Journal · 21 July 2025

North star metrics in practice

A north star is a sentence with a number attached, not a poster in the kitchen. In App Analytics it should name a job the product exists to finish, counted in a way a sceptical finance partner in the United Kingdom would recognise. “Engagement” is not a job. “Weekly completed journeys” might be. “Minutes in app” almost never is, unless you sell attention on purpose and are honest about it.

In practice we workshop three candidates, kill two, and write the survivor into the event grammar with a version. If the definition later changes — and it will — the version changes too, so history does not pretend the star never moved.

Guards around the star

A single number invites cheating. Pair it with two guards: quality (did the journey succeed without a refund or a crash) and cost (did we buy the users who completed it). Guards are not extra north stars. They are the reason the star does not drift into notification spam.

Where teams get stuck

Founders want a metric that always rises. Good stars wobble. A museum app’s star may be seasonal. A banking app’s star may be brief. If your culture cannot bear a flat month, you will pick a vanity count and call it strategy. Cohort Signal Lab will not bless that choice; we would rather you sit with a quieter pulse.

For a worked example against your stream, the catalogue or a note to the desk is the next step. Do not rename your Mixpanel board “North Star” until the sentence is written down.