The Beneish M-Score, applied to a real 10-K

Most forensic-accounting frameworks ask you to memorize a list of red flags. The Beneish M-Score does something more useful: it gives you a single number, derived from eight ratios, that you can compute from any 10-K in about ten minutes. The model was trained on a sample of SEC AAER targets between 1982 and 1992, and in the original paper it correctly classified 76% of manipulators with a 17.5% false-positive rate.

That accuracy isn’t the point. The point is that the model forces you to look at the eight ratios it cares about, and it turns out those eight ratios are exactly the ones that matter regardless of whether you bother computing the composite score.

The first ratio: DSRI (Days Sales in Receivables Index)

If receivables are growing faster than sales, the company is either booking revenue earlier in the cycle or getting paid later — both bad. NKLA’s DSRI for FY25 is 1.92, meaning days-sales-outstanding rose from 47 to 88 year-over-year.

The second ratio: GMI (Gross Margin Index)

Eroding gross margins, paired with growing sales, are a textbook cocktail for revenue-recognition aggressiveness. NKLA’s GMI is 1.41 — meaning prior-year gross margin (already negative) deteriorated 41% further.

The third ratio: AQI (Asset Quality Index)

Tracks the proportion of “soft” assets — non-current assets other than property, plant, and equipment — relative to the total. Rising AQI suggests the balance sheet is accumulating capitalized expenses, intangibles, and other assets whose realizable value is harder to verify than that of physical equipment.

The fourth ratio: SGI (Sales Growth Index)

High sales growth, by itself, isn’t manipulation. But the Beneish model treats it as an aggravating factor: the empirical record shows that companies under pressure to sustain a growth rate are more likely to lean on aggressive accruals when the rate becomes hard to deliver.

The remaining four ratios

DEPI (Depreciation Index), SGAI (SG&A Index), LVGI (Leverage Index), and TATA (Total Accruals to Total Assets) round out the eight. The most diagnostically powerful of these is TATA — high accruals as a fraction of assets is the cleanest single indicator of earnings being booked but not converted to cash.

Reading the composite

The composite M-Score is a weighted combination of the eight ratios. Beneish’s threshold is -1.78: scores below that suggest a low probability of aggressive earnings management; scores above suggest a higher probability. We use a slightly stricter band internally — below -2.22 reads as comfortably clean, between -2.22 and -1.78 as adequate, between -1.78 and 0 as elevated, and above 0 as alarming.

What to take away

The M-Score is not a fraud detector. It is an accruals-quality scan, and accruals-quality is the single most predictive forensic signal in the financial-statement literature. A clean M-Score on a struggling company tells you the struggle is being reported honestly. An alarming M-Score on a growing company tells you to look at where the growth is being booked — and whether the cash is following.