Earnings Quality and Red Flags

Framing note, and it governs the whole article. This is a list of questions, not a list of accusations. Every indicator described below has ordinary, innocent explanations, and in the great majority of cases the innocent explanation is the correct one. The article does not teach fraud detection and should not be read as doing so — it teaches a reader which questions the preceding nine articles equip them to ask, and, just as importantly, why a checklist of this kind produces far more false alarms than findings.

One's figure is repeatable, backed by cash, and built on modest estimates. The other's depends on a favourable estimate, a one-off gain, and revenue that has not yet been collected. Earnings quality is the difference — and it is a question about the character of a number rather than its size. A company with high-quality earnings may be a poor business, and a company with mediocre earnings quality may be an excellent one. Keeping those two judgements separate is most of the skill.

What quality actually means: three dimensions

Persistence. Will this recur? Earnings from selling products to repeat customers are persistent; earnings including an asset-sale gain, a legal settlement, or a one-off tax benefit are not. The question is never "how much did they earn" but "how much of it will happen again."

Cash conversion. Does the profit become money? The first article in this pillar established the four reasons profit and cash diverge and the ratio that tracks it. Earnings that persistently fail to convert are the most-watched quality indicator — and, as that article showed, the divergence has innocent causes at least as often as concerning ones.

Estimation intensity. How much of the figure rests on judgement? A company whose earnings depend heavily on impairment assumptions, percentage-of-completion estimates, or long-dated provisions has earnings that are more opinion than a company whose revenue is collected at the point of sale. This is not a criticism — some industries cannot be accounted for any other way — but it changes how much weight a single year's figure deserves.

The questions this pillar equips you to ask

Eight questions, each drawn from an earlier article, and each with its innocent explanation attached — because the innocent explanation is usually the right one.

1. Is cash conversion deteriorating over several years? Innocent explanation: growth consumes working capital, and a fast-growing company funding larger receivables and inventory will show exactly this pattern while being entirely healthy.

2. Are receivables growing faster than revenue? Innocent explanation: a shift toward larger customers with longer standard terms, or a change in geographic mix. Worth asking about; rarely alarming on its own.

3. Do the "one-time" items in adjusted earnings appear every year? An item excluded annually is not one-time by any ordinary meaning of the word. Innocent explanation: genuinely recurring restructuring at a company in a long transition — though this is the question where the innocent explanation is weakest, and the stock-compensation article showed a single adjustment moving reported profit by 18% after tax.

4. Has an estimate or policy changed, and which way did it move profit? Innocent explanation: new information genuinely arrived, and standards and business models change. Both depreciation policy and revenue recognition are supposed to be revised when the facts do.

5. Is the company capitalising more of what it spends? Innocent explanation: a genuine shift toward assets with longer lives, or development work that now meets the capitalisation criteria under the applicable framework.

6. How much of the balance sheet is goodwill, and what does the impairment test assume? Innocent explanation: the company grew by acquisition, which is a strategy rather than a fault — as that article set out.

7. What did the auditor identify as the hardest judgement? Not a red flag at all — key audit matters are a map of where estimation concentrated, and reading them is the highest-value use of an audit report.

8. Has the company restated, and was it earnings or classification? Innocent explanation: most restatements are technical, and that article showed two identically sized corrections meaning completely different things.

Why a checklist like this misleads more often than it helps

This is the most important section in the article, and it is an argument against using the preceding one carelessly. Serious accounting problems are rare. Apply a set of indicators to a large number of companies and the arithmetic of base rates takes over.

Suppose — purely for illustration — 5% of companies have a genuine problem, and a checklist correctly flags 80% of those while wrongly flagging 20% of sound companies. Out of 1,000 companies: 40 real problems are caught, 10 are missed — and 190 sound companies are flagged. Of 230 total flags, only 17% are genuine, so 83% of the alarms are wrong. If the true prevalence is 2% rather than 5%, precision falls to 8%; if it is 10%, precision reaches only 31%. The checklist is not badly designed — this is what happens to any test for a rare condition, and it is the same arithmetic Pillar 22 applied to backtests: search enough candidates and the extremes appear regardless.

Three consequences. A flag is a prompt to read, not a conclusion. The indicator's job is to direct attention to a disclosure, and the disclosure usually resolves it. Multiple independent flags mean more than one, because the false-positive rate compounds far less quickly than intuition suggests. And the strongest signals are patterns over time rather than levels in a year — a single year's ratio is noise, while a four-year deterioration with no explanatory disclosure is a question with substance.