Methodology

How each signal is produced, and what it does and does not mean.

“Signal” means an observable change in what a company disclosed — not a prediction. This site reports that a disclosure changed, quotes the filing that changed it, and links the source. It does not forecast outcomes, and the association rates on the sequences page are descriptive rather than causal. Counts describe the filing history covered so far, which is still being extended backwards.

How much is flagged

A count of events means little without the population it came from. Across the filing days processed so far:

Filings in the daily index: 357,632
Read in full: 23,815
Events published: 4,085
Flagged: 17.2%

Most filings are never candidates: fund and securitisation paperwork makes up the bulk of daily volume and cannot carry these signals. The flag rate is deliberately low — these are meant to be rare events.

Where the data comes from

Every weekday the pipeline reads the SEC's EDGAR daily index — the official list of everything filed that day. From operating companies it keeps the item-coded current reports (8-K), the periodic reports whose narrative it compares (10-K, 10-Q, 20-F, 40-F and their amendments), late-filing notices (NT 10-K, NT 10-Q, NT 20-F), and SEC review correspondence (UPLOAD, CORRESP). Fund and securitisation paperwork — the majority of daily volume — is discarded, because it cannot carry these signals. The index is read from two of EDGAR’s files, generated separately in different formats; if the first is missing or looks cut short, the second is used, and a business day that yields nothing from either is left for the next run rather than recorded as empty.

Restatements and auditor changes

These come from the SEC's own structured item codes, read from each filing's header. Item 4.02 is "Non-Reliance on Previously Issued Financial Statements"; Item 4.01 is "Changes in Registrant's Certifying Accountant". Because the filer selects these codes themselves, a match is a fact about what was disclosed, not an inference. These are labelled SEC item code.

The sub-classification is the signal. The same item code covers materially different events, so the filing itself is read to separate them. Under Item 4.02(a) management or the board reached the conclusion; under 4.02(b) the auditor told them, which is more serious because the company did not find it itself. For Item 4.01, an auditor resigning is a stronger signal than a company dismissing one, a disclosed disagreement stronger still, and a move from a large firm to a much smaller one is ranked above a rotation between equals.

What Item 4.02 does not mean. It means the company told investors that earlier figures should no longer be relied upon. It is not by itself evidence of fraud or misconduct. Restatements frequently arise from technical accounting corrections.

Late filings

A company that cannot file on time is often the earliest public warning, and it usually precedes a restatement rather than following it. These come from Form 12b-25 (filed as NT 10-K or NT 10-Q).

Beyond the fact of lateness, the form asks whether the company anticipates a significant change in results of operations, answered as a checkbox. A "yes" is the company pre-announcing that the delayed numbers will look materially different, and those are marked elevated. So are late annual reports, and companies that are also behind on their other periodic reports.

Going concern

This is the signal most often done badly, so it is worth being precise.

Searching filings for the phrase "going concern" does not work. The phrase appears in the Risk Factors of most speculative issuers as standing boilerplate, unchanged for years. It also appears in forward-looking-statement disclaimers. None of that is news.

Instead the pipeline locates the actual accounting note — the disclosure made under ASC 205-40 — and reads its conclusion. That note always opens by reciting both possible outcomes as methodology, so the conclusion sentence, not nearby wording, decides the classification. Each filing is placed on a ladder:

No going-concern disclosure
Risk-factor language only — boilerplate, no accounting conclusion
Substantial doubt raised, alleviated by management's plans
Substantial doubt about the ability to continue as a going concern

An event is published only when a company moves between rungs, compared against its own previous filing of the same type. A company that has disclosed substantial doubt for eight consecutive quarters generates no event, because nothing changed.

SEC comment letters

When the staff reviews a company's annual or quarterly report they may write to it with questions about the accounting. Those letters (UPLOAD) and the company's replies (CORRESP) are published on EDGAR. The comment letters page groups them by the accounting topic raised.

A comment letter is not a finding. The SEC is required to review every reporting company at least once every three years, so receiving one is routine, and most reviews close with no change to the accounts. A letter records a question that was asked — nothing about the answer.

They are historic, not current. The SEC releases correspondence only after a review closes, no earlier than 20 business days afterwards. In practice a median of 209 dayspasses between a letter being written and appearing on EDGAR, so every entry shows both dates.

Letters are grouped into reviews. One review is one conversation — the staff write, the company replies, sometimes several times over months — and EDGAR publishes each letter separately. Grouping on the company and the filing under review means a single review reads as one exchange rather than eight near-identical entries.

