SaaS Churn Analysis Framework
You cannot fix churn until you understand why users leave. This framework classifies churn types, walks through cohort analysis, and shows how to build a simple prediction model.
Most SaaS teams track churn rate but do not diagnose its causes. This guide gives you the tools to move from "our churn is X%" to "here is why users leave and here is what we are doing about it."
Three Types of Churn
Voluntary churn
Customer actively cancels their subscription.
How to diagnose: Exit surveys, cancellation flow interviews, support ticket analysis
Common causes: Unmet expectations, poor onboarding, found a better alternative, budget cuts
Involuntary churn
Subscription lapses due to payment failure.
How to diagnose: Failed payment logs, dunning email open rates, card expiry reports
Common causes: Expired credit cards, insufficient funds, incorrect billing details
Quiet churn
Customer stops using the product but does not cancel (zombie accounts).
How to diagnose: Usage decline alerts, login frequency drop-off, feature engagement decay
Common causes: Product no longer needed, champion left the company, found a workaround
Cohort Analysis in Five Steps
Segment by signup month
Group customers by the month they signed up. This reveals whether churn is improving or worsening over time.
Plot retention curves
For each cohort, plot the percentage still active at month 1, 3, 6, and 12. Look for drop-off patterns.
Compare by acquisition channel
Break cohorts by how they found you (organic, paid, referral). Some channels attract higher-churn users.
Overlay feature usage
Correlate feature adoption with retention. Identify which features retained users adopt that churned users do not.
Calculate revenue impact
Multiply churn rate by average contract value per cohort. This turns a percentage into a dollar figure stakeholders care about.
FAQ
What is a good churn rate for SaaS?
Monthly churn under 2% (annual under 20%) is typical for SMB SaaS. Enterprise SaaS with annual contracts often achieves under 5% annual churn. Compare against your own trend, not just benchmarks.
How do I build a churn prediction model?
Start simple: identify 3-5 usage signals that differ between retained and churned users (login frequency, feature adoption, support tickets). Score each account weekly. A logistic regression on these signals often outperforms complex ML models for early-stage SaaS.
Should I try to win back churned customers?
Yes, selectively. Customers who churned due to timing or budget are good win-back candidates. Customers who churned due to product fit issues are not. Run a win-back email at 30 and 90 days post-cancellation.
Key Takeaways
- Classify churn as voluntary, involuntary, or quiet — each requires different intervention.
- Cohort analysis reveals whether churn is a systemic problem or specific to certain acquisition channels.
- Correlate feature usage with retention to identify the features that keep users active.
- Calculate churn in revenue terms to get stakeholder attention and justify retention investment.
Related Reading
Related from other topics
Need Help Reducing Churn?
Heck Design Group helps SaaS teams diagnose churn and build retention systems that work.