Cohort Analysis

Cohort analysis groups users by something they share at the start — signup week, acquisition channel, first action — and tracks how each group behaves over time. A single aggregated number blends all cohorts together; cohorts pull the real movement back into view.

The most useful version is the retention cohort chart. Stack signup weeks down the rows and 1-, 2-, 4-, 8-week retention across the columns, and you can immediately see whether new cohorts are retaining better after a product change. A single average can't show that.

Cohort size matters. If a weekly cohort has fewer than ~50 users, noise dominates and you should usually fall back to monthly cohorts. In low-volume B2B contexts, cohort analysis can hit a hard floor where qualitative review becomes more useful than the chart.