datarekha
Case & Behavioral Hard Asked at DuolingoAsked at SpotifyAsked at SnapAsked at Airbnb

30-day retention dropped from 42 % to 31 % over the last two months. How do you diagnose the root cause?

The short answer

A retention drop investigation requires distinguishing between an acquisition-mix shift (newer cohorts are lower quality) and a genuine product regression (existing cohorts are performing worse). The two look identical in aggregate retention but have completely different fixes. Cohort analysis — plotting the D30 survival curve for each weekly acquisition cohort — is the first move.

How to think about it

Step 1 — Separate acquisition-mix shifts from product regressions

Plot D30 retention for each weekly cohort (users acquired in week W, retained to day 30). Two patterns:

  • Pattern A — New cohorts are worse: cohorts acquired in the last 8 weeks show lower D30 than older cohorts at the same age. Cause: a change in acquisition channel (e.g., a new paid UA campaign attracting lower-intent users). Fix: audit the acquisition funnel, not the product.
  • Pattern B — All cohorts degraded simultaneously: both old and new cohorts show D30 drops starting from the same calendar date. Cause: a product change or external event. Fix: look at product releases deployed on or near that date.

Step 2 — Segment the drop

If Pattern B, segment retained vs churned users by:

  • Platform (iOS / Android / web)
  • Geography
  • User type (new / returning / paid)
  • Core feature usage (did they use feature X before churning?)

The segment with the steepest drop narrows the hypothesis.

Step 3 — Funnel decomposition

Break retention into its components: login rate, core-action completion rate, notification opt-in rate. The first metric in the funnel that drops reveals the breakpoint.

Worked example. D30 drops from 42 % to 31 %. Cohort plot shows Pattern B starting the week of April 7. Platform segment: Android retention down 23 pp, iOS flat. Feature segment: users who did not complete the new onboarding redesign (shipped April 5 on Android) churn at 2.4x the rate. Root cause: the redesigned Android onboarding broke a key activation step. Fix: roll back the Android onboarding; reactivation push to the affected cohort.

Step 4 — Quantify impact before acting

Compute the expected lift from fixing the root cause: if the broken cohort (800K users) recovers from 19 % to 42 % D30, that is 184K additional retained users at the 30-day mark — worth quantifying for prioritisation against other roadmap items.

Keep practising

All Case & Behavioral questions

Explore further

Skip to content