Known limitations

Everything is read from filings made to the SEC in US GAAP. Where that does not exist, the answer is no answer rather than a worse one.

• Scores for most companies that file on Form 20-F or 40-F. Their statements are read and shown as filed, in the currency they were filed in, along with the ratios that can be worked out from those figures alone. What they do not get is a score, or any figure that divides their statements by a share price. Scoring IFRS figures against bands built for US GAAP produces a confident number that means nothing, and a price here is in US dollars per depositary share while the statements are in the company's own currency per ordinary share, so the two cannot be divided into one another. A foreign issuer that reports under US GAAP in US dollars has neither problem and is scored like any other company, which is a small minority of them.
• Funds, commodity trusts and volatility products. A trust that files a 10-K still has no revenue, no operating margin and no invested capital, so a business-quality score is a precise answer to a question that does not apply.
• Companies whose registrant covers several tickers, where the filing describes the registrant rather than any one of them.
• Anything that has never filed. Private companies, and listings too new to have reported.
• The financial strength tests skip banks, insurers, REITs and real estate companies. They are Piotroski's nine, and he excluded financial firms because the tests read working capital, gross margin and long-term borrowing, none of which mean the same thing on a bank's balance sheet. Those companies still get a scorecard, a valuation and everything else; only that one card says why it is empty.
• The financial strength tests also need two comparable annual reports on Form 10-K, so a company in its first year or two of reporting has none yet.
• The distress score skips financial companies for the same reason, and skips any filer that does not report one of the five figures it needs. Deere presents an unclassified balance sheet, so it has no working capital and therefore no score; a partial sum would be a number on no scale rather than a lower one.

A blank is a decision. Every figure is checked against the period the filer actually reported it for, and one that cannot be tied to a current period is removed rather than shown.

• A figure whose source line the company stopped using is dropped. Left in, it silently reports a number from years ago as though it were current, which is the single worst defect this product has had.
• Return on equity and debt to equity are withheld when a company's book equity is negative. The arithmetic works and the sign means the opposite of what the label says, so the ratio is not shown at all.
• Free cash flow is missing for some property companies. Capital spending for a business whose product is buildings has no settled definition, and picking one would change the answer without saying so.
• Fourth quarters are often absent. A quarter is only shown when it can be derived from figures the filer tagged consistently. Where the company changed how it tags mid-year, the quarter is refused rather than assembled from two different things.
• The earnings-quality figure carries no verdict, and that is deliberate rather than unfinished. Sloan's finding ranks companies against each other and this site has no ranked universe to place one in, so the number is published with its inputs and nothing grades it. Two traps if you read it yourself: the typical value is around minus four and a half percent rather than zero, because depreciation costs no cash; and a company making a loss produces a strongly negative figure that looks clean for the wrong reason.
• The distress score reads a healthy airline, grocer or restaurant chain as riskier than it is. Those businesses collect from customers before paying suppliers, so negative working capital is how they are meant to run, and Altman's model counts it against them. Delta and Walmart both read low for that reason and neither is in trouble. The zone is a starting point for a question, not an answer.
• Individual financial strength tests drop out where the line behind them is not reported, which is why that card counts out of the tests it could run rather than always out of nine. AT&T tags no gross profit; Deere presents no classified balance sheet and so has no current ratio. A company that reports no long-term debt is read as unknown rather than as zero, so a genuinely debt-free company loses that test rather than being credited with a fall it did not have.

Arithmetic produces every number on the site. The AI writes the sentences around them, and it is never asked to calculate. That split bounds the damage: an explanation can be wrong, and the figure beside it is still right.

• The written analysis can misread a filing, emphasise the wrong thing, or sound more certain than the evidence supports. Read it as a starting point, not a conclusion.
• Analyses are cached and shared. The one you see may have been generated for someone else earlier, and the date it was generated is shown above it.
• Nothing on this site is investment advice, and the ratings are not recommendations.

• Quotes refresh roughly every 30 seconds and can be briefly unavailable at busy moments. When that happens the page says which, rather than showing a blank price.
• Filings appear here after the SEC publishes them, which is not instant.
• Everything except the live price is computed from the most recent filing, so it describes the last period the company reported and not today.

The disclosure pages are a record of what members of the US House reported to the Clerk of the House. They are not a record of what anyone owns, and there are four things they cannot tell you.

• Amounts are the ranges the form asks for, not figures. A report says $1,001 - $15,000 rather than what was actually spent, and that band is shown exactly as filed. It is never converted into a single number.
• A disclosure date is not a transaction date. Members have 45 days to report a covered transaction, so the two are routinely weeks apart. Both are shown as separate columns.
• About one report in eight is filed on paper and published as a scan with no machine-readable text. Those transactions are not listed. The reports are still linked, they are counted on the page, and a member who files that way is told so on their own page, because otherwise a thin record would read as somebody who does not trade.
• A ticker is shown as the filer typed it, and is only linked to a company page when the filed symbol and the company name agree. Where they do not, the match is labelled as probable and left unlinked: attaching a public official's trade to the wrong company is worse than leaving it unattached.

Both chambers are covered, and they publish differently. The House files a PDF per report, some of them scans; the Senate publishes a table, so its reports are read in full. Senators also hold far more municipal bonds, treasuries and funds, which have no company page to link to, so a smaller share of their transactions carries a ticker here.

Things that work but are not finished, while Materiality is in development.

• Only the newest browsers on desktop and Android have been tested end to end. Safari and older browsers should work and have not been verified.
• Confirmation and password-reset emails are subject to a daily ceiling. If one does not arrive, ask for another tomorrow or write to us and we will confirm the account by hand.
• Some pages are dense on a small screen. They work, and they were designed on a large one.

This is the most useful thing you can do, and a specific report beats a general one every time. The ticker, the figure, and what you expected instead is enough to find almost anything.

Report a problem, or write to norrowstudiossupport@gmail.com.

If you want to check a figure yourself first, the methodology page says exactly how each one is calculated, and every stock page links to the filing it was read from.