How to predict company bankruptcy: first match the Altman formula to the company. Use the original Z-Score for public manufacturers and Z″ as a starting screen for non-manufacturing industrial firms. Treat either result as an early warning, then verify liquidity, debt maturities, cash flow, and filing footnotes before making an investment decision.
Suppose you screen a public software company with weak working capital but a high market value. A generic calculator returns 3.541. The non-manufacturer Altman model, using the same financial statements, returns 1.3554.
The company did not change between calculations. The model did. Stop at the first result and you could dismiss a balance-sheet warning before reading the debt footnotes.
Many free calculators display one field labeled “Altman Z-Score” and leave that choice invisible. The shortcut matters because the original formula includes sales divided by total assets, a ratio that changes sharply across industries. A retailer, software company, railroad, and manufacturer can have very different asset turnover even when their underlying financial condition is similar.
Choose the Z-Score Model by Company Type
Match the company to the model before opening a calculator.
| Company | Starting model | Why |
|---|---|---|
| Publicly traded manufacturer | Original Z | Matches the public-manufacturer setting used to develop the original model. |
| Private manufacturer | Z′ | Uses book value of equity and re-estimated coefficients rather than inserting a proxy into the public-company formula. |
| Non-manufacturing industrial company | Z″ | Removes sales divided by total assets to reduce the industry effect from asset turnover. |
| Bank or insurer | Do not use these variants as the primary screen | Financial institutions have sector-specific balance sheets, capital rules, and risk measures. Use a model designed for that sector. |
Scope limit: Z″ is not a universal replacement for every non-manufacturer. The Altman material discussed here concerns industrial corporates. Banks and insurers need sector-specific analysis.
Edward Altman’s later review explains why Z″ drops the sales-to-assets variable: asset turnover is industry-sensitive. The revision also changes every coefficient and uses book value of equity in the leverage ratio. That is a different model, not the original formula with one line deleted.
For an emerging-market corporate, Z″ can be a starting point, but Altman’s broader emerging-market process adds currency, industry, competitive-position, and debt-service judgments. A retail investor should not treat a mechanical score as the entire credit analysis.
What the Altman Z-Score Accuracy Claims Actually Show
The original paper studied 66 publicly traded manufacturing companies: 33 bankrupt firms and 33 non-bankrupt firms. Using one financial statement before distress, the model correctly classified 63 of 66 firms in the same sample used to estimate the coefficients.
That result was strong, but it was also in-sample. Altman explicitly warned that reclassifying the firms used to build the model creates upward bias and that secondary-sample testing is essential.
| Financial statement timing | Bankrupt firms correctly classified | Non-bankrupt firms correctly classified | Overall |
|---|---|---|---|
| One statement before distress | 31 of 33 (94%) | 32 of 33 (97%) | 63 of 66 (95%) |
| Two statements before distress | 23 of 32 (72%) | 31 of 33 (94%) | 54 of 65 (83%) |
Source: Altman (1968), Tables 2 and 3. The one-statement result is an in-sample classification test, not a universal out-of-sample accuracy guarantee.
A later study by John Grice and Robert Ingram tested whether the original model generalized to newer periods and non-manufacturing firms. Its published abstract reports negative answers on both questions. That supports a narrower conclusion: model performance depends on period and industry. It does not support converting one overall accuracy figure into a precise non-manufacturing error rate.
What the score measures: the Z-Score classifies firms from a weighted combination of accounting and market ratios. It does not produce a calibrated probability that a company will file for bankruptcy.
How to Predict Company Bankruptcy With the Correct Z-Score
Start with the latest annual filing, then check whether a newer quarterly filing materially changes liquidity, retained earnings, operating profit, equity, or liabilities. The guide on how to read a 10-K shows where those numbers sit and how to reconcile the statements with the footnotes.
Original Z-Score for public manufacturers
Z = 1.2X1 + 1.4X2 + 3.3X3 + 0.6X4 + 1.0X5
- X1: Working capital / Total assets
- X2: Retained earnings / Total assets
- X3: EBIT / Total assets
- X4: Market value of equity / Total liabilities
- X5: Sales / Total assets
Four inputs come from the financial statements. X4 also requires market value of equity, usually the company’s market capitalization on a clearly stated measurement date. That means the original public-company score cannot be reproduced from the 10-K alone.
Z″ for non-manufacturing industrial companies
Z″ = 6.56X1 + 3.26X2 + 6.72X3 + 1.05X4
- X1: Working capital / Total assets
- X2: Retained earnings / Total assets
- X3: EBIT / Total assets
- X4: Book value of equity / Total liabilities
Z″ removes sales divided by total assets and uses book equity rather than market equity. All four inputs can therefore be tied directly to the company’s filing, provided the analyst states how EBIT and book equity were defined.
Keep the measurement date consistent. Do not combine a year-end balance sheet with a market capitalization from months later and call the result a year-end score. Record the filing period, share-price date, units, and any adjustments.
Worked Example: One Company, Two Formulas
Consider a fictional public software company. The figures below are intentionally simple so the model mechanics remain visible. They are not estimates for a real security.
| Input | Value | Resulting ratio |
|---|---|---|
| Working capital | -$30 | X1 = -0.05 |
| Retained earnings | $30 | X2 = 0.05 |
| EBIT | $42 | X3 = 0.07 |
| Total assets | $600 | Denominator for X1 to X3 |
| Total liabilities | $300 | Denominator for X4 |
| Market value of equity | $900 | Original X4 = 3.00 |
| Book value of equity | $300 | Z″ X4 = 1.00 |
| Sales | $900 | X5 = 1.50 |
The original public-manufacturer formula produces:
Z = 1.2(-0.05) + 1.4(0.05) + 3.3(0.07) + 0.6(3.00) + 1.50 = 3.541
The non-manufacturer formula produces:
Z″ = 6.56(-0.05) + 3.26(0.05) + 6.72(0.07) + 1.05(1.00) = 1.3554
The scores diverge because the formulas answer the question with different variables, equity definitions, and coefficients. The original result is boosted by market equity and sales-to-assets. Z″ excludes sales-to-assets and uses book equity. Since the hypothetical company is a non-manufacturer, Z″ is the model aligned with the intended company type.
