What is Cohort Analysis?
Cohort analysis is an analytical technique that groups users who share a common characteristic (typically install date) and tracks their behavior over a defined time period. Instead of looking at all users as one mass, cohort analysis lets you understand how specific groups of users behave, enabling more accurate LTV predictions and campaign optimization.
💡 Example
Users who installed your app during a January promotion form one cohort. You track their Day 1, Day 7, and Day 30 retention separately from February users. This reveals whether the January campaign attracted high-quality users or just volume.
Retention Cohort Example
| Cohort | Users | Day 0 | Day 1 | Day 7 | Day 30 |
|---|---|---|---|---|---|
| Jan Week 1 | 10,000 | 100% | 42% | 18% | 8% |
| Jan Week 2 | 12,500 | 100% | 38% | 15% | 6% |
| Jan Week 3 | 8,000 | 100% | 45% | 22% | 11% |
| Jan Week 4 | 15,000 | 100% | 35% | 12% | 4% |
Jan Week 3 shows best retention despite lower volume - investigate what drove those installs.
Types of Cohorts
Acquisition Cohorts: Users grouped by when they installed (most common). Helps compare campaign effectiveness over time.
Behavioral Cohorts: Users grouped by actions taken (completed tutorial, made purchase, reached level 10).
Channel Cohorts: Users grouped by acquisition source (Facebook, Google, organic) to compare channel quality.
Geographic Cohorts: Users grouped by country/region to understand market differences.
Use Cases for Cohort Analysis
📊 LTV Prediction
Project lifetime value based on early cohort behavior patterns. Day 7 revenue often predicts Day 90+ revenue.
💰 ROAS Measurement
Track cohorted ROAS over time to understand true campaign profitability beyond install day.
🎯 Channel Comparison
Compare user quality across acquisition channels by analyzing cohort retention and monetization.
🔄 Product Changes
Measure impact of app updates by comparing pre and post-update cohort performance.
Best Practices
Choose Meaningful Timeframes: D1, D7, D14, D30 are standard retention checkpoints. For monetization, extend to D60, D90, D180.
Ensure Statistical Significance: Cohorts need sufficient size (1,000+ users typically) for reliable analysis.
Control Variables: When comparing cohorts, try to isolate one variable (channel, creative, time) for clear insights.
Act on Insights: Use cohort data to adjust UA spend toward higher-quality sources and away from poor performers.