The Altman Z-Score
A single number, built in 1968, that tries to answer a blunt question: is this company heading for bankruptcy?It's genuinely useful — and it's also the most famous example in all of finance of a formula being confidently, spectacularly wrong about the wrong kind of company. Learn both halves and you'll never misread it again.
01Where it comes from
In 1968, a young NYU finance professor named Edward Altman did something Graham never had the computing power to do: he took dozens of companies that had gone bankrupt, dozens that hadn't, and used statistics to find the combination of financial ratios that best separated the two groups ahead of time. The result was the Z-Score — the first widely adopted quantitative bankruptcy-prediction model.
It worked. In his original sample the model flagged roughly 72% of bankruptcies two years before they happened, and banks, auditors, and credit analysts have leaned on it ever since. Where Graham's number asks “is this cheap?”, Altman's asks the prior, more urgent question: “will this company still be here?”
02The formula
Five ratios, each weighted by how strongly it predicted survival in Altman's data. The weights aren't arbitrary — they're what the statistics chose.
Add them up and the score lands in one of three zones:
Notice what four of the five ratios divide by: total assets. Remember that — it's the trapdoor.
03On the company it was built for
Altman built the Z-Score on manufacturers, so let's point it at one — Linamar, the auto-parts maker from the last module, computed from its latest filings:
Z = 2.90
A sensible, believable read. Linamar sits just under the 2.99 “safe” line — the grey zone — which for a cyclical, capital-heavy parts maker is exactly the honest answer: solid, not bulletproof, worth watching. Every ratio is doing real work, and no single term dominates. This is the Z-Score operating inside its design envelope.
Interesting footnote: Linamar looked cheap on the Graham Number (−36%) but only grey on Altman. Two lenses, two shades of the same picture — which is the whole reason this app never leans on one number.
04When the Altman Z-Score lies
Now watch it break. Here is the exact same formula on Royal Bank of Canada — one of the largest, most profitable, most stable banks on the planet:
Z = 0.27 → deep in the “distress” zone
A score of 0.27 would, taken literally, mean RBC is on the brink. It obviously isn't. So what happened? Every single term collapsed— and the reason is the trapdoor from earlier: four of the five ratios divide by total assets, and a bank's assets are a $2.4-trillion loan book.
Against that gigantic asset base, working capital, retained earnings, EBIT, and sales all look microscopic — not because the bank is weak, but because a bank's balance sheet is nothing like a factory's.Its “assets” are loans it made, its “liabilities” are your deposits. The model's core assumption — that assets are plant and equipment used to generate sales — simply doesn't hold. Altman knew this: he later published separate variants (the Z′ and Z″ models) for non-manufacturers, and stated plainly the original was never meant for financial institutions.
Where a low score means nothing
- Banks & insurers. Enormous asset bases and deposit/policy liabilities guarantee a low, meaningless score — as with RBC above.
- REITs.Property-heavy balance sheets with high leverage-by-design read as “distress” even for healthy, well-covered landlords.
- Asset-light & early-stage. Little in the way of retained earnings or hard assets skews X₂ and X₃, so young or capital-light firms score low regardless of prospects.
This is why the research page shows a company-type banner: the moment you open a bank, a REIT, or an unprofitable firm, it flags that the Graham checklist and the Altman Z-Score are the wrong lenses — before a misleading “distress” reading can scare you off a perfectly sound business. The score isn't broken; it's being asked the wrong question.
Watch the Z-Score — and the warning that catches it
Open Royal Bank and you'll see the low Altman Z and the company-type banner that tells you to ignore it. Then try a manufacturer and watch the same number suddenly make sense.