What is Cohort Analysis?
Cohort analysis groups users by a shared characteristic (usually signup date) and tracks their behavior over time. Instead of looking at all users together, you compare how the January cohort behaves vs the February cohort vs the March cohort.
This reveals whether your product is improving. If each new cohort retains better than the last, you're building a better product.
Why Cohort Analysis Matters
Aggregate metrics hide trends. Your overall retention might look flat, but cohort analysis might reveal that new cohorts retain much better while old cohorts are churning. Or vice versa.
Cohort analysis answers questions like: Are we getting better at retaining customers? Did that product change help? Which acquisition channels bring the best customers?
Running Cohort Analysis
Group users by signup week or month. Track retention, revenue, engagement, or any key metric over time. Visualize as a cohort table or retention curve. Compare cohorts to identify trends and the impact of changes.
How to Run a Cohort Analysis Step by Step
Step 1: Define your cohorts. Group customers by the month they signed up. January signups = January cohort.
Step 2: Choose your metric. Retention rate, revenue, or engagement. For SaaS, revenue retention per cohort is most valuable.
Step 3: Build the cohort table. Track each cohort's metric over time:
| Cohort | Month 0 | Month 1 | Month 2 | Month 3 | Month 6 |
|---|---|---|---|---|---|
| Jan (50 customers) | 100% | 88% | 82% | 78% | 68% |
| Feb (62 customers) | 100% | 90% | 85% | 80% | — |
| Mar (55 customers) | 100% | 92% | 87% | — | — |
Step 4: Read the story. Two things matter: (1) Are newer cohorts retaining better than older ones? Yes. March retains 92% at month 1 vs January's 88%. Something improved (onboarding change? product update?). (2) Where's the biggest drop? Month 0→1 loses 8-12%. That's where to focus.
Step 5: Act on the insights. If month-1 retention jumped after a product change in February, you've validated the change works. If a specific cohort churns faster, investigate what was different about those users.
Common mistakes founders make:
- Using too-large cohort windows (monthly is standard, quarterly hides trends)
- Not normalizing to 100% at month 0 (makes comparison impossible)
- Comparing cohorts of vastly different sizes
- Only doing retention cohorts. Revenue cohorts can show different patterns (customers stay but spend less)
Formula
Cohort Retention = Customers Remaining ÷ Original Cohort Size × 100
Track by time period (month, quarter) and segment (source, plan, size)
Example
Your SaaS company tracks January cohort retention:
- Month 0: 100 customers (100%)
- Month 1: 85 customers (85%)
- Month 3: 70 customers (70%)
- Month 6: 55 customers (55%)
- Month 12: 45 customers (45%)
The January cohort retains 45% after one year. Compare against February and March cohorts to see if retention is improving.