What is the Altman Z-Score?

The Altman Z-Score is a formula that combines five financial ratios into a single number to estimate how likely a company is to go bankrupt within the next two years. A high score signals financial health; a low score signals distress. It's one of the oldest and most widely used bankruptcy-prediction tools in finance, and credit analysts, lenders, investors, and auditors still reach for it because it's transparent and easy to read.

NYU Stern professor Edward Altman developed it in 1968. He used a statistical technique called multiple discriminant analysis to find the combination of ratios that best separated companies that later went bankrupt from those that survived, then weighted each ratio by how strongly it predicted failure. The result is one score that rolls up liquidity, profitability, leverage, and efficiency into a single health check.

Below is the formula, what each ratio measures, how to read the score, the variants for private and non-manufacturing companies, and what the model can and can't do.

The original formula

The classic Z-Score, built for publicly traded manufacturing companies, is:

Z = 1.2·X₁ + 1.4·X₂ + 3.3·X₃ + 0.6·X₄ + 1.0·X₅

Those weights aren't arbitrary — they came out of Altman's discriminant analysis as the values that best distinguished failing companies from healthy ones. The five ratios:

How to read the score

For the original model, the score falls into three zones:

One thing matters more than the single reading: the trend. Companies rarely jump straight from safe to bankrupt. A Z-Score that declines over several quarters is often a clearer warning than any one quarter's number, so analysts watch the direction of travel, not just the level.

A slide toward the distress zone rarely stays contained to the score. As failure risk rises, so does the company's cost of capital — lenders and investors demand more to bear the added risk, which raises the very discount rate used to value the business.

A quick example

Take a manufacturer with working capital / total assets of 0.20, retained earnings / total assets of 0.20, EBIT / total assets of 0.10, market value of equity / total liabilities of 0.89, and sales / total assets of 1.0:

Z = (1.2 × 0.20) + (1.4 × 0.20) + (3.3 × 0.10) + (0.6 × 0.89) + (1.0 × 1.0) = 0.24 + 0.28 + 0.33 + 0.53 + 1.0 ≈ 2.38

A 2.38 lands in the grey zone: not in crisis, but not clearly safe either. The next move is to look at which ratios are dragging — here, the modest EBIT margin — and check whether the score is rising or falling year over year.

The variants: private and non-manufacturing companies

The original model only fits public manufacturers, so Altman built two adaptations. Using the wrong version for the company type is a common mistake.

Z'-Score (private companies). Private firms have no stock price, so Altman replaced X₄'s market value with book value of equity and recalibrated every coefficient:

Z' = 0.717·X₁ + 0.847·X₂ + 3.107·X₃ + 0.420·X₄ + 0.998·X₅

The zones shift: distress below 1.23, grey 1.23–2.90, safe above 2.90. This version tends to produce more conservative scores.

Z''-Score (non-manufacturing). Service firms, tech, and other asset-light businesses don't fit the sales-to-assets logic, so this version drops X₅ entirely and reweights the rest:

Z'' = 3.25 + 6.56·X₁ + 3.26·X₂ + 6.72·X₃ + 1.05·X₄

Removing the sales-to-assets ratio reduces distortion where asset intensity varies widely between firms, and the +3.25 constant — the detail most summaries drop — rescales the result back onto the standard scale. Because the coefficients and constant differ from the original, so do the zones:

Read each with its own cutoffs — a 2.5 is "grey" on the original scale but "safe" on Z''. The Z''-Score works for both private and public non-manufacturers.

Accuracy, and what the score gets wrong

In Altman's original 1968 sample the model was strikingly accurate close to failure — about 95% correct one year before bankruptcy and roughly 72% two years out — and later out-of-sample tests (Altman's own 2000 revisit of the Z-Score and ZETA models, and later academic reviews covering hundreds of distressed companies) have generally landed in the 82–94% range one to two years ahead. That's good enough to screen, not to underwrite: the score sorts a population into "look harder" and "probably fine," and its power fades the further out you forecast.

Two error types matter, and they cost differently. A Type I error — a company the model calls safe that then fails — is the expensive one for a lender or auditor: the missed default. A Type II error — a healthy company flagged as distressed — mostly costs time. Both rise as the horizon lengthens and as you move away from the population the model was fit on.

