Cohort Retention Analysis
A method of analyzing retention by grouping users based on when they signed up (or another shared characteristic) and tracking their behavior over time.
What Is Cohort Retention Analysis?
Cohort retention analysis groups users by a shared starting point — typically signup week or month — and tracks what percentage of each cohort remains active over subsequent periods. The output is a triangular table (cohort down, time-since-signup across) that reveals whether your product is getting better, worse, or staying the same over time.
Also Known As
- Analytics teams: cohort analysis, triangle table
- Growth teams: retention matrix
- Product teams: cohort tracking
- Finance teams: customer vintage analysis
How It Works
A team pulls signups by month for Jan through Jun. For each month's cohort, they compute the percentage still active at Month 1, Month 2, Month 3, etc. The table shows Jan's cohort at 31% Month 3 retention, Feb at 33%, Mar at 29% (they shipped a bad release), Apr at 38% (they fixed it and added better onboarding), May at 41%, Jun at 43%. The upward trend across cohorts at the same tenure point tells them the product is genuinely improving — not just churning different people at different rates.
Best Practices
- Do align cohorts to meaningful periods (weekly for fast-moving consumer apps, monthly for SaaS).
- Do segment cohorts by source — organic and paid often retain very differently.
- Do read the table diagonally to spot external events (platform changes, marketing spikes) that affected multiple cohorts simultaneously.
- Don't compare cohorts with very different sample sizes without flagging the variance.
- Don't look at one cohort in isolation. The signal is in the trend across cohorts.
Common Mistakes
- Using calendar-based active users instead of cohort-based retention. You can have "growing MAU" while every cohort retains worse.
- Ignoring the seasonality of when a cohort entered. Q4 signups often behave differently than Q1 signups.
Industry Context
Cohort analysis is standard practice in SaaS and consumer apps. In ecommerce, cohorts are often defined by first purchase date and tracked by repeat purchase rate. B2B uses cohort analysis for renewals (what % of year-1 customers renew into year 2, year 3). It's less useful in transactional businesses where users have no ongoing relationship.
The Behavioral Science Connection
Cohorts isolate generational effects — users acquired under different conditions (product state, marketing messages, competitive landscape) behave differently. This mirrors construal level theory: users' mental model of your product is anchored to what it looked like when they joined, which affects their later behavior.
Key Takeaway
Cohort analysis is the only way to answer "is the product getting better?" Aggregate metrics lie because they blend many cohorts together. If you can only build one analytics view, build the cohort retention triangle.