Every company on CrossAssetData carries a grade out of ten, read from its own filings with the SEC. This page says what the grade reads, how the pieces add up, and what it is not.
Three tenses, and in each a look at the business from three sides: the company itself, its customers, and the people who own it. The words below are the site's own method file, the same words every company file and the desk read.
| ItselfHow the business holds up on its own | Its customersWhat it delivers to the people who buy from it | Its shareholdersHow it treats the people who own it | |
|---|---|---|---|
| The pastFrom the record | Surviving crises | Track record | Shareholder treatment |
| The presentAs of today | Financial health | Industry position | Leadership |
| The futureReasoned forward | Room to expand | Growth drivers | Expectations vs reality |
Has it survived what it has been through?
What it reads. We read every filing of the company’s last ten years, from its own SEC index, for warning events: bankruptcy or receivership notices (Form 8-K, item 1.03), debt default or acceleration notices (item 2.04), notices about its exchange listing (item 3.01), warnings not to rely on past financial statements (item 4.02), auditor changes (item 4.01), material cybersecurity incidents (item 1.05), and late-report notices (Forms NT 10-K and NT 10-Q). Each kind found adds to a burden: three points for the first four kinds, one for the rest, halved when the latest of that kind is more than three years old. The burden is ranked against the company’s industry family, so fewer points read higher; a clean record shares the top rank with every other clean record in its family. Recomputed from the filing record on every data build. Going-concern language, an auditor’s stated doubt that the company can keep operating, is not in the scan yet; where the desk has read a filing, its read says so.
What has it actually done, year after year?
What it reads. Up to ten fiscal years of filed revenue and net income, read from EDGAR’s company facts: how fast revenue grew a year over the whole span, the share of years it rose, the share of years with a profit, and the deepest fall from a revenue peak. Each part is ranked against the company’s industry family and the metric is the average of the parts that can be read, at least two of them; fewer than three filed years of revenue read nothing. Recomputed on every data build.
How has it treated the people who own it?
What it reads. Two readings of how the company treated the people who own it, from its own filed statements: the diluted share count over the last three fiscal years (new shares dilute every owner; a shrinking count does the opposite), and the cash returned as dividends and buybacks against the operating cash flow of those years (read only when that cash flow is positive, capped at all of it). Each is ranked against the company’s industry family and the metric is the mean of the two; one alone leaves it unseen. Beside it, as the record and not the score, the year of Form 4 filings: every purchase and sale of the company’s stock by its officers and directors on the open market, with the sales made under scheduled 10b5-1 plans marked as such. The insiders’ own trades do not move the score, because the record shows insiders buy most where companies are in trouble.
Is it financially healthy right now?
What it reads. Four standard checks on the newest annual balance sheet and income statement on file, read from EDGAR’s company facts: leverage, what it owes per dollar of assets (assets less shareholders’ equity, over assets); return on equity, the profit earned per dollar of shareholders’ money (not read when equity is below zero); gross profitability, gross profit per dollar of everything the company owns (revenue less the cost of revenue where gross profit is not tagged); and net share issuance, how the diluted share count moved against the prior fiscal year. Each check is ranked against the company’s industry family and the metric is the average of the checks that can be read, at least two of them. Statements whose fiscal year ended more than two years ago are not read as the present. Recomputed on every data build.
How strong is its position in its industry?
What it reads. The company’s standing in its industry family in its newest filed year: its revenue against the family’s (ranked only among the members that report in the family’s main currency), its net margin, and its revenue growth a year over the last three filed years. Each part is ranked against the family and the metric is the average of the parts that can be read, at least two of them. Recomputed on every data build.
What are the people in charge doing?
What it reads. Three readings of the people in charge, from the SEC’s insider filings (Forms 3, 4 and 5) and the company’s own current reports: the officer and director changes reported on Form 8-K over three years, set against the number of people in charge now (fewer changes for each person reads higher); the median years the current officers and directors have been filing; and the chief executive’s years under that title. Years count from a person’s first filing in the SEC’s data sets, which begin in 2006, so each is a floor. Each part is ranked against the company’s industry family and the metric is the average of at least two of them. The insiders’ own trades are printed as the record and do not move the score.
How much room does it have to grow?
What it reads. The room the newest filed statements leave the company to grow: its cash less its debt against everything it owns, its cash from operations less its capital spending against revenue, its gross margin, and its research spending against revenue where a research line is filed. Each part is ranked against the company’s industry family and the metric is the average of the parts that can be read, at least two of them. Recomputed on every data build.
What could actually drive growth from here?
What it reads. Whether growth is speeding up and what feeds it, from the filed years: revenue growth over the last three years against growth over the whole filed span, the newest year’s revenue growth, and the growth of research spending and of capital spending over up to three years. Each part is ranked against the company’s industry family and the metric is the average of the parts that can be read, at least two of them, from at least three filed years of revenue. Where the company tags its revenue by line of business, the lines are printed beside it as the record.
Did its newest year beat what its own record pointed to?
What it reads. The newest filed year against what the company’s own record pointed to: its revenue against the pace of the three years before it carried one more year, its net income against the average of those three years, and its net margin against its average over them. Each part is ranked against the company’s industry family and the metric is the average of the parts that can be read, at least two of them, from at least four filed years of revenue. No price or market value enters it.
Each metric is built from parts, and each part is a figure from the filings: a rate of growth, a margin, a count of warning events. A part is ranked against the other companies in the same industry family, as the SEC's industry codes group them, and turned into a percentile: near 100 is the top of the family, 50 the middle, and ties share a rank. The metric is the average of the parts that can be read, at least two for every metric built from parts, put on a scale of ten.
Every figure is read again from EDGAR, the SEC's filing system, on every data build, and each company file names the filing behind it.
Every metric on a company file says which it is. A metric read as filed is arithmetic on the company's own filings, and the file prints the figures and the filing they come from.
Where the filings cannot answer a metric, because the company is new to the record, reports too few years or leaves a line out, the desk estimates it rather than leave a hole. The estimate comes from the one part that is on file, else from the typical reading of its industry family, else from the typical reading of every company on file. It is labeled "CrossAssetData estimate" wherever it prints, with its reasoning, a confidence out of 100 and what would overturn it, usually the next filing that carries the missing figures.
Each file counts both, for example "7 of 9 filed, 2 estimated".
Each tense is a weighted average of its three metrics, and the weights depend on the kind of business, read from its industry code: a lender, a retailer, a subscription business and about two dozen other kinds. The headline grade blends the three tenses with shares set the same way, so a company with no revenue yet leans on the future and a lender on the past.
Each file names the kind of business and prints the shares.
A metric is graded on every company only after it passes a test run on the record itself, never on prices. The test goes back to each fiscal year from 2018 to 2024, reads the metric from only what was on file at that date, and sets it against what the company's later filings showed: warning events, new shares issued, profitability and growth. What counts as a pass is written down and sealed with a fingerprint, a hash, before each run, so the bar cannot move once the results are in; a metric that fails stays out of the grade until a changed version passes a new run. The results are kept as private research: the site prints no accuracy or track-record figures until the live record has earned them.
The grade covers companies. A crypto asset, a Treasury yield or a currency is read from its own record and carries no grade until a reading of its own passes the same kind of test.
The grade is research on a company's own record. It is not a price, a price target or a forecast of the stock, and it does not say whether to buy, sell or hold anything. The fair price estimate on a company file is a separate opinion with its own method, released each quarter; it does not feed the grade.
Every stock is a company. We grade the company behind the stock. The disclosures say what that makes CrossAssetData and what it is not.