But it has real limits, and a careful analyst treats it as a screen, not a verdict:

• The zones aren't absolute. Cutoffs vary by industry and economic conditions, and distress thresholds have arguably loosened over time.
• Model fit matters. The original is calibrated to manufacturers; applying it to a bank, insurer, or early-stage tech company is misleading. Match the variant to the company.
• It's a snapshot from the financials. Accounting choices, one-time items, and off-balance-sheet exposures distort the inputs. Garbage in, garbage out.
• Trends beat points. A single score means less than its direction over time.

A worked example: Rivian in the distress zone

The Z-Score earns its keep on capital-intensive manufacturers with thin profits — exactly where a pre-profitability automaker lives. Rivian's Q1 2026 inputs produce a Z of −1.61, deep below the 1.81 distress threshold, with operating losses (factor C) doing most of the damage.

The score discriminates the peer cohort precisely — legacy automakers sit in safe or grey, pure-play EV startups all sit in distress — but it doesn't say which distressed firms can survive on cash and equity raises and which can't.

And direction matters as much as level: a deteriorating score is more predictive than the level itself. Rivian's has climbed steadily off its 2023 trough — the trend a single reading would miss entirely.

See the full interactive example: Altman Z-Score — Rivian Q1 2026 ↗

A cautionary case: Enron, and the number nobody read

The most-taught example of a Z-Score that was screaming while the market wasn't listening is Enron. Applying the non-manufacturer Z'' model to Enron's reported financials, academic reconstructions find the score sliding toward — and into — the distress zone in the late 1990s, years before the December 2001 bankruptcy. One widely cited analysis noted that Enron's Z-Score implied a credit profile around a 'B' rating — barely above junk — while the rating agencies still carried it at investment-grade 'BBB' until four days before it filed.

The lesson isn't that the Z-Score is magic — Enron's statements were partly fraudulent, and no ratio model fully sees through cooked books. It's that even the numbers Enron did report were deteriorating in plain sight, and a five-minute screen would have flagged "look harder" long before the collapse. The score didn't fail; it was simply never run — or never believed — against a name everyone assumed was safe.

Two caveats keep it honest: Enron was a trading-and-services firm, so the manufacturer Z is the wrong variant — the analyses use Z''; and because the financials were later restated, treat the specific figures as an illustration of the method, not audited history. The reusable habit is the point — run the screen on the names everyone is sure about, watch the trend, and take a falling score seriously even when the rating agency hasn't moved.

Running it at scale

Calculating one Z-Score by hand is a five-minute exercise. The work appears when you have to do it across a portfolio of borrowers, a watchlist, or a data room full of targets — each with financials in a different format, each needing the right variant applied, the inputs tied back to source statements, and the trend tracked over several periods.

If you need Z-Scores across a set of companies, you can run them with the Altman Z-Score agent using MacrosLM. Drop in the financial statements, and it pulls the inputs for each company, applies the correct variant (original, private Z', or non-manufacturing Z'') based on the company type, computes the score and zone, tracks the trend across periods, and ties every ratio back to the line item it came from through an evidence layer. Click any score and see the statements behind it. Scoring a whole portfolio becomes a reviewed list instead of a spreadsheet you build company by company.

The judgment stays with you: whether the right variant was applied, whether an input was distorted by a one-time item, what a grey-zone score actually means for that borrower. The agent does the calculation and the tie-out so the analyst spends time interpreting the result, not assembling it.

Bottom line

The Altman Z-Score turns five financial ratios into one number that flags bankruptcy risk, sorting companies into safe, grey, and distress zones. The original is built for public manufacturers; the Z' and Z'' variants adapt it for private and non-manufacturing firms. It's durable because it's simple and transparent, but it's a screening signal, not a final answer: match the right model to the company, watch the trend, and treat a low score as a reason to dig deeper rather than a conclusion.

---

This article is for general information and is not investment, credit, or financial advice. The Altman Z-Score is a screening model with known limitations and should be one input among many in any credit or investment decision.