Match each number it can read in a document to the SEC filings, with a status and a citation for each.
Upload a document: each number the check can read is matched to the latest filing that reports its period (or the figure as first filed) and listed with its status and the filing it was checked against. Free pilot.
Matches figures to filings and describes what they report about the company · Free pilot
The things people actually arrive here holding
The company restated. Is my old figure still the filed one?
A later filing can report a different number for the same period. Both filings are named by accession, so a figure copied from an older report is told apart from the current one.
Read the method →Does my firm's own public record hold together?
The RIA desk reads the firm's ADV, its website and the filings of its vendors, and lists each finding with the source quote behind it. It is an evidence aid for the CCO, not a compliance determination.
Read the method →How do I know this isn't just an AI opinion?
The signals are deterministic Python over the XBRL a company filed, and every figure carries the EDGAR document and as-of date behind it. Where a filing does not disclose something, the metric is withheld and says why, rather than being estimated.
Read the method →Not a score. The line item that moved, and what it changes.
Where the evidence is, rather than a number quoted at you
The proof page
The point-in-time backtest and the case studies, with the limits stated beside them rather than in a footnote at the bottom.
See the proof →The case library
The signal classes this engine reads, replayed on documented accounting failures using only the filings that were public before each collapse.
Open the case library →The calibration panel
How the grade bands separated across the backtest panel, and where that separation is at its weakest — read live from the engine on every page load, never typed into this copy.
Read the live panel →Match each number it can read in a document to the SEC filings, with a status and a citation for each.
Upload a PDF or paste the text. Each number the check can read is matched to the latest filing that reports its period (or the figure as first filed) and listed with a status and the filing it was checked against, in a form you can download; numbers it cannot read are listed as not read. Free pilot; an evidence aid, not a compliance determination.
Every figure traceable to the filing · No credit card required
Illustrative example of the output format — not a live or current verdict for these names.
The same forensic checks, replayed before the collapse
We don't claim to predict the future. We show something checkable: the kinds of signal our engine reads — among them returns on invested capital, cash returned beyond free cash flow, share-count growth, accruals and inventory against cost of sales — were sitting in the public SEC filings before Carvana, Bed Bath & Beyond and Peloton came apart. Reproducible, no AI guesswork.
And we don't hide behind a fake price target. Every name gets a range of 12-month outcomes, read off past panel names with this name's volatility, grade, size and position in its range, with a fat downside tail, not a single number. That range is a spread of what happened, re-scored on /proof, not a forecast. What we can measure is below, with its limits. The whole track record is public.
See the full proof & calibration →The answer is in the filing. Nobody has time to read it.
Whether a business is healthy, whether the earnings are real, whether the price makes sense — it's all disclosed in the 10-K and 10-Q every company is legally required to file. But that's 200 pages of footnotes per name, per quarter, and the signal that matters — cash that never matches profit, leverage outrunning coverage, revenue outrunning receivables — sits buried where a quick glance never finds it. Stockonomy runs the forensic checks an analyst would do by hand, on every filing, and shows the work.
The forensic work-up an analyst would do by hand
Built on the same primary SEC data institutions pay for — turned into a reproducible read on the risk and the reward, with every figure traceable back to the filing.
Forensic distress & earnings-quality screens
How close the balance sheet sits to the shape companies had the year before they restructured, and whether this year's accruals, margins and asset mix moved together in the combination seen at filers later found to have overstated earnings — the same screens institutions run — plus cash-conversion, accruals, DSO and DIO checks. The first question is always 'is this business sound underneath the price', not 'is it going up'.
Every number traceable to the filing
Figures come from XBRL, each one source-linked to the exact EDGAR document and as-of date. A value the primary filing doesn't support isn't estimated or filled in from a data vendor — it's left out, so a number that appears on the page is a number that came from the filing.
