Where the numbers come from
Company financials come from the reports companies file with the US Securities and Exchange Commission (10-K and 10-Q). We keep every figure's source — filing, period and filing date — so any number on the site can be traced back to the report it came from. Companies label the same item in different ways; we map those labels carefully, and where a figure can't be determined reliably we leave it out rather than guess. Market prices come from third-party market data and may be delayed.
How figures are calculated
Every metric is calculated in code, the same way for every company: growth rates, margins, free cash flow (cash from operations minus capital spending), twelve-month totals and comparisons with a company's own history and its peers. Where a company reports only year-to-date totals, quarterly figures are derived from them. Share counts and earnings per share are restated after stock splits so years compare like for like.
Pre-earnings estimates
Pre-earnings articles show analyst consensus, numeric guidance from the company's latest earnings release when available, and a separate historical-surprise model. The model uses the median EPS surprise from the latest four to eight reported quarters, shrunk toward zero as if four neutral quarters had been observed. Its beat probability uses a Beta(2, 2) prior. We show it only when at least four quarters have comparable actual and consensus EPS data. It is a transparent, small-sample reference estimate, not a calibrated forecast; it does not change or replace company guidance or analyst consensus. Missing or incomparable data is omitted.
Signals
Our signals flag notable changes in a company's reported figures — for example growth speeding up or slowing down, margins widening or narrowing, earnings running ahead of cash, share counts rising or falling, a balance sheet growing faster than the business, or results well above or below analysts' expectations. The set is drawn from published research on financial statements, and we regularly test it against history, keeping signals that prove informative and retiring those that don't. Each signal is an observation about reported data, never a recommendation on its own.
The fundamentals grade
Each covered company gets a letter grade (A+ to E) summarising where its reported growth, profitability, cash generation and balance sheet sit against the other companies we cover in its industry. It describes the financials, not the share price, and is not a rating of the shares. A company that fails our data-quality checks gets no grade.
Financial-health checks
The health section on each company page runs seven checks on the latest balance sheet:
- Short-term assets cover short-term liabilities.
- Short-term assets cover long-term liabilities.
- Cash exceeds debt.
- Debt is below 40% of shareholders' equity.
- Debt relative to equity is no higher than five years ago.
- A year's operating cash flow covers at least 20% of debt.
- Operating profit covers interest at least five times.
A check counts only when the company reports the figures it needs. The five-measure profile shows the company against its peers on growth, profitability, cash generation and balance sheet, plus where the share price sits against our value range.
Moat score
Warren Buffett describes a good business as a castle protected by a moat: an advantage that stops rivals competing its profits away. The moat score reads the evidence for that in up to ten years of the company's own filings, in five pillars worth 20 points each:
- Returns on capital: the median cash operating profit as a share of total assets (return on assets for banks and insurers). Full marks at 15% or more.
- Pricing power: the median gross margin (full marks at 60%) and how little it has moved from year to year.
- Durability: the median operating margin (full marks at 30%) and how well the worst year held up against it.
- Cash conversion: the share of years with positive free cash flow and the median free-cash-flow margin.
- Discipline: the change in the diluted share count over five years and stock-based pay as a share of revenue.
The score is the points earned as a share of the points available (banks and insurers are scored without pricing power and cash conversion, which do not describe them). We then deduct 15 points where the industry is highly exposed to AI disruption, because its product is information work that AI systems increasingly perform, and 5 points where exposure is medium. Exposure comes from a fixed table of industries, not a model's opinion. Bands: Strong 70 and above, Moderate 45 to 69, Weak below 45. Companies with fewer than five years of comparable figures are not assessed. The score is a rules-based reading of reported figures; it does not feed our valuations or ratings and is not investment advice.
Highlights
Strengths and things to watch are reported facts that cross fixed thresholds: fast or slowing growth, high or narrowing margins, strong or weak cash generation, buybacks or dilution, heavy stock-based pay, debt levels, how the shares are valued against our model, how volatile they are, insider selling, short interest, results against analysts' expectations, and risks the company itself names. Each line states the figure behind it.
Valuations, ratings and price targets
How we value companies, and how ratings and price targets are set, is explained in full on how we value companies, with the conflicts and disclosures in our recommendations disclosure.
How AI is used — and not used
AI helps with drafting and readability. It receives verified, pre-calculated figures and their sources; it never supplies numbers, does the arithmetic or invents citations. Every number in a draft is checked against our calculations, and a person reviews and approves every article before it is published.
Corrections
When we find a material error we correct it and note the correction with a date. To report an error, see the contact details on the about page.