Cohort analysis

Cohort analysis groups users by their signup period and tracks behavior over time. Reveals trends that aggregate metrics hide.

What it shows

• Is retention improving cohort-over-cohort?
• When does decay stabilize?
• Are newer cohorts retaining differently?

Standard view

Y-axis: cohort (signup month). X-axis: months since signup. Cell value: % retained.

Healthy curves

Sharp initial drop (users who never activate), then flatten into a retention plateau.

Tools

• Mixpanel, Amplitude cohort reports
• Google Analytics 4 (basic)
• Custom SQL for tailored views

What to do with this

• Track retention by cohort, not aggregate MAU, aggregate numbers hide the decline that should alarm you
• Compare retention curves across 6-12 cohorts, if newer cohorts trend worse, your product or audience has drifted
• Look for the flattening point in each cohort, if a curve never flattens, that cohort has a PMF problem
• Benchmark against your category, 30% month-3 retention is weak for B2B SaaS but great for consumer social
• Review cohort analysis weekly, when a cohort's curve bends the wrong way, the root cause is usually still fixable