Cohort analysis

What is cohort analysis?

Cohort analysis is a technique that groups people by a shared starting event and then follows each group over time. The most common cohort is the month someone signed up. Everyone who joined in January is one cohort, February another, and you watch how each behaves in the weeks and months after.

The power is in what it reveals that an average hides. Total active users can look flat while a serious problem grows underneath, because new signups mask the old ones leaving. Split by cohort and the truth shows: maybe every group loses 40% of its users by month two, a retention cliff a blended chart would never expose. A mobile app team using cohort analysis can see that users who joined after a redesign stick around longer than earlier cohorts, which says the redesign worked far more convincingly than a topline number. Cohorts also expose whether changes you made are actually moving the needle for new users versus flattering yourself with legacy ones.

Cohort analysis needs patience and clean event data. A cohort only tells its story after enough time passes, and if the underlying events are tracked inconsistently, the groups won't be comparable.

Cohort analysis at Dallonses

We reach for cohort analysis when a client's headline numbers look stable but something feels off. Grouping users by when they started, or by which campaign brought them, turns a flat average into a story about who stays, who leaves, and when. That's usually where the real data driven customer insights are hiding.

The technique is only as good as the events behind it, so our work starts by getting the tracking and the data model right. We've built cohort views for global brands that exposed a retention drop the dashboards had been averaging away for months. Once a team can see how each group behaves over time, decisions about retention and onboarding stop being guesses.

Topline numbers hiding a problem? Let's split them by cohort and find out.

Talk to us about cohorts