Three teams report "active users" and get three numbers. None of them is wrong; there was never one definition.
Three people count the crowd at a party. One counts everyone who came, one counts who's still there, one counts who bought a drink. None is wrong; nobody agreed the question.
Undefined metrics turn every meeting into a reconciliation exercise instead of a decision.
A metric is not a number, it is a written contract: the event that counts, the population it applies to, the time window, the deduplication rule, and the exclusions. Without that written down, every team reimplements it slightly differently and the organisation spends its meetings reconciling numbers instead of acting on them. The discipline is boring and it is the highest-leverage thing an analytics function does — one definition, implemented once in the modelling layer, referenced everywhere, with a named owner who can change it.
Define a metric as a contract: the precise event, the population, the window, the dedup rule, and the exclusions — plus an owner. Implement it once in the semantic or modelling layer so every dashboard references the same logic. A good metric also has to be sensitive enough to move within a decision cycle and hard to game.
How to Set a Great North Star Metric — Exponent, 5:26