Cooking-the-books count
What it measures¶
James Montier, a strategist at an investment firm, published in 2008 a simple count of six warning flags — his C-score, named after "cooking the books". Each flag compares this year with last year and is either raised or not:
- the gap between net income and operating cash flow grew;
- days sales outstanding rose — customers took longer to pay;
- days of inventory rose — goods sat in stock longer;
- other current assets grew against revenue;
- depreciation fell against property, plant and equipment (here the net figure, since the gross one is not stored);
- total assets grew by more than 10% — often a sign of growth bought through acquisitions.
The check asks: are two flags or fewer raised?
Why it is on the page¶
Each flag is a way a company can make one year's profit look better: booking sales early, building stock, stretching the life of equipment, buying growth. Montier found that companies raising many of them did worse for their owners afterwards. The flags are simple enough to verify by hand from the statements.
Thresholds, their source and the population they were fitted on¶
| Result | When |
|---|---|
| ✓ passed | 2 or fewer of the 6 flags are raised |
| ✕ failed | 3 or more are raised |
| ○ not evaluable | so many flags lack their inputs that the missing ones could change the result |
A flag whose inputs are missing is counted neither way. The count is decided only when the missing flags could not change the outcome: 0 raised with 1 unknown still passes; 2 raised with 1 unknown does not decide.
- Source: Montier (2008), "Cooking the books, or, more sailing under the black flag".
- Fitted on: US and European companies, 1993–2007.
- Evidence grade: practitioner convention — published by a practitioner, not peer-reviewed or independently replicated.
- Basis: absolute.
A worked example¶
Honeywell International, fiscal year 2025 (statements stored on 25 October 2026): 0 of 6 flags raised, one of them unknown — passed, since the unknown flag could not make it three.
Apple, the same year: 3 of 6 flags raised — failed.
What “No verdict” means here¶
The check is not evaluable when the flags that could not be judged — for missing inventory, receivables or property lines — are enough to change the outcome. The missing input is named, for example inventory FY2025, and the check leaves the area's count.
What the verdict is and is not¶
A count of three is a reason to read the statements more closely, not a finding of anything improper: a fast-growing or acquiring company can raise several flags honestly. It is not a forecast and not an instruction to trade the shares.