Most teams do not fail because they lack metrics. They fail because they optimize the wrong one for too long.
Start from the value moment, not the dashboard
Ask one concrete question: what behavior proves the product delivered value?
- Collaboration tool: teams completing shared work.
- Marketplace: successful matches that close.
- Learning app: lessons completed and revisited.
If the behavior does not correlate with retention, it is not your north star.
Stress-test candidates with three checks
- Customer value: users feel progress when this rises.
- Business link: revenue or retention moves in the same direction.
- Team agency: product, design, and engineering can influence it weekly.
If one check fails, treat that metric as supporting KPI, not north star.
Keep a one-page metric brief
Document:
- Metric definition in one sentence.
- Included and excluded events.
- Segment that owns the target.
- Guardrail metrics.
- Review cadence and owner.
This prevents metric drift where each team uses a different interpretation.
Review cadence that avoids thrash
Review weekly for trend, monthly for diagnosis, quarterly for recalibration. Do not replace your north star after one bad month. Change it only when your product model changes.
