Churn is the metric every subscription deck quotes and almost no one benchmarks honestly. Across MWM's catalog, the median app has churned 72.7% of its users by day 1, 90.8% by day 7, and 96.1% by day 30. Even the best-retaining decile still loses 89% by day 30. Churn, like its mirror image retention, is not a number you drive to zero — it is a curve you bend.
But "reduce churn" is two completely different jobs depending on which churn you mean, and conflating them is why most churn programs stall. This guide separates them and gives you the levers for each.
The two churns you must measure separately
- Engagement churn — a user stops opening the app. This is the dominant churn for free, freemium, and ad-supported apps, and it's the inverse of retention: the 96% D30 figure above is engagement churn.
- Subscription (revenue) churn — a paying user cancels or fails to renew. This is the churn that directly destroys revenue, measured as a monthly cancel rate on the paying base, not as D30 retention.
An app can have terrible engagement churn and healthy subscription churn (a small, loyal paying base) or the reverse. Fixing one does nothing for the other. Before anything else, know which one is bleeding.
Voluntary vs involuntary churn
Within subscription churn, split again:
- Voluntary — the user chose to cancel (price, value, or simply done).
- Involuntary — the payment failed: expired card, insufficient funds, bank decline. The subscription lapses even though the user never decided to leave.
Involuntary churn is typically 20 to 40% of total subscription churn and is the most fixable, because the user still wants the product — their card failed, not their intent. Billing retries, grace periods, account-hold states, and dunning (payment-recovery messaging) recover a large share. Most teams pour effort into voluntary-churn save flows while quietly leaking involuntary churn they could recover with a billing-retry configuration change.
Measure churn like an operator
- Engagement churn is the inverse of N-day retention — track it by install cohort, never blended into one site-wide number.
- Subscription churn is the monthly cancel rate of the paying base; separate voluntary from involuntary, and track gross churn against net (which nets out reactivations and plan upgrades).
- Uninstall rate is the hard floor of engagement churn: a user who uninstalled is gone, while one who merely stopped opening can still come back.
Cut engagement churn (the first-week fight)
The steepest engagement churn is D1 to D7 — the catalog runs 72.7% to 90.8%, meaning most of what survives day 1 is gone by day 7. That is an activation-and-habit problem, and the full playbook lives in How to Improve App Retention. The short version: get users to the value fast, build a habit loop in the first week, and run re-engagement before users go dormant, not after they're already gone.
Cut subscription churn (where the revenue is)
This is where churn work pays for itself directly:
- Recover involuntary churn first — the highest-ROI move. Billing-retry logic, grace periods, account-hold, and dunning sequences win back users who never meant to leave. This is configuration and messaging, not product work.
- Build a real cancel flow. At the moment of cancel intent, offer a pause (pause beats cancel), a win-back offer, or a downgrade tier — not a dead-end confirm button.
- Close the value-realization gap. Most voluntary churn is users who never got the value they subscribed for. Tie activation and retention back to trial conversion and the paywall so you convert users who will stay, not users who will cancel in month one.
- Default to annual. Annual plans replace twelve renewal decisions with one and structurally lower churn — see the mobile app monetization playbook.
The category lens
Engagement churn shape varies sharply by category. Median D30 churn across the catalog:
| Category | D1 churn | D7 churn | D30 churn |
|---|---|---|---|
| Social & Communication | 68.1% | 87.7% | 94.1% |
| Lifestyle & Well-being | 76.3% | 90.4% | 95.2% |
| Productivity & Tools | 77.0% | 91.1% | 95.5% |
| Education & Knowledge | 75.1% | 91.4% | 96.4% |
| Media & Entertainment | 75.1% | 92.0% | 96.6% |
| Game | 63.4% | 90.7% | 97.1% |
Games churn the most by day 30 (97.1%) despite the lowest day-1 churn (63.4%) — novelty pulls users back once or twice, then the curve collapses. Social churns least (94.1%) because communication loops manufacture reasons to return. If you're a game, your fight is the D7→D30 decay; if you're social, protect the network loop that's already working for you.
Why churn is the lever that compounds
Every point of churn you cut compounds into LTV: a retained user keeps monetizing, and lower churn raises the lifetime value that funds acquisition. Churn is the leak in the bucket — and at 96% median D30, most buckets are mostly hole. The teams that win don't chase a magic low number. They separate the two churns, recover involuntary churn for free, and bend the engagement curve one cohort at a time.