How a grade is built

• Company figures: from the annual reports companies file with the US Securities and Exchange Commission (forms 10-K, 20-F, 40-F), in their machine-readable form. No value is estimated, smoothed or taken from analysts.
• Prices: daily closing prices. The valuation range is shown in euros, converted at the current exchange rate.
• Universe: US-listed companies (NYSE, Nasdaq, NYSE American, NYSE Arca, Cboe) that report both total assets and revenue to the SEC. That is 3,261 tickers, 2,409 of them with a complete grade. The list is under Universe.
• As of: grades from 13 Oct 2026. Every company page states its price date and the latest filing used.

• Each of the four grades runs from 0 to 5. It is built from two or three key figures. Each figure is placed on five steps on its own: reaching none gives 0, reaching all gives 5.
• The grade is the mean of these steps, rounded half up: 3.5 becomes 4. Every company page shows the calculation with the actual values under the derivation of each grade.
• The four grades are added without weighting. If a grade is missing because data is missing, it is not estimated: there is then no total, and the page names what is missing.
• The sentence at the top of each company page calls quality “high” from 15 points, “moderate” from 10 and “low” below that.
• The grades are absolute. They are not adjusted to the sector: in a weak industry a weak grade stays weak. For utilities and real estate companies high debt is part of the business; their balance sheet grade is therefore not raised, but the page states that it does not compare with other industries.

Balance sheet

Measured on the latest fiscal year. A company holding more cash than financial debt gets the full grade for net debt without a ratio being computed.

Swipe sideways for grades 2 to 5.

Profitability

Measured on the latest fiscal year. With negative equity, return on equity is left out.

Swipe sideways for grades 2 to 5.

Cash flow

Free cash flow is operating cash flow minus capital expenditure. The first two figures cover the latest fiscal year, the third up to five years. Banks and insurers get no cash flow grade: their cash flow follows lending and deposits, not earning power.

Swipe sideways for grades 2 to 5.

Growth

Over as many years as are available, at most five. Computed only when the first and last values are positive.

Swipe sideways for grades 2 to 5.

• For each method a multiple is formed, for example price divided by earnings per share. It is computed for each of the last up to 10 fiscal years, using the price on the balance sheet date.
• From these years the lower and upper quarter are taken (25th and 75th percentile) and applied to the figures of the latest report. That gives one price band per method. At least four years with a price and a positive basis are needed.
• A method is dropped if its multiple varies by more than a factor of 4 between its highest and lowest year: there is then no usual level. It is also dropped if its midpoint differs by more than a factor of 2 from the midpoint of the others.
• The range is the median of the lower and the median of the upper bounds of all remaining methods. Reliable means at least three methods. With one or two it says “limited”; with none, no range is shown.
• The range is not a fair value and not a price target. It says how today’s price relates to the company’s own past. In the historical test, the shares furthest below their own range actually did worst.

Which methods for whom

The industry comes from the SEC’s official industry code. For banks, liabilities are the business itself rather than its funding; revenue, free cash flow and enterprise value do not compare there. For real estate companies depreciation lowers profit without the buildings losing value at that pace; the measure there is FFO.

Caps per method

Years in which a company had almost no profit or revenue produce multiples in the thousands. They say nothing about the usual level and are not counted when above these caps.

A durable advantage, often called a moat, comes from brand, switching costs, networks or scale. The causes appear in no report, but their trace does: a high return on the assets employed that holds over years. It is measured over up to 10 years, with at least 6 years. These points do not enter the total.

• Return on assets, level (up to 3 points): operating income divided by total assets, median of the years. From 7 % one point, from 12 % two, from 18 % three. Total assets are the denominator because, unlike equity, they cannot be shrunk by share buybacks.
• Return on assets, persistence (up to 3 points): share of years with at least 10 %. From 40 % one point, from 60 % two, from 80 % three. If the last two years were both below, at most one point: an advantage that is over is history.
• Pricing power (up to 2 points): gross margin. One point if its median is at least 40 %, one if it fell by no more than 5 percentage points from the first to the last third of the years.
• Stability in a downturn (up to 2 points): operating margin in the downturn years 2020 and 2022. One point each if it held at least two thirds of its usual level (median of the years).
• Classification: “broadly evidenced” from 80 % of possible points and only if all four parts could be computed; “present” from 50 %, “weak” from 20 %, below that “no trace”.

• No forecasts: neither price targets nor estimated future earnings.
• No assumptions about cost of capital or discount rates, so no model whose result hinges on an assumption nobody can check.
• No language model in the result. The same figures always give the same grade.
• No personal recommendation. The pages know neither your situation nor your goals.

Method version 2.7.0 applies. Every company page states the version it was computed with. A rule change gets a new number and is listed here.