In 1968, a finance professor named Edward Altman took a stack of bankruptcy filings and asked whether the companies' own balance sheets had seen it coming. They had. Five ratios, weighted and summed, separated the failures from the survivors years in advance more reliably than any single number could. That sum is the Altman Z-Score, and half a century later it's still the fastest read on one question: how financially fragile is this company right now?
It is also one of the most misquoted numbers in investing — because there isn't one Z-Score. There are three, and using the wrong one can brand a perfectly healthy company a bankruptcy risk.
What the score measures
Every variant of the formula reads the same story off the balance sheet, scaled by total assets so companies of different sizes are comparable:
- Working capital — can the company cover its near-term bills without scrambling?
- Retained earnings — has it accumulated profits over its life, or accumulated losses? This is the term that quietly penalizes young, unprofitable companies.
- Operating profit (EBIT) — does the business itself earn anything, before financing and taxes?
- Equity vs. total liabilities — how much cushion do owners have before creditors are impaired?
- Revenue vs. assets — in the manufacturer variants only: how hard do the assets work?
Weak working capital, a history of losses, thin operating profit, and a debt-heavy structure — each drags the score down. The zones then turn the number into a verdict: safe, grey, or distress.
Three formulas, not one — and why it matters
Altman's original 1968 model was calibrated on public manufacturers and uses market capitalization in the equity term. In 1983 he published two revisions that matter far more in practice. Z' replaces market cap with book equity, so it can be computed from filings alone. Z'' goes further and drops the revenue-to-assets term entirely — because for software, services, and pharma companies, asset turnover is noise, not signal. A software company's assets are mostly goodwill from acquisitions; dividing revenue by goodwill tells you about its M&A history, not its solvency.
Each formula has its own zone lines. For Z' (manufacturers): above 2.9 is safe, below 1.23 is distress. For Z'' (everyone else): above 2.6 is safe, below 1.1 is distress. A bare "Z = 1.8" with no formula named is not information.
Here's why we care so much about this distinction. When we audited our own forensic pipeline, the number that set off the investigation was Salesforce — a large, profitable, cash-generative software company — scoring 1.18: "distress" under the manufacturer formula. Nothing was wrong with Salesforce. The formula was wrong for Salesforce, on two counts: the revenue-to-assets term punished a goodwill-heavy balance sheet, and its working capital looked artificially terrible because of a quirk we'll come to next. Scored with Z'', the same filings tell a very different story. Any screener that runs one formula across the whole market is manufacturing distress signals out of sector differences.
The deferred-revenue trap
One more distortion hides inside "working capital," and it bites exactly the companies investors screen most: subscription businesses. When customers prepay for a year of software, accounting books that cash as a current liability called deferred revenue — the company "owes" a year of service. Economically, it's the opposite of distress: customers handed over cash in advance. For a large SaaS company this can be enormous — for Salesforce, deferred revenue is on the order of 15% of total assets — enough to flip measured working capital deeply negative and drag the whole score into the distress zone on its own.
The honest fix is to treat prepaid subscriptions as what they are — revenue waiting to be recognized, not a bill waiting to be paid — before computing the working-capital term. That single adjustment moves cash-rich subscription businesses out of fake distress while leaving genuinely strained balance sheets exactly where the model put them.
Who the score should never be run on
Banks, insurers, REITs, and most financial companies. A bank's "working capital" and "asset turnover" are category errors — its balance sheet is made of loans, deposits, and reserves, and the model was never calibrated on anything like it. A healthy bank can land in the distress zone simply for being a bank. Roughly a quarter of the companies we cover fall in this bucket, and the only honest score for them is no score at all, clearly labeled — never a number that invites a false conclusion.
How to read "distress" without panicking
Here is the fact that recalibrates most people's reading of this model: across the thousands of companies we score, nearly half of all scored companies land in the distress zone. That is not a market on the edge of collapse — it's the model doing exactly what it was built to do, applied to a market full of pre-profit biotechs and early-stage growth companies. Negative retained earnings plus negative operating profit puts a company in the distress zone almost by definition. For a clinical-stage biotech, the score is stating the obvious: this company lives on its cash runway, not its earnings.
So read the zone as a description of balance-sheet posture, then ask the question the score cannot answer: is that posture funded? A distress-zone reading sends you to three places in the 10-K — the cash and investments line against the annual burn rate, the debt maturity schedule in the notes, and the going-concern language in the audit report. A distress-zone company with four years of cash and no debt due is a stage, not an emergency. A distress-zone company with nine months of cash and a maturity wall next year is the thing Altman built the model to catch.
The score earns its keep in two quieter ways, too: trajectory — a company sliding from safe through grey toward the line, year after year, is a slow-motion warning that headline earnings often hide — and pairing — a weak Z-Score next to an elevated Beneish M-Score is one of the most dangerous combinations in forensic finance, because companies under solvency pressure have the strongest motive to flatter their earnings.
How AnalystBook computes it
We compute a Z-Score for every company we cover where the model applies, straight from the machine-readable numbers in their SEC filings. The engine picks the formula by sector — manufacturers get Z', everyone else gets Z'', and banks, insurers, and REITs get an honest not applicable instead of a misleading number. It uses the book-equity variants deliberately, so a score never depends on a stock quote; it makes the deferred-revenue adjustment described above; and when a filing doesn't tag operating income, it reconstructs EBIT from pretax income plus interest rather than silently skipping the company. Every score is shown with its formula and zone, and links back to the exact financial lines it came from — so when a company reads "distress," you can see in one click whether you're looking at a maturity wall or just a biotech being a biotech.
That's the honest version of the Z-Score: not a bankruptcy prophecy, but the fastest way to sort a watchlist by balance-sheet fragility — provided the right formula did the sorting. For research purposes only; not investment advice.