Cohort analysis is the gold standard for measuring retention because it controls for time. Group customers by their signup month (or week), then measure what percentage are still active at Month 1, Month 2, etc. This reveals whether your product is getting better or worse at retaining people over time. Compare cohort curves — if newer cohorts retain better, your product improvements are working. If they’re flattening at a lower point, you have an engagement ceiling problem.

An engagement score compresses multiple usage signals into a single health metric. Common approach: identify 3–5 “core actions” that correlate with retention (e.g., creating a project, inviting a teammate, running a report). Score each action 0–100 based on frequency relative to your power users, then compute a weighted average. Weight actions by their correlation with 90-day retention, not by gut feel. Recalibrate quarterly as your product evolves.

The “aha moment” concept: users who complete certain actions early are dramatically more likely to stay. Facebook’s was “7 friends in 10 days,” Slack’s was “2,000 messages sent.” Find yours by running a correlation analysis between early actions (first 7/14/30 days) and 90-day retention. Look for actions with both high correlation and reasonable completion rates. Build your onboarding around driving users to these milestones.

Don’t just measure how much churn — understand why and when. Map your churn timing curve: is attrition front-loaded (onboarding problem) or does it spike after month 6 (value realization problem)? Categorize churn reasons: price, competitor, missing feature, poor support, no longer needed. Cross-reference with segments: are enterprise customers churning for different reasons than SMBs? This diagnostic view turns churn from a single scary number into an actionable improvement roadmap.