Reading Korean mobile cohorts across OS versions
Device and OS mix in Korea shifts quickly after major releases. Ignoring that skew can make a release look like a retention disaster—or a miracle.
When a new iOS or Android release rolls out in Korea, early adopters are rarely a random sample. They skew toward higher engagement, newer devices, and different carrier mixes. Comparing a post-release cohort to the previous month without stratifying by OS version often misattributes product changes to platform shift.
Our standard readout splits retention by major OS version and by whether the user upgraded within seven days of install. We also flag crash-free session rates in the same cut—because a retention dip that tracks crash spikes is an engineering story, not a feature story.
For Seoul-based consumer apps, we additionally check weekday versus weekend install mixes around holidays. A holiday-heavy cohort can look "worse" on day-3 simply because usage patterns differ, not because onboarding failed.
Document these cuts in the experiment and release checklist so leadership sees stratified charts before declaring a regression.