Cohort Analysis Explained: How to Build and Interpret Cohort Reports

What is a cohort and why cohorts matter

A cohort is a group of users who share a common characteristic at a specific point in time — most commonly, the time period in which they first used your product. An acquisition cohort groups all users who signed up in the same week or month. An activation cohort groups all users who completed a specific action in the same period. The reason cohorts matter: aggregate metrics like 'monthly active users' mix together users from different time periods who may behave very differently. A product that's acquiring 1,000 new users per month but has 80% D30 churn looks fine in aggregate DAU — but a cohort view reveals a company pouring water into a leaky bucket. Cohort analysis exposes the leak.

Retention cohorts: the most important chart in product analytics

A retention cohort report shows, for users acquired in a given period, what percentage were still active at each time interval after acquisition (D7, D14, D30, D60, D90). The visualization is a grid — rows are acquisition cohorts (Jan users, Feb users, etc.), columns are time intervals, and the values are retention percentages. Two patterns matter most. First, the 'curve shape' — does retention flatten out above 0%, indicating a core retained user base, or does it trend toward zero, indicating no product-market fit for the acquired cohort? Second, the 'cohort improvement trend' — are more recent cohorts (recent rows) retaining better at each interval than older cohorts? Improving cohort retention over time is the most important leading indicator of product-market fit momentum.

Behavioral cohorts: going beyond acquisition date

Behavioral cohorts segment users by actions they took, not just when they signed up. A behavioral cohort might be: 'users who completed onboarding in their first 7 days,' 'users who invited a teammate in week 1,' or 'users who used Feature X in their first session.' The power of behavioral cohorts is that they let you test the impact of specific product behaviors on long-term outcomes. If users who invite a teammate in week 1 retain at 65% at D30 while users who don't retain at 22%, you've identified a high-impact activation target. This analysis is the foundation of activation milestone discovery and is the core use case for tools like Amplitude Compass.

How to build a cohort analysis in Amplitude

In Amplitude, go to Retention → New Retention Analysis. Set your start event (typically 'any event' or your signup event for an acquisition cohort, or a specific action for a behavioral cohort). Set your return event (the action that counts as 'retained' — typically 'any event' for broad retention, or a core feature action for engaged retention). Set the measurement to weekly intervals, set the look-back window to 8-12 weeks. Enable 'Usage Interval' mode to see the classic retention grid. Add a segment or group-by property (acquisition channel, user plan tier, or any user property) to compare retention across segments. The resulting grid is your retention cohort report.

Common misinterpretations and how to avoid them

The most common misinterpretation of cohort analysis is confusing improving aggregate metrics with improving cohort retention. A product that improves its D30 retention from 15% to 20% while also increasing acquisition 3x will show rising DAU — but the underlying retention improvement is masked by the acquisition volume. Always analyze cohort retention separately from growth volume. A second common error: comparing cohorts of different sizes without normalizing. A cohort of 50 users has much more statistical noise than a cohort of 5,000 — small cohorts should be grouped by month rather than week to produce stable retention estimates. Third: attributing cohort retention differences to the wrong variable. If January cohorts retain better than December cohorts, it might be seasonality, a product change, or an acquisition mix shift — always investigate before concluding.