Methodology

Red Flags Detection

Automated monitoring of SEC filings for disclosure events that historically correlate with material stock underperformance. Red flags are extracted from 8-K, 10-K, 10-Q, and NT filings using pattern matching and NLP.

Validation Status

Red-flag detection claims are tied to the committed SEC corpus and the late-filing backtest in the claims registry, including the revised evidence when observed returns differ from older launch copy.

Backed claimfiling-red-flag-10kq-categories-6

6 count

README red-flag categories are backed by deterministic rules in src/services/filing-red-flags/rules/{going-concern,liquidity,auditor,legal,executive-transitions,accounting}.ts and the real SEC corpus conformance test at tests/conformance/filing-red-flags/real-sec-corpus-f1.conformance.test.ts, which validates category coverage over 40 committed 10-K/10-Q SEC excerpts labeled in tests/fixtures/sec/real-red-flag-corpus/manifest.json. Supporting coverage includes tests/conformance/filing-red-flags/rules.conformance.test.ts, tests/conformance/filing-red-flags/litigation-differential.conformance.test.ts, tests/unit/services/filing-red-flags/text-extractor-detector.golden.test.ts, tests/adversarial/filing-red-flags/parser-adversarial.test.ts, and e2e/authenticated/filing-red-flag-surfacing.spec.ts. Evidence is limited to deterministic rule coverage over the curated corpus and related harnesses.

Test evidence: tests/conformance/filing-red-flags/real-sec-corpus-f1.conformance.test.ts

Real-data backtestlate-filing-underperformance-8.2

-1.7 percent

Real Form 4 + split-adjusted daily_returns backtest (corrected bd-rbkns methodology, >=18c7ed4ba: coverage-gap and delisting skips counted, tolerance-only confirmation) still refutes the legacy README magnitude: observed -1.70% 90-day market-adjusted return for late filings over 2021-03-15..2026-03-09, 95% CI [-2.17%, -1.23%], n=18,823 late filings. The small negative effect itself is very strongly supported (classical clustered p=1.28e-12; the classical artifact's result.evalue field is the kappa=1/2 p-to-e calibrator of that test). The evalue block here tracks the parallel Grunwald-Vovk betting e-process from late-filing-evalue-latest.json, which tests the winsorized conditional-mean null in filing-date order (presentation adjudication: bd-l6i53). The claimed -8.2% magnitude remains refuted under all of these.

Computed 2026-07-05; evidence tests/scripts/late-filing-underperformance.test.ts; docs/claims/late-filing-latest.json

Source: docs/claims.json, generated 2026-06-03.

Detection Categories

Each red flag category is assigned a severity level and contributes to the attention queue scoring. Critical flags trigger immediate alerts for subscribers.

CategorySeverityDescriptionSource

Going ConcernCriticalAuditor expresses substantial doubt about the company's ability to continue as a going concern10-K, 10-Q auditor opinion

Material WeaknessHighDeficiency in internal controls over financial reporting that could result in material misstatement10-K Item 9A, 10-Q Item 4

Auditor ChangeMediumChange of independent auditor, especially mid-cycle or with disagreements disclosed8-K Item 4.01

RestatementHighRestatement of previously issued financial statements due to error or irregularity8-K Item 4.02, 10-K/A

Late FilingMediumFailure to file 10-K or 10-Q by deadline; NT filing indicates delayNT 10-K, NT 10-Q

CFO/CEO DepartureMediumDeparture of key executive, especially if unexpected or during sensitive period8-K Item 5.02

Delisting NoticeCriticalNotice from exchange regarding non-compliance with listing standards8-K Item 3.01

SEC EnforcementHighSEC investigation, Wells notice, or enforcement action disclosed10-K Risk Factors, 8-K Item 8.01

Scoring Integration

Red flags contribute to the attention queue score with a base weight of 0.12. The contribution scales with severity:

critical_multiplier = 1.0
high_multiplier = 0.7
medium_multiplier = 0.4

red_flag_score = max(severity_multiplier for active flags)
attention_contribution = 0.12 * red_flag_score

Only the highest-severity active flag is counted to avoid double penalizing companies with multiple related disclosures (e.g., going concern typically accompanies material weakness).

Data Sources

SEC EDGAR: Real-time RSS feed for 8-K, 10-K, 10-Q, NT filings
Refresh cadence: Every 15 minutes during market hours
Historical lookback: 24 months for scoring purposes
Pattern matching: Rule-based extraction with NLP fallback

Limitations

US-listed equities only; foreign private issuers may have different disclosure forms
Pattern matching may miss novel disclosure language or non-standard formatting
Some flags (e.g., SEC investigation) may be disclosed late or not at all
Historical correlation does not guarantee future predictive power

Citations & Research Basis

SEC EDGAR company filing feeds are the source of 8-K, 10-K, 10-Q, and NT filing disclosures.
Internal conformance tests over a real SEC corpus validate category coverage before these flags appear in launch surfaces.
Late-filing impact estimates are treated as empirical backtest results, not static marketing copy.

See also: Attention Queue | Insider Signals