Cohort Analysis
Cohort analysis is a method of grouping users who share a starting point or characteristic, such as the month they signed up, and tracking how each group behaves over time. Comparing cohorts side by side shows whether retention, engagement or revenue is improving for newer users, a trend that totals across the whole user base often hide.
How Cohort Analysis Works
A cohort analysis needs three choices:
- The cohort definition. Acquisition cohorts group users by when they started, such as sign-up week or month. Behavioral cohorts group users by something they did, such as completing setup on day one. Amplitude describes acquisition cohorts as showing when retention changes, and behavioral cohorts as helping explain why.
- The metric. Retention is the most common, but cohorts can track revenue, orders, feature use or any repeated behavior.
- The interval. Days, weeks or months since each cohort started, matched to how often the product is naturally used.
The result is usually a cohort table. Each row is a cohort, each column is a period since that cohort started, and each cell shows the metric, such as the percentage still active. Read a row from left to right to see one cohort's curve. Read a column from top to bottom to see whether that period is improving for newer cohorts. Effects tied to a calendar date, such as an outage or a seasonal spike, show up along a diagonal because they hit every cohort at a different age.
Why Cohort Analysis Matters
Aggregate numbers mix old and new users together. A product with fast-growing sign-ups can show rising total active users while each new group retains worse than the last. Cohort analysis separates those effects, so teams can see whether the product itself is getting better.
It also makes changes measurable. If a new onboarding flow ships in March, comparing March and later cohorts with earlier ones shows whether retention improved. Revenue cohorts show how spending grows or shrinks for each group, the base for metrics like net revenue retention. Behavioral cohorts help explain churn by showing which early actions separate users who stay from those who leave. For turning cohort and segment signals into roadmap choices, see mapping product health metrics and user behavior to roadmap decisions.
Cohort Analysis Example
A habit-tracking app builds a monthly retention table:
| Sign-up cohort | Users | Month 1 | Month 2 | Month 3 |
|---|---|---|---|---|
| January | 1,000 | 40% | 30% | 26% |
| February | 1,200 | 42% | 33% | 29% |
| March | 1,500 | 48% | 38% | not yet |
Reading down the Month 1 column, retention rose sharply for March, the month a redesigned first-day experience launched. Reading across, each cohort still loses users after the first month, so the team's next question is what happens between months 1 and 2.
Cohort Analysis vs. Segmentation and Retention Analysis
Segmentation splits users by an attribute, such as plan or country, at one point in time. A cohort is a group followed over time, which makes cohort analysis inherently about change. Retention analysis is narrower in topic but broader in method: it focuses only on why users stay or leave, and uses cohort analysis as one of its main tools alongside retention curves and behavioral comparisons.