The Altman Z-score: what it says about insolvency risk, and what it does not
The formula is over fifty years old and it works surprisingly well. How it is computed from the trial balance, which thresholds are used, and why it must not be read as a verdict.
In 1968, Edward Altman took 66 companies - half bankrupt, half healthy - and looked for the combination of ratios that separated them best. What came out was a five-term formula that predicted bankruptcy two years ahead in roughly 80% of cases. Five decades later, the result still holds.
The five components
For unlisted companies, the variant in current use is:
Z' = 0.717·X1 + 0.847·X2 + 3.107·X3 + 0.420·X4 + 0.998·X5
| X1 | Working capital / Total assets. How much of the asset base is financed long term. It measures the buffer. |
| X2 | Retained earnings / Total assets. How much value the company has accumulated over time, not only this year. |
| X3 | Operating profit / Total assets. The profitability of the core business, before interest and tax. It carries the largest weight, and rightly so. |
| X4 | Equity / Total liabilities. How far asset values can fall before liabilities exceed assets. |
| X5 | Turnover / Total assets. How intensively the assets are used. |
How to read the result
| Below 1.23 | High-risk zone. A significant probability of difficulties within the next two years. |
| 1.23 - 2.90 | Grey zone. The model does not pronounce; what matters is the direction, not the level. |
| Above 2.90 | Safe zone. |
In the report we use a visual band with the company's position on it, precisely because a single number creates a false impression of precision. The difference between 1.20 and 1.30 is not the difference between bankruptcy and safety.
Where the model behaves badly
Very young companies. X2 - retained earnings - is zero or negative in the first years, which drags the score down no matter how well the business is doing. A two-year-old company growing healthily can land in the red zone without having any problem at all.
Asset-light services. X5 and X3 both have "total assets" as denominator. A consultancy with three laptops and 2 million in turnover obtains enormous X5 values, and the score becomes unnaturally good.
Real estate and holdings. Large assets and low turnover produce exactly the opposite effect.
Unadjusted accounting. The model reads what the books say. Overstated inventory or unadjusted receivables raise the score precisely in the companies where they ought to lower it.
How we use it anyway, to good effect
Not as a verdict, but as a signal and a trend. A Z that falls three periods in a row says something real even if it stays in the green zone. A Z below the threshold but rising says something different from one below the threshold and falling.
That is why the report shows the components separately, not only the sum: when the score moves, you see at once which of the five terms moved it. A Z falling because of X3 is a profitability problem; one falling because of X4 is a leverage problem. They are fixed differently.
See this on your own trial balance
You upload the PDF of the SAGA trial balance and receive, within a few minutes, all the indicators above calculated and interpreted, plus the bank scoring and an action plan. The preview is free and does not even ask for your email address.
Read more
Marginal cost and marginal revenue: what each extra leu of sales brings
We added marginal cost and marginal revenue analysis to the report: how much profit each extra leu brings, operating leverage, the safety margin to break-even and sales scenarios.
The report, richer: portfolio, debts, dividends and how much a bank could lend you
Five new chapters in the report: unpaid customers and suppliers, debts to banks and leasing, dividends with up-to-date tax, how much credit you could support and a year-by-year chart. Plus the report in your language.
Your figures, set beside the sector's: the partnership with ica.ro
An indicator means nothing until you compare it. From now on, the report places your company's figures next to the median of the CAEN sector and next to its own history, using public data brought in through our partner ica.ro.