdf",
"source_id_norm": "s28.q4cdn.com/517092977/files/doc_financials/2023/q4/financial-statements-fy23-q4-final.pdf",
"source_date": "2023-05-18",
"tier": "E0"
}
]
},
"C-36": {
"claim": "BCE reported non-cash asset impairment charges totalling $2,190 million for 2024, mainly related to Bell Media's TV properties.",
"territories": [
"3"
],
"load": "SUPPORTING",
"tier": "E0",
"sources": [
{
"source": "BCE Inc. 2024 fourth quarter and full-year results news release",
"source_id": "https://www.prnewswire.com/news-releases/bce-reports-2024-q4-and-full-year-results-announces-2025-financial-targets-302369604.html",
"source_id_norm": "www.prnewswire.com/news-releases/bce-reports-2024-q4-and-full-year-results-announces-2025-financial-targets-302369604.html",
"source_date": "2025-02-06",
"tier": "E0"
}
]
},
"C-37": {
"claim": "The Canadian Press reported BCE recorded $2.11 billion of asset impairment charges in Q3 2024, mainly related to Bell Media properties.",
"territories": [
"3"
],
"load": "SUPPORTING",
"tier": "E1",
"sources": [
{
"source": "Sammy Hudes, The Canadian Press, published by BNN Bloomberg",
"source_id": "https://www.bnnbloomberg.ca/business/company-news/2024/11/07/bce-reports-q3-loss-on-asset-impairment-charge-cuts-revenue-guidance/",
"source_id_norm": "www.bnnbloomberg.ca/business/company-news/2024/11/07/bce-reports-q3-loss-on-asset-impairment-charge-cuts-revenue-guidance",
"source_date": "2024-11-07",
"tier": "E1"
}
]
},
"C-38": {
"claim": "Lightspeed reported total shareholders' equity of US$3,304,419 thousand at September 30, 2022.",
"territories": [
"1"
],
"load": "LOAD-BEARING",
"tier": "E0",
"sources": [
{
"source": "Lightspeed Commerce Inc. condensed interim consolidated statement of financial position at September 30, 2022",
"source_id": "https://s28.q4cdn.com/517092977/files/doc_financials/2023/q2/Financial-Statements-FY23-Q2-Final.pdf",
"source_id_norm": "s28.q4cdn.com/517092977/files/doc_financials/2023/q2/financial-statements-fy23-q2-final.pdf",
"source_date": "2022-09-30",
"tier": "E0"
}
]
},
"C-39": {
"claim": "Lightspeed shareholders' equity was US$3,399.3 million at March 31, 2022, falling to US$2,539.5 million at December 31, 2022.",
"territories": [
"2"
],
"load": "SUPPORTING",
"tier": "E0",
"sources": [
{
"source": "Lightspeed Commerce Q3 fiscal 2023 earnings release balance sheet, SEC EDGAR",
"source_id": "https://www.sec.gov/Archives/edgar/data/1823306/000182330623000008/earningsreleasefy23q3.htm",
"source_id_norm": "www.sec.gov/archives/edgar/data/1823306/000182330623000008/earningsreleasefy23q3.htm",
"source_date": "2023-02-02",
"tier": "E0"
}
]
},
"C-40": {
"claim": "Lightspeed's disclosed market-capitalization trigger date matches its impairment recognition date, giving a disclosed lag of zero quarters.",
"territories": [
"2"
],
"load": "LOAD-BEARING",
"tier": "E2",
"sources": [
{
"source": "Inference from the Lightspeed Q3 fiscal 2023 earnings release records in this ledger",
"source_id": "https://www.sec.gov/Archives/edgar/data/1823306/000182330623000008/earningsreleasefy23q3.htm",
"source_id_norm": "www.sec.gov/archives/edgar/data/1823306/000182330623000008/earningsreleasefy23q3.htm",
"source_date": "2023-02-02",
"tier": "E2"
}
]
},
"C-41": {
"claim": "BCE recorded $2,113 million of non-cash asset impairment charges against Bell Media in the quarter ended September 30, 2024.",
"territories": [
"2"
],
"load": "SUPPORTING",
"tier": "E1",
"sources": [
{
"source": "BCE third quarter 2024 results news release distributed by PR Newswire",
"source_id": "https://www.prnewswire.com/news-releases/bce-reports-third-quarter-2024-results-302298110.html",
"source_id_norm": "www.prnewswire.com/news-releases/bce-reports-third-quarter-2024-results-302298110.html",
"source_date": "2024-11-07",
"tier": "E1"
}
]
},
"C-42": {
"claim": "Lightspeed Commerce recognized a goodwill impairment charge of US$748.7 million in the three months ended December 31, 2022.",