This does not prove the company will fail. It tells the analyst that a generic calculator could create a very different classification and that the underlying balance sheet deserves closer inspection.
Use the Score Without Overtrusting It
After calculating the appropriate model, move straight to the mechanisms that can actually force a company into distress.
- Liquidity: Is working capital negative because of a durable business model, or because near-term obligations exceed accessible cash and receivables?
- Debt maturity: How much debt comes due within the next 12 to 36 months, and can the company refinance it under current terms?
- Cash conversion: Does EBIT turn into operating cash flow, or are receivables, inventory, capitalized costs, and stock compensation masking weak economics?
- Interest coverage: Can recurring operating profit cover interest expense without relying on asset sales or new financing?
- Equity quality: Is low or negative book equity caused by accumulated losses, acquisitions, buybacks, or accounting structure? The cause matters.
- Auditor and covenant language: Read going-concern disclosures, covenant waivers, debt footnotes, and subsequent events.
Z-Score should sit beside, not replace, leverage review. The debt-to-equity ratio guide shows why negative book equity can make a familiar leverage ratio undefined or misleading. A company can also look financially stable and still be overpriced; distress and valuation are separate decisions.
A score also cannot see a scheduled event. A single-asset company can screen as solvent and still fail if one regulatory decision goes against it, which is why the published Phase 3 approval rate for the relevant therapeutic group belongs next to the balance sheet rather than after it.
Do not average multiple Z-Score variants. A blended score has no published coefficient set, cutoff calibration, or validation sample. Choose the applicable model, document it, and keep the result separate from any other distress indicators.
Frequently Asked Questions
How to predict company bankruptcy with the Altman Z-Score?
Use the model that matches the company, calculate it from consistently dated inputs, and treat the result as a screening signal. The Z-Score cannot predict bankruptcy with certainty; accuracy varies with the sample, industry, period, and distance from failure.
Which Z-Score should I use for a software or service company?
Z″ is the closer fit for a non-manufacturing industrial company because it removes the industry-sensitive sales-to-assets variable. Banks and insurers remain outside the scope of this guide.
Does a high Z-Score mean a stock is a good investment?
No. A high score can indicate lower measured financial distress while saying little about valuation, competitive durability, governance, dilution, or expected return. Financial stability is only one part of an investment thesis.
Can every input be found in the 10-K?
Z″ uses accounting inputs that can be tied to the filing. The original public-company formula also uses market value of equity, which requires a dated market capitalization rather than a 10-K line item.
Should I use the latest quarter or the latest annual filing?
Start with the annual filing for complete footnotes, then update the inputs when a newer quarter materially changes the balance sheet or operating result. Do not mix dates without disclosing the measurement basis.
The Bottom Line on Predicting Company Bankruptcy
The Altman Z-Score remains useful because it forces several dimensions of financial condition into one repeatable screen. Its weakness is the temptation to treat one formula and one historical accuracy figure as universal.
For investors asking how to predict company bankruptcy, the usable process is straightforward: choose the model before calculating the score. Use the original Z for publicly traded manufacturers, Z′ for private manufacturers, and Z″ for non-manufacturing industrial companies. Exclude banks and insurers from this shortcut. Then read the debt, liquidity, cash-flow, covenant, and auditor disclosures that explain why the score looks the way it does.
Use the score to decide where to look next, not to avoid reading the filing.
Which company would you test first, and which model actually matches its business?
- Altman (1968), Journal of Finance: original 66-firm public-manufacturer sample, formula, classification tables, and explicit discussion of in-sample bias.
- Altman, Revisiting the Z-Score and ZETA Models: later accuracy-by-horizon table, private-company Z′, and non-manufacturer Z″ formula.
- Grice and Ingram (2001), Journal of Business Research: tests of temporal and industry generalizability.
Method: Formula coefficients and historical classification counts were transcribed from the cited papers. The worked example was independently recalculated from the displayed inputs: original Z = 3.541 and Z″ = 1.3554.
Limits: The example is hypothetical. Z-Scores are classification tools, not calibrated bankruptcy probabilities. Accounting definitions, filing dates, industry structure, and economic conditions can change the result.
AI-assisted tools supported source discovery and structural editing. Danny Hwang reviewed the cited papers, formulas, calculations, model-scope statements, and final conclusions.
Financial Disclosure
This article is educational and does not provide personalized investment, accounting, credit, or legal advice. A Z-Score can miss distress and can also flag healthy firms. Verify every input against the company’s filings and use sector-appropriate analysis before making an investment decision.
TheFinSense did not receive compensation from any company or data provider mentioned in this article.
Update history
-
v1.1
2026-07-17
CORRECTIONRemoved the unsupported claim that the original model has a 42.2% non-manufacturing error rate and deleted the hypothetical $162,330 portfolio-loss projection. Rebuilt the article around verified model scope, accuracy by horizon, and a reproducible formula comparison.
-
v1.0
2026-04-11
PUBLISHOriginal publication.
Educational quantitative analysis based on published data. Not investment, tax, or legal advice. Consult a licensed professional before acting on any calculation. About TheFinSense.