Honesty · 2 min · Updated 27 Sept 2026

Why we print “unknown” instead of a number

A figure we cannot verify is not a zero and not a guess. The rules we follow when data is missing, and what they cost us.

A number on a screen looks like a fact. If it was filled in because the real one was missing, it is worse than no number, because nothing tells you so. So we follow one rule: if the result depends on a figure we do not have, we do not replace it with zero or with an average. We say it cannot be valued, and why.

What happens without that rule

Before it, a company whose debt we did not know was valued as if it owed nothing. An airline with about $4 billion of debt came out at more than three times its share price, “undervalued”. Counting its debt, the same model valued it below the price. The conclusion flipped.

Where you will see “unknown”

No valuation: when the debt, the cash, the company’s own growth, its margin or, for a bank, its book value is missing. The dossier gives the reason.
No verdict: when the value comes out more than 5 times the price. We keep the arithmetic, but we do not call it undervalued.
No Leak Score number: the score is only published when signals worth at least 60 % of the rubric could be checked. Until then the dossier shows how many of the twelve were checked. A signal we cannot verify counts neither for nor against.
No Graham value: the formula is left out when earnings are negative or falling fast.

The rule in reverse

Refusing to value a company for lack of a figure the method does not use is just as false as inventing it. A bank is valued on book value, so a missing cash figure does not stop its valuation. Each method asks only for what it uses.

Two things on every dossier

The date of the price the verdict uses, and the version of the model. When we change the model, every valuation is recomputed that night, and an analysis saved before the change is redone the next time you open it.

Where we are still blind

Coverage is not complete. Foreign companies file their annual reports on Form 20-F or 40-F, often under IFRS and in their own currency; we read those too, translating them to dollars at the US Treasury rates, but many are still being loaded and until then they are valued from market data only. Where the filings are thin, fewer signals can be checked. The disclaimer page lists what the model assumes.