Cohort analysis
Cohort analysis groups users by something they have in common — most often the day they were acquired — and then tracks how each group behaves over the days and weeks that follow. Instead of one blended average, you get a set of comparable curves. It is the technique that lets you answer “are newer users better or worse than older ones?” — a question a single aggregate number can never answer.
How it is calculated
First you define the cohort key. The classic one is the acquisition cohort: everyone who installed or signed up in the same week or month. But cohorts can be cut by any shared property — acquisition channel, first device, country, or a behavior like “completed onboarding.”
Then you measure a metric for each cohort at equal ages, not calendar dates. Age 0 is each cohort’s start day; age 7 is seven days into that cohort’s life. Laying the cohorts out as rows and ages as columns produces the familiar triangular cohort table, where you read down a column to compare cohorts at the same maturity.
The metric in the cells is usually retention or revenue, which is exactly why cohorting underpins both retention rate and lifetime value.
Why it matters
Averages mix together users at every stage of their lifecycle, so they hide change. If your overall retention looks flat, cohort analysis might reveal that recent cohorts are actually retaining worse and only the legacy base is propping up the average — a problem you would otherwise miss until it was severe. Cohorts also isolate the effect of a change: ship a new onboarding, and only cohorts acquired after the change show the impact, cleanly separated from everyone who came before.
In games and apps
Cohorting is how you judge acquisition quality. Two channels can deliver the same install volume while their cohorts diverge sharply in day-30 retention or LTV — and only the cohort view exposes it. Live-ops teams cohort by join date to see whether a seasonal event created lasting habits or just a one-day spike, and product teams cohort by behavior to test whether reaching a milestone causes better retention.
In Keentics
Keentics builds cohorts on your raw event stream, by acquisition date or any user property, and renders the cohort table for retention or revenue with no sampling. Any funnel drop-off group can become a cohort you follow forward. Cohorts power the retention analysis and LTV analysis features — see the pricing page for plan details.
Related: Active users · ARPPU · ARPU · Attribution