Price-aware verdict, with the reasoning
Whether the price is justified by the quality, stated outright. One coherent verdict — sound and priced in line with its quality, pricey but defensible, or carrying more risk than the reward supports — built as a causal chain from filing to line item to consequence, not a black-box score.
The whole filing, read and queried
A deep read of the real 10-K, 10-Q, 8-K and Form 4 behind every name, plus unlimited Q&A that retrieves over the complete filing — ask about a segment, a covenant, dilution, or a footnote and get the passage, not a guess.
Management tone, tracked across years
We compare management's own words across years of filings to surface when confidence quietly shifts — new hedging around a segment, a customer, or liquidity. Context on the people running the business, grounded in what they actually wrote.
Insider activity & dilution, in context
Form 4 buying and selling (10b5-1 plans separated out) and share-count dilution or buybacks — read as signals of how the people closest to the business see its value, explained rather than just listed.
Reproducible proof, not back-fit claims
The forensic checks are deterministic Python, calibrated point-in-time against history — and replayed on data truncated before each well-known collapse so you can check the signal classes were visible in the filings beforehand.
A weekly sector teardown
Each week we run an entire sector through the same engine and email a plain-English read on which names look sound, which look stretched, and which the screens flag. A two-minute skim.
How it works
Enter a ticker
Any name you're researching or already hold. Stockonomy pulls its real EDGAR filings — 10-K, 10-Qs, 8-Ks and Form 4s — and the XBRL behind them.
The engine runs the forensics
Distress and earnings-quality screens, cash-conversion and accrual checks, sector-relative valuation, capital allocation and management tone — deterministic, calibrated, and source-linked to the filing.
You get one coherent verdict
A single price-aware read — sound and priced in line with its quality, pricey but defensible, or carrying more risk than the reward supports — written as a causal chain you can follow, not a black-box score.
Track it, and get flagged on change
Add names to a watchlist and Stockonomy re-checks them after each new filing, emailing only the material changes — a downgrade in quality, a forensic disclosure, an insider cluster.
The analysis is free during the pilot.
The full forensic work-up — on any US company that files a 10-K, any time you ask — is free during the pilot, for individuals, advisors and firms, with no card taken. We don't sell anything that identifies you.
Why it costs us little
The forensic core is deterministic Python over free, public SEC EDGAR data — not an expensive model running on every screen. The heavy computation is cached and shared across users, so the cost of one more analysis stays near zero.
Where it stands
The analysis is free during the pilot, for individuals, advisors and firms; no card is taken.
What we won't do
We don't sell data that identifies you — not your email, not your individual activity — and we take no order flow. We're not an ad business monetizing your attention, and the forensic verdict itself is never for sale: no company can pay to change its grade. Our incentive is that the analysis is right.
Free during the pilot. Run a practice or a platform? Pre-publication figure check, pilot →
Publishing figures from filings? Check them before they go out.
When a note quotes a company's revenue, margin or debt, each figure can be checked against the SEC filing it came from, by a check that does not use the model that wrote the note. In the pilot a one-page record is prepared with you for each document, and your firm holds it.
Stockonomy for professionals →Free while in pilot. A single figure can be checked free, without an account: check a figure.
Check a firm's public record against the filings
The RIA desk builds an evidence pack, checks marketing claims, compares ADV with what the firm says elsewhere, and watches the vendors you name for cybersecurity 8-Ks and changes of accounting firm. Each finding carries its source quote. An evidence aid, not a compliance determination.
Open the RIA desk →A plain-English read on a whole sector — every week
Each week we run an entire sector through the same engine — financial-health screens and valuation — and email you which names look solid, which look overpriced, and which carry more risk than the reward justifies. A two-minute read. No account needed. Unsubscribe anytime.
Informational only · not investment advice
The thesis lives or dies in a filing you'll never finish reading
Stockonomy runs the forensic work-up for you — so you know the risk you're taking and whether the price justifies it, on any name you research or hold.
Analyze a stock — freeNo credit card · Every figure traceable to the filing