A sharp dip in activation after a redesign is often a missing event, not a missing feature. Before rewriting the onboarding flow, verify the activation definition still fires on the new screens.
Compare like cohorts: same channel, same first action, same app version. Seasonal patterns in Korea and campaign-heavy weeks can bend curves without signalling product failure.
When the drop is real, ask what changed in the first session. Time-to-value and second-day return usually move before day-30 retention does.
Keep a short anomaly checklist beside your weekly report. The checklist saves more roadmap time than another chart.