Most staff letters review registration statements rather than periodic reports. Those are a different activity — a company raising money, not a review of published accounts — so the "Re:" line is parsed and only periodic-report reviews are kept.

Material weakness

Every annual report states management's conclusion on internal control over financial reporting under Item 9A. A material weakness means a deficiency severe enough that a material misstatement might not be prevented or detected in time.

As with going concern, only a change is published: control reported effective and then not, or a weakness reported and then cleared. A company carrying the same weakness for several years produces no event.

Where a weakness clears, the site reports how long it had been reported. That cannot be read off the filing announcing the fix — it requires walking back through the company's own annual reports until control was last reported effective. Where that history is unreadable or runs out, no duration is shown rather than a floor presented as a fact.

Note that a filing reciting the definition of a material weakness is not disclosing one; the definition appears in filings that report no weakness at all.

Finance chief departures

Item 5.02 covers appointments and departures of directors and officers. It is one of the commonest 8-K items, so the item code alone is noise. Only departures of the CFO, chief accounting officer or controller are kept — the roles that own the financial statements — and the role and the departure must appear in the same sentence, or a director's resignation gets joined to whichever CFO is named further down.

Whether a successor was named, an interim appointed, or the seat left vacant is recorded, as is whether the filing states there was no disagreement. That last sentence is standard and its presence is reassuring, not adverse — reading it as a disclosed disagreement is a mistake this pipeline made and corrected.

A finance chief leaving is an ordinary event. It is published because it is worth knowing alongside the other signals, not because it implies anything is wrong.

Accounting standard newly cited

The pipeline reads every Accounting Standards Update (ASU) each filing refers to, and reports one that appears in this filing but not in the same company’s previous comparable filing, where the sentence describes it as adopted rather than merely issued.

This measures what the company discloses, not necessarily what it did this period. A filing can cite a standard for the first time long after adopting it — one filer cited ASU 2016-02 in 2026 in a sentence saying it had adopted it in 2019. Where the filing states an adoption date, it is published beside the citation, and anything two or more years old is marked as a first citation rather than a change of policy. Amendments are excluded, because an amendment that re-files a note would appear to cite everything in it for the first time.

An earlier version compared the wording of the whole policy note. It was abandoned: filings have no consistent note structure, so the extractor swept adjacent notes into the comparison and mostly measured moving dollar figures, flagging 15 of 18 periodic filings in a single day.

Revenue recognition beta

The pipeline extracts the revenue-recognition note, strips out figures, dates and quarter labels, and compares the wording against the same company's previous comparable filing. Comparisons are always like-for-like — a 10-Q against a 10-Q — because a quarterly note is far shorter than an annual one and comparing across them would flag every company every year.

These are marked beta. They are derived from text comparison rather than a structured code, so they carry more noise. Treat them as a prompt to read the filing, not as a conclusion.

The role of AI

Detection is entirely deterministic: pattern matching, item codes and text comparison. No model decides whether something is a signal.

When enabled, a language model adds a plain-English summary and screens beta text-diff findings for cosmetic rewrites. Anything it writes is labelled AI summary and always appears alongside the verbatim filing text and a link to the source. If the analysis layer is off or fails, events still publish — with evidence and quotes intact.

Limitations

Only filings from the days the pipeline has processed. It is not a historical database.
Section extraction depends on filing structure. Unusual formatting can cause a note to be missed.
Companies with no prior comparable filing produce no comparison signals: a change cannot be shown without a baseline.
Amended filings are compared against the original series, which can surface wording differences that are procedural rather than substantive.
Comment letters describe reviews that closed months earlier, so they are a record of what was asked at the time, not of anything current.
Counts show what has been flagged over the days processed so far, which is not a base rate to compare a company against. History is still being filled in backwards, so the period covered grows each night.

Corrections

Every entry links to its source. If an entry misreads a filing — a mischaracterised disclosure, a quote taken out of context, a company matched to the wrong filing — it should be corrected, and corrections are welcome from anyone, including the filer.

Raise it at https://github.com/srinivasledger/filing-signals/issues with the accession number and what the entry should say. There is no editorial gatekeeping: the detection rules are open, and the fix lands in the next scheduled run.

Entries are generated automatically and are not reviewed by a person before publication, so mistakes reach the page. That is the reason every entry carries the filing text it was derived from and a link to the original: the source is always checkable against the claim.