SubmetryBlog

Subscription churn rate: cancellation is not expiration

Calculate subscriber churn with a consistent cohort, distinguish cancellation from lost access, and validate your RevenueCat event data.

A spike in subscription cancellations can be an early warning, but it does not automatically mean the same number of paying customers have already left. To calculate a useful subscription churn rate, define who is being counted, when access ends and which period you are measuring.

This guide uses a simple customer-level gross churn definition. It is a method you can document and reproduce, not a claim that every analytics dashboard uses the same formula.

1. Define the starting population

For a calendar-month view, take the customers with paid access at the start of the month. Count each customer once, even if they have multiple active products. Exclude free trials and keep test or sandbox data separate.

Then count members of that starting group who lose all paid access during the month. Under this gross-loss definition, count the first qualifying loss once and report later reactivations separately.

Subscriber churn rate = starting customers who lose paid access ÷ customers with paid access at the start × 100.

Do not add new customers acquired during the month to that denominator. Their early retention belongs in a separate acquisition cohort. A metric based on active customers at month-end answers a different question.

2. Separate cancellation, expiration and billing trouble

RevenueCat distinguishes a cancellation notification from expiration; an unsubscribe can leave access active until the paid period ends. Check its event definitions when mapping subscription states.

Maintain separate operational views for cancellation intent, ended paid access and payment recovery. Treat a failed payment as a signal to inspect the subscription state rather than automatically counting a lost customer. Also check for another active product before recording customer-level churn.

Imagine a starting group of 1,000 paying customers. During the month, 80 turn off renewal, but only 30 members of the starting group lose all paid access. Customer churn under the definition above is 30 ÷ 1,000 × 100 = 3%. Reporting 8% from cancellation notifications would describe a different event. These numbers are illustrative, not a benchmark.

3. Keep subscriber churn and revenue churn separate

Losing ten low-priced subscriptions and ten high-priced subscriptions can produce the same subscriber count but very different revenue loss. Report subscriber churn alongside an explicitly defined revenue metric rather than swapping one for the other.

For a monthly-normalized revenue view, use the recurring revenue associated with the starting customer base and state how discounts, annual plans, refunds and currency are handled. Do not divide raw annual receipts by a monthly customer count or treat gross receipts as net proceeds after store fees.

4. Validate identity and event coverage

An event feed can be connected while a retention report is still incomplete. Check these items before interpreting the rate:

  • The app events and subscription events use a shared customer identity.
  • Event retries do not create duplicate losses.
  • The selected environment and time zone are consistent.
  • You have enough history to establish the starting cohort and observe its outcomes.
  • Late events, reactivations and refunds follow a documented rule.

Follow Submetry's RevenueCat integration guide for identity and webhook setup, and its event guide for client instrumentation. Importing apps does not automatically backfill historical purchases. A successful test notification proves connectivity, not a complete revenue history.

Submetry's documented renewal/churn percentages use matured observed paid periods and a seven-day outcome window. That differs from the calendar-month customer formula above. Check the metric definition and coverage before comparing the two.

5. Investigate one segment before changing the product

Break the same metric down by plan, platform, acquisition cohort or app version. Keep the observation window consistent and show the number of customers behind each percentage. A movement from one loss to two in a tiny segment is not the same evidence as a sustained shift across a large cohort.

Choose an action that matches the signal: review onboarding for early departures, inspect product value before renewal, or investigate payment recovery for billing-related losses. Change one thing, record when it changed and compare equivalent cohorts. Explore Submetry to connect product behavior with subscription events, while keeping the definition of every rate visible.