What Is a Churn Cohort? (Plus How to Use One)

A churn cohort is a group of customers tracked over time based on a shared starting point, usually their signup month. Cohort analysis tracks what percentage of each cohort remains active in each subsequent month.

It is the single most useful unit of analysis for retention work because it reveals patterns that blended averages hide.

Why cohorts beat blended averages

A blended monthly churn rate of 5% sounds clean. But it might mean any of these things:

• 12% churn in customers under 90 days, 4% in 3-12 months, 1% in 12+ months (early-stage problem)
• 2% in 0-12 months, 8% in 12+ months (a renewal problem)
• 5% across all tenures, no improvement (steady decline)

Each scenario needs completely different interventions. The blended number can't tell you which one is happening. Cohort analysis can.

How a cohort retention chart works

You build it like this:

1. Pick a cohort definition (usually signup month: "January 2026 cohort")
2. Track what % of the cohort is still active in each subsequent month
3. Plot the percentages on a chart over 12-24 months
4. Repeat for multiple cohorts to compare trends

Example: 1,000 customers signed up in January. By March (month 2), 850 are still active. That's 85% month-2 retention for the January cohort.

For the full setup, see the cohort analysis guide.

What good vs bad cohort shapes look like

Four common shapes:

• Sharp drop, then flat: Healthy. The initial drop is expected (some early churn is normal). The flat tail means you retain the engaged users indefinitely.
• Gentle slope downward: OK but improvable. You're slowly losing customers over time. Engagement and resurrection work can flatten the curve.
• Steady slope toward zero: Bad. You'll lose 100% of users eventually. Indicates a value-decline problem.
• Smile curve (drops, flattens, then rises): Excellent. Expansion revenue and engagement increases are outpacing churn. Network-effect businesses (Slack, Airbnb) show this shape.

What to look for in your cohorts

Three diagnostic questions:

1. Are newer cohorts retaining better or worse than older ones? Better = your retention work is paying off. Worse = something has gotten harder (acquisition, product fit, competitive landscape).
2. Where does most of the drop happen? First 30 days = activation problem. Months 3-6 = engagement problem. Year 1+ = competitive or value-decline problem.
3. Is there an outlier cohort? A specific month with much better or worse retention than its neighbors. Usually correlates with a product change, marketing campaign, or external event you can identify.

Tools that build cohort charts automatically

• Revenue cohorts: ChartMogul, Baremetrics (Stripe-native)
• Behavioral cohorts: Amplitude, Mixpanel, PostHog
• Customer success cohorts: Vitally, Gainsight, ChurnZero
• DIY: Spreadsheet with COUNTIF formulas, or SQL in your data warehouse

For deeper retention diagnostic work, also see the retention curve simulator and how to calculate churn rate correctly.

Curious how your retention setup scores?

Take the 60-second Churn Health Check. It scores your measurement maturity (including whether you use cohorts) and tells you what to fix next.