Cohort Analysis
A technique for grouping users by shared characteristics or signup timeframes to compare behavior and retention patterns across different user segments.
Definition
Cohort analysis groups users by shared characteristics or timeframes to compare behavior patterns. By tracking how different cohorts perform over time, teams can measure retention, identify trends, and evaluate the impact of product changes.
Common Cohort Types
Acquisition Cohorts
Users grouped by when they signed up (January cohort, Q1 cohort). Most common for retention analysis.
Behavioral Cohorts
Users grouped by actions they took (completed onboarding, used feature X). Good for understanding feature impact.
Property Cohorts
Users grouped by attributes (plan type, company size, source). Good for segment comparisons.
Reading a Retention Cohort Table
| Cohort | Week 0 | Week 1 | Week 2 | Week 3 | Week 4 |
|---|---|---|---|---|---|
| Jan | 100% | 45% | 32% | 28% | 25% |
| Feb | 100% | 48% | 35% | 30% | 27% |
| Mar | 100% | 52% | 40% | 35% | - |
This shows February users retained better than January users - suggesting a product improvement.
What Cohort Analysis Reveals
- Retention trends - Are newer users retaining better?
- Feature impact - Do users who try feature X retain longer?
- Seasonality - Do certain signup periods perform differently?
- Product-market fit - Is retention improving over time?
Tools for Cohort Analysis
Analytics platforms with cohort features:
- Amplitude - Behavioral cohorts with lifecycle analysis
- Mixpanel - Retention cohorts with breakdown
- PostHog - Open-source cohort tracking
Frequently Asked Questions
What’s a good retention rate for SaaS?
Week 1 retention of 40-60% is typical for B2B SaaS. Month 3 retention of 30-40% is healthy. But benchmarks vary significantly by product type and user segment.
How do I improve cohort retention?
Identify what high-retaining users do differently. Build features or onboarding that guides more users to those behaviors. Compare cohorts before and after changes to measure impact.
How many users do I need for cohort analysis?
You need enough users per cohort for statistical significance - typically 100+ per cohort for meaningful comparisons. Smaller cohorts can show trends but be careful drawing conclusions.