Methodology

Public companies file their financial statements with the SEC tagged in a machine-readable format called XBRL. Each night this site reads the SEC’s index of that day’s filings, pulls the full reported history for any company that filed an annual or quarterly report, and stores it.

Nothing is estimated and nothing comes from a third-party data vendor. Each page names the filing its most recent figures came from and links to that company’s filings on EDGAR so you can check.

There is rarely one tag for a line item. Companies pick among several valid ones, and the choices have shifted over the years — revenue was tagged one way before the 2018 revenue-recognition standard and another way after. For each figure this site tries a prioritised list of tags and takes the first the company actually reported, which lets a single company’s history span several tagging conventions.

Where a period appears in more than one filing, the most recent filing wins, so restatements are reflected rather than the original figures. Only periodic reports count: annual and quarterly reports, never press releases or proxy statements, which sometimes carry figures on a different basis.

Each company is compared against others sharing its SEC industry classification. For every metric, a company’s position within that group becomes a percentile, flipped where a lower number is better. Category scores average those percentiles and map onto a letter.

Two guards apply. A metric needs at least five classified peers that reported it, because a percentile against two companies is noise. And a category needs at least three contributing metrics — two for dividends, which has only three in total — or no grade is published for it at all. A missing grade means the comparison was not strong enough to make, not that the company scored badly.

Every grade shows the peer count and how many metrics it used, because a B+ against six companies is a much weaker claim than the same letter against two hundred.

Officers, directors and 10% shareholders must report trades in their own company’s stock on SEC Form 4, within two business days. Those filings are read nightly and stored with the SEC’s transaction code exactly as submitted.

Most reported insider activity is not buying or selling by choice. It is compensation: shares granted, options exercised, stock withheld to cover tax. Those happen on a schedule regardless of what an insider thinks about the business. Measured across the filings loaded, only about 6% were open-market purchases. The insider trading page shows transaction codes P and S only — someone deciding to buy or sell on the open market — and company pages label every transaction with what it actually was.

Buying and selling are not symmetrical. An insider buys because they expect to make money. They sell for many reasons that say nothing about the business: diversifying, a tax bill, a house, or a trading plan scheduled months in advance. Both are shown, but selling is much weaker evidence than buying and should not be read as its mirror image.

A transaction can be dated long before the filing that reports it. Late filings are common, and one filer was seen reporting five years of small quarterly purchases in a single submission. Coverage is therefore described by filing date rather than transaction date.

Coverage is uneven. Gross profit is not a required tag and many companies never report it. EBITDA is not an accounting concept at all, so where it appears it has been derived from operating income and depreciation. Debt is reported for roughly three in five company-years; where a company splits borrowings across tags this site does not match, the figure can be understated rather than missing. Anywhere a value was not reported, the page says so instead of showing a zero.

Industry classifications are coarse. The SEC uses a scheme last meaningfully revised in 1987, which companies assign themselves. One code covers both the largest software companies and two-person firms. Comparisons are still informative, but they are not precise, which is why peer counts are always shown.

Fourth quarters are usually calculated. Most companies never report Q4 separately because the annual report covers it, so it is derived by subtracting the first three quarters from the full year — and only when every figure involved came from the same tag. Those rows are marked.

Per-share figures can change basis. Companies report each year as their filings stated it, and an annual report restates only its two prior years. After a stock split, older years keep the pre-split share count. Where that happens, per-share figures for the affected years are marked as not comparable rather than shown alongside newer ones. Totals are unaffected.

There are no valuation metrics. The SEC publishes financial statements but not share prices, and price data cannot be redistributed without a licence. So there is no price-to-earnings ratio or similar here. Those will appear only if a properly licensed source is added.

The pipeline runs nightly, reading the previous day’s filings and recalculating the metrics and grades that depend on them. Pages refresh daily. A company that filed this morning will usually appear by the next day.