Product
Cohort analysis
A method of segmenting users by the period they joined and tracking a metric (retention, revenue, engagement) over time — so improvements in the product show up as newer cohorts outperforming older ones.
By Daniel Reyes · Last updated July 22, 2026
In plain English
Instead of looking at 'March DAU', look at 'users who signed up in March, and how many are still active each month after.' It's the only honest way to see if the product is getting better.
Example
Jan cohort: 100 signups → 40 active in month 1, 25 in month 3, 20 in month 6. Feb cohort ships an onboarding rebuild → 100 signups → 60 in month 1, 45 in month 3. Feb curve above Jan = product improved.
Why it matters
Aggregate metrics hide product decay: growing signups can mask a shrinking retention curve. Cohorts surface the truth. Every board deck for a subscription business should have a retention-cohort chart.
Common mistakes
- Confusing signup cohorts with activation cohorts — pick one convention and stick with it
- Reading cohorts too early — you need at least 3-6 months of data before drawing conclusions on retention shape
- Comparing cohorts of different sizes without normalisation