How do you define a reliable business metric?

Direct answer

What a strong answer covers

How it works

A metric is a contract. It names the business question it answers, the entity being counted, the event that qualifies, the numerator and denominator with their filters, the time zone and calendar used for periods, how attribution works when several events could claim credit, and what is excluded such as test accounts or refunds. It has an owner who approves changes, a versioned definition in code or a semantic layer so every dashboard computes the same number, and a validation against a trusted source such as finance data. Without that, two teams present two revenue figures in the same meeting.

Example

Monthly active users was defined as distinct user IDs with at least one session event of 10 seconds or more, in the account’s local calendar month, excluding internal and test accounts, computed in dbt as a single model that all dashboards read. The first reconciliation against the billing system revealed a 6 percent gap from bots, which the definition then excluded explicitly.

Common mistakes

• Letting each dashboard implement its own version of the same metric.
• Changing a definition without versioning and breaking every historical comparison.
• Defining a rate without stating who is in the denominator.

A useful answer structure

Answer from your own work.

Do not memorise a script: use only examples you can explain honestly in follow-up questions.