Churn analysis: churn rate, causes and a worked calculation
Churn analysis examines how many customers leave, when and why. The article describes the correct way to calculate the churn rate, how to annualise it, the Kaplan-Meier survival curve, the split between voluntary and involuntary churn, and common mistakes.
Amounts and values in the examples are illustrative and show how the calculation works. In projects, we use the client's data.
Dominik Dębski · Lemur Media · · reading time: 4 min
Reviewed by Dominik Dębski, 1 November 2026
What is churn analysis?
Churn analysis is the measurement and explanation of customer departures (churn): what share of its base a company loses in a period, at what point in the relationship customers leave and why. The churn rate answers the question "how many", the survival curve answers "when", and the cause analysis answers "why". We cover revenue retention and cohort tables in separate articles; here we focus on the number of customers and the reasons they leave.
How do you calculate the churn rate correctly?
The churn rate is calculated as the number of customers at the start of the period who left during that period, divided by the number of customers at the start of the period. Customers acquired during the period are included in neither the numerator nor the denominator.
| Monthly data | Value |
|---|---|
| Customers at the start | 1 000 |
| New customers in the month | 200 |
| Total departures | 50, including 10 customers acquired in the same month |
| Correct churn | 40 ÷ 1 000 = 4,0% |
| Mistake 1: all departures divided by the opening base | 50 ÷ 1 000 = 5,0% |
| Mistake 2: all departures divided by the closing base including new customers | 50 ÷ 1 200 = 4,2% |
Annualising requires compounding, because each month churn applies to a smaller base: 1 − (1 − 0.04)12 = 1 − 0.6127 = 38.7%. Multiplying 4% × 12 gives an incorrect 48%.
How do you show when customers leave?
The timing of departures is shown by a survival curve, calculated with the Kaplan-Meier estimator. The method was described by Edward L. Kaplan and Paul Meier in the Journal of the American Statistical Association (1958). Its advantage is that it correctly accounts for censored observations: customers who are still active but have been with the company too briefly to be observed for longer.
| Month | Customers at risk | Departures | Censored after the month | Survival |
|---|---|---|---|---|
| 1 | 100 | 5 | 5 | 1 × (1 − 5/100) = 0,950 |
| 2 | 90 | 9 | 11 | 0,950 × (1 − 9/90) = 0,855 |
| 3 | 70 | 7 | 0,855 × (1 − 7/70) = 0,770 |
After three months, about 77% of customers remain active. Simply dividing the remaining customers by 100 would understate the result, because it would treat censored observations as departures.
How do you analyse the reasons for churn?
Reasons are first split into voluntary and involuntary churn, because they require different actions. Involuntary churn occurs when a subscription ends because of a failed payment, e.g. an expired card, even though the customer wants to keep using the service. It is reduced with payment retries and payment reminders. Voluntary churn requires changes to the product, service or price.
- Value: the customer did not achieve the outcome for which they bought the product.
- Price: cost out of proportion to usage.
- Change on the customer side: acquisition, liquidation, change of decision-maker.
- Competition: switching to another provider.
- Service: unresolved tickets, quality problems.
Which signals precede churn?
A customer's departure is usually preceded by changes in product usage and in the relationship with the company, visible in the data weeks or months in advance. The list of signals is built by comparing customers who left with customers who stayed, over the same period before departure.
| Signal | Example metric | Action |
|---|---|---|
| Drop in activity | Number of active users at the customer 30% lower than three months earlier | Account manager contact and usage review |
| Service problems | Tickets open for more than 14 days | Escalation to the customer service manager |
| Change of decision-maker | New customer-side contact in the CRM | Introductory meeting and a reminder of results achieved |
| Payments | Failed card charge or late payment | Payment retry and reminder |
| No first value | Customer has not achieved an outcome within 60 days of starting | Onboarding plan with deadlines |
The thresholds in the table are examples. Each signal is assessed by how well it has distinguished, in the past, customers who left from those who stayed.
What are the common mistakes?
- Multiplying monthly churn by 12 instead of compounding.
- Including new customers in the denominator.
- Combining voluntary and involuntary churn in one metric without a split.
- Recording churn reasons as free text without a fixed list of categories.
- A single metric for the whole base without a split by segment and cohort.