Value investing method
Value investing strategy: refuse or rank
A value investing strategy needs a rule for figures that do not fit. A bank's gross margin is refused, its return on assets is ranked only inside financials.
By Javier SanzPublished Updated : The three refusal rules this page stated are stated once now, on /methodology, and the page keeps the table they produce and the two places that table looks unfinished.
Founder of ValueScreener. Built Ninety Nine, a securities services provider, and sold it to a tier-one global bank; ran operations and internationalization at Alpaca, the brokerage API.
Who writes thisEditorial processLinkedIn (opens in a new tab)X (opens in a new tab)
A bank's gross margin is refused outright, its return on assets is published and ranked only against other financials, and its leverage is published with a note attached. Those are the three mechanisms this product has for a figure that does not fit the company, read out of the rules it runs on 2026-09-07.
Why there are three answers rather than one, and which kind of wrongness each answers, is stated on the methodology page. What is on this page is the table that rule produces, which nothing else publishes: which figure gets which treatment for which kind of company, read out of the rules this product ran on one dated build.
| Figure | Bank | Insurer | Equity REIT | Specialty finance | Utility | Operating company |
|---|---|---|---|---|---|---|
| Gross margin | refused | refused | refused | plain | plain | plain |
| Current ratio | refused | refused | refused | plain | plain | plain |
| Debt to equity | refused | ranked, noted | ranked, noted | plain | ranked, noted | plain |
| Interest coverage | refused | plain | plain | plain | ranked | plain |
| Return on assets | ranked, noted | ranked, noted | ranked | plain | plain | plain |
| Return on equity | ranked, noted | ranked | plain | plain | ranked | plain |
| Return on invested capital | refused | refused | plain | plain | ranked | plain |
| Altman Z-Score | refused | refused | refused | refused | plain | plain |
| Piotroski F-Score | refused | refused | refused | plain | plain | plain |
| Beneish M-Score | refused | refused | refused | plain | plain | plain |
| Price to earnings | plain | plain | refused | plain | plain | plain |
| Free cash flow yield | refused | refused | plain | plain | ranked, noted | plain |
"Ranked" means the number is shown and its peer percentile is computed against that company class rather than against the overall market. "Noted" means a sentence is attached explaining why it reads the way it does. "Plain" means the figure is treated like any other company's.
Reading down a column
A bank loses ten of the twelve figures above to a refusal or a re-ranking, and keeps two plain. Its price to earnings is an ordinary price to earnings: a bank has earnings and a share price and dividing one by the other means what it usually means. That is the point of doing this per figure rather than per company. A single "financial sector" flag would have suppressed the P/E too, and there is no reason to.
An insurer keeps its debt to equity and its interest coverage while a bank loses both. That is deliberate and it is the sharpest example in the table of a rule that had to be thought about rather than pattern-matched. A bank's deposits are its raw material and counting them as borrowing is a category error. An insurer's float is a liability to policyholders, also not borrowing, but the debt an insurer issues is ordinary debt and the ratio over it means what it says.
A specialty finance company loses exactly one figure, the Altman Z-Score, because the model excluded financial firms. Everything else about it is real: it files ordinary commercial statements, has a real gross margin, a real current ratio and real free cash flow. Payment networks and ratings businesses sit in this class.
A utility loses nothing to a refusal. Five of its twelve figures are re-ranked against other utilities, because rate regulation sets the return, the rate base sets the leverage and the capital programme sets free cash flow, and ranking any of those against the whole market measures the regulatory compact rather than the management.
The two places the matrix looks unfinished
The Altman Z-Score row. It is refused for four company classes and plain for utilities, alongside operating companies. The utilities cohort's 75th percentile Z-Score on the 2026-09-07 build is 1.33, and Altman's distress floor is 1.81, so at least three quarters of that sector reads as distressed by a model calibrated on manufacturers. Every other structurally-affected figure for a utility is re-ranked. This one is not.
The return on assets and return on equity rows for a utility. Return on equity is ranked against utilities and return on assets is not. Both are returns on a rate base whose size a regulator sets, so it is hard to see the argument for treating them differently.
Neither of those is a bug in the sense of a broken computation. They are two places where the rule table has a gap, and a page about how a product handles figures that do not fit is the right place to say so.
The composite refuses whole or not at all
The Value Score combines a price block, a quality block built on return on invested capital and gross margin, and the forensic trio. Read the matrix and you can see what happens to a bank: the trio is refused three times over, return on invested capital is refused, gross margin is refused. Five of the inputs are gone.
So the composite is not published for banks, insurers, brokers or mortgage trusts at all, and a sentence appears in its place. On 2026-09-07 a value score was held for 3,853 of the 5,257 listings in the pool that day.
Why a score is withheld whole rather than recomputed over the inputs that survived is the same page's to state, and it does.
The fourth mechanism, one level up
The same discipline applies to the cohorts themselves, and the floor it sets is on the methodology page with the rest. What it costs is visible in the output: on the 2026-09-07 build, the Altman Z-Score refuses two whole sectors on that rule, Financial Services and Real Estate, because after the class exclusions too few companies remain in either. Both sectors publish percentiles for other figures on the same build. The gap between one metric's sector list and another's is the class-exclusion table made visible in the output.
Why any of this is worth the trouble
Three refusal counts from the same build, for scale: 865 companies are excluded from the Altman cohort, 617 from gross margin, 324 from debt to equity, and none at all from return on equity or return on assets. Those counts and the sentence each refusal prints are set out company by company on the page linked at the top of this one.
The reason to build it this way rather than to suppress a sector wholesale is that a wholesale rule is wrong in both directions at once. It hides figures that are perfectly good, which makes the product less useful, and it leaves untouched the figures that are quietly wrong for a reason the sector label does not capture, which makes it less trustworthy. The utility Altman row above is exactly the second failure, in a product that took the trouble to build the machinery to avoid it.
What this leaves out
The matrix is read out of the rule tables this product runs and describes behaviour as of 2026-09-07. It covers twelve figures and six company classes. The product computes 118 scored indicators and recognises nine company classes, so this is a window on the rules rather than the whole of them.
Company class is determined from the industry string, which is a data vendor's judgement where the vendor supplied one and, since 2026-09-12, the SEC's own words for the filed SIC code where it did not. A company on the boundary between two classes may be classified in a way its own filings would not obviously support, and the consequence is a figure suppressed or shown that a careful reader would decide the other way.
The three refusal counts are from one distribution build and each is taken over the population holding that figure at that build. They are not counts of the market and they are not comparable with each other.
Nothing here is a view on any company, any sector or any company class. It is a description of when a number is printed, when it is printed against a narrower comparison, and when a sentence is printed instead.