19 March 2026

When Finance asks for one number

The request is reasonable in origin. Models need inputs. Boards have limited minutes. The failure is to average ages, channels, and immature weeks into a percentage that cannot be reconstructed.

We suggest a sentence with four parts, spoken or written: the population, the birthday, the age at which you read “alive,” and the sample size of the latest mature cohort. Example: “Of people whose first paid event was in a week that has reached age four, 147 of 788 were alive in that age band.” That is one number (18.7%) with a spine.

Do not offer a trailing twelve-week mean unless every week in the mean is mature and the mix is stable. If mix shifted toward paid, the mean will move even if each channel’s ledger did not.

If the model still wants a single input, agree the age once (week four is common in consumer; week eight in slower B2B) and freeze it in the data room glossary. Changing the age to make the number prettier is the same sin as changing the birthday.

Logo churn, net revenue retention, and seat expansion are neighbouring reports. They are not substitutes for a user-level cohort. Say that once, calmly, and hand over the table. Our desks practise that handover because it is where most packs still collapse.