How do I measure retention properly?

Measure retention by first defining what it means for your context-specify whether you're tracking Day 7, Day 14, monthly, or another timeframe, and whether you're using bounded or unbounded retention. Then segment users into cohorts of retained vs. non-retained and examine both their behavioral patterns (what actions they take) and persona characteristics (demographics, acquisition channel, job title).

When running experiments to improve retention, don't rely on Day 7 or Day 30 metrics-those are lagging indicators affected by too many variables. Instead, pick experiment metrics that respond fast to user interaction, move only because of the experiment you're running, and are expressed as percentages rather than absolute counts. The goal is clear, immediate signals about what actually changed, not a delayed picture of outcomes.

Finally, ensure you have enough users in both cohorts before analyzing-without sufficient data, your insights won't be meaningful.