Cohort
A cohort is a group of users sharing a common starting trait, most often an install date, traffic source, or geo, tracked together over the same stretch of their lifecycle for comparison. It's how retention and LTV get measured honestly, instead of lumping users of very different ages inside the product together.
How cohort analysis works
Instead of one blended retention number for the month, users get split into groups by install day, and each group is tracked separately for how many stay active by day 1, 3, 7, and 30. That makes it possible to compare fresh users against older ones over the same time window, instead of mixing different lifecycle stages together.
Cohorts in PWA arbitrage
Comparing cohorts day by day reveals whether a specific source is degrading over time — if last week's cohorts show noticeably lower retention than earlier ones, that's a signal to check the source before overall ROI actually drops. APEX push sends are worth evaluating by cohort too: comparing retention between a cohort that received a send and one that didn't gives an honest read on whether a specific message sequence is actually working.
FAQ
- How is cohort analysis different from regular period-based stats?
- Regular period stats blend users of very different ages — yesterday's installs and last month's — which hides the real trend. Cohort analysis compares users at the same point in their lifecycle, giving a more honest read.
- How often should cohorts be checked in PWA arbitrage?
- At minimum weekly for active sources, to catch a gradual decline before it shows up in overall campaign ROI. New sources or geos are worth checking more often, right from the first few days.
Cloaking, anti-bot, push, split tests and your own domains — in one service.
Get started free