Cohort analysis is a method of behavioral analytics that divides users into groups (cohorts) based on shared characteristics — typically the date they first used a product — and tracks each group's metrics over time. By comparing cohorts rather than looking at aggregate numbers, it reveals patterns that would otherwise be invisible.
**Why Cohort Analysis Matters**:
Aggregate metrics lie. Consider a product with stable MAU of 100,000. This looks healthy, but cohort analysis might reveal that each monthly cohort loses 80% of users within 60 days — the stable MAU is only maintained because aggressive acquisition keeps replacing churned users. Without cohort analysis, this 'leaky bucket' would be invisible until acquisition slows.
**Types of Cohorts**:
**Acquisition cohorts** (most common): Group users by when they first joined. 'January 2026 cohort' = all users who signed up in January 2026. Track each cohort's retention, engagement, and revenue over subsequent weeks/months.
**Behavioral cohorts**: Group users by actions they took. 'Users who completed onboarding in first session' vs. 'Users who skipped onboarding.' Compare retention between these groups to identify which behaviors predict success.
**Demographic/segment cohorts**: Group users by characteristics — plan type (free vs. paid), source (organic vs. paid), geography, device type. Reveals which segments have the best unit economics.
**Reading a Cohort Retention Table**:
| Cohort | Month 0 | Month 1 | Month 2 | Month 3 |
|--------|---------|---------|---------|----------|
| Jan | 100% | 40% | 30% | 25% |
| Feb | 100% | 45% | 35% | 28% |
| Mar | 100% | 50% | 38% | 32% |
This table shows improving retention across cohorts — March users retained better than January users at every stage. This could indicate product improvements, better onboarding, or improved targeting.
**Key Patterns to Look For**:
1. **Retention curve shape**: Does the curve flatten (good — a stable base of retained users) or continue declining (bad — even long-term users eventually leave)?
2. **Cohort-over-cohort improvement**: Are newer cohorts retaining better than older ones? This indicates product improvements are working
3. **Drop-off points**: Where does the steepest decline occur? This identifies critical moments in the user journey
4. **Resurrection patterns**: Do users in older cohorts sometimes reactivate? This might justify re-engagement campaigns
5. **Segment differences**: Which cohorts (by source, plan, or behavior) have the best retention? Double down on those
**Applications Beyond Product Metrics**:
- **Marketing**: Compare customer lifetime value across acquisition channels
- **Sales**: Track revenue retention by contract vintage
- **Education**: Compare student outcomes by enrollment period or program
- **Healthcare**: Track patient outcomes by treatment start date
- **Content**: Analyze reader engagement by subscription date
**Best Practices**:
- **Choose the right time granularity**: Weekly cohorts for fast-moving products, monthly for slower ones
- **Define metrics clearly**: Standardize what 'retained' means across all cohorts
- **Allow sufficient maturation**: Don't draw conclusions from immature cohorts
- **Segment deeply**: The most valuable insights come from comparing sub-cohorts
- **Combine with qualitative data**: Numbers show what happened; user research explains why
**Limitations**:
- **Sample size**: Recent or highly segmented cohorts may be too small for reliable analysis
- **Survivorship bias**: Analyzing only retained users ignores the experience of churned users
- **External factors**: Seasonal effects, market changes, and competitor actions can affect cohorts independently of product quality
- **Complexity**: Maintaining and analyzing many cohorts across many metrics can become analytically overwhelming