"territories": [
"1"
],
"load": "LOAD-BEARING",
"tier": "E0",
"sources": [
{
"source": "Lightspeed Commerce Inc. audited consolidated financial statements, fiscal year ended March 31, 2023",
"source_id": "https://s28.q4cdn.com/517092977/files/doc_financials/2023/q4/Financial-Statements-FY23-Q4-Final.pdf",
"source_id_norm": "s28.q4cdn.com/517092977/files/doc_financials/2023/q4/financial-statements-fy23-q4-final.pdf",
"source_date": "2023-05-18",
"tier": "E0"
}
]
},
"C-43": {
"claim": "IAS 36 mandates only annual goodwill testing at a fixed date, so an indicator arising mid-year can defer recognition by up to three quarters.",
"territories": [
"4"
],
"load": "SUPPORTING",
"tier": "E2",
"sources": [
{
"source": "Inference from Grant Thornton, IAS 36 - If and when to undertake an impairment review",
"source_id": "https://www.grantthornton.global/en/insights/articles/IFRS-ias-36/ifrs-ias-36-If-and-when-to-undertake-an-impairment-review/",
"source_id_norm": "www.grantthornton.global/en/insights/articles/ifrs-ias-36/ifrs-ias-36-if-and-when-to-undertake-an-impairment-review",
"source_date": "2021-08-03",
"tier": "E2"
}
]
},
"C-44": {
"claim": "Ballard Power Systems recognized a goodwill impairment loss of US$40,277 thousand on testing performed as at September 30, 2024.",
"territories": [
"1"
],
"load": "LOAD-BEARING",
"tier": "E0",
"sources": [
{
"source": "Ballard Power Systems Inc. audited consolidated financial statements, year ended December 31, 2024, goodwill note",
"source_id": "https://www.ballard.com/wp-content/uploads/2025/03/Q4-24-Financial-Statements-FINAL.pdf",
"source_id_norm": "www.ballard.com/wp-content/uploads/2025/03/q4-24-financial-statements-final.pdf",
"source_date": "2025-03-12",
"tier": "E0"
}
]
},
"C-45": {
"claim": "Hayn Hughes report goodwill write-offs lag economic impairment of goodwill by an average of three to four years.",
"territories": [
"4"
],
"load": "LOAD-BEARING",
"tier": "E0",
"sources": [
{
"source": "Hayn & Hughes, Leading Indicators of Goodwill Impairment, SSRN abstract page",
"source_id": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=850705",
"source_id_norm": "papers.ssrn.com/sol3/papers.cfm",
"source_date": "2005-11-21",
"tier": "E0"
}
]
},
"C-46": {
"claim": "Ballard Power Systems recognized a goodwill impairment loss of US$23,991 thousand during 2023 relating to Ballard Motive Solutions.",
"territories": [
"1"
],
"load": "LOAD-BEARING",
"tier": "E0",
"sources": [
{
"source": "Ballard Power Systems Inc. audited consolidated financial statements, year ended December 31, 2024, comparative period",
"source_id": "https://www.ballard.com/wp-content/uploads/2025/03/Q4-24-Financial-Statements-FINAL.pdf",
"source_id_norm": "www.ballard.com/wp-content/uploads/2025/03/q4-24-financial-statements-final.pdf",
"source_date": "2025-03-12",
"tier": "E0"
}
]
},
"C-47": {
"claim": "Corus Entertainment recognized a $350.0 million goodwill impairment charge in the quarter ended August 31, 2022.",
"territories": [
"1"
],
"load": "LOAD-BEARING",
"tier": "E0",
"sources": [
{
"source": "Corus Entertainment Inc. annual MD&A with audited consolidated financial statements, year ended August 31, 2023",
"source_id": "https://assets.corusent.com/wp-content/uploads/2023/10/Corus-Entertainment-Annual-MDA-and-Financial-Statements-Oct-30-2023-vf.pdf",
"source_id_norm": "assets.corusent.com/wp-content/uploads/2023/10/corus-entertainment-annual-mda-and-financial-statements-oct-30-2023-vf.pdf",
"source_date": "2023-10-30",
"tier": "E0"
}
]
},
"C-48": {
"claim": "BCE's market capitalization exceeded its book equity at every 2024 quarter-end.",
"territories": [
"3"
],
"load": "SUPPORTING",
"tier": "E3",
"sources": [
{
"source": "Model prior, no external source retrieved in this run",
"source_id": "none",
"source_id_norm": "none",
"source_date": "",
"tier": "E3"
}
]
}
}