← Glossary · Retention & cohorts

Cohort analysis

Also: Cohort matrix · Cohort report

Cohort analysis groups users by when they were acquired and tracks each group's behavior — retention, revenue, LTV — by days since install. It is how UA teams compare the quality of users bought in different weeks or from different channels.

The standard view is a matrix: acquisition week on the rows, D1 / D7 / D14 / D30 / D60 / D90 on the columns, and a cumulative value in each cell. Read down a column to see cohort quality over time; read across a row to see the revenue curve.

The bottom-right of the matrix is always empty — recent cohorts have not reached later days. Those cells must be marked immature, not averaged in as zeros.

Common mistake

A "last 30 days LTV" dashboard tile. It blends cohorts at every age and moves when acquisition volume moves, not when the product does.

Go deeper: Reading a cohort matrix

Related terms

In Roasy

Roasy computes this across every ad network, Adjust, and RevenueCat on one screen — real revenue, cohort windows, and the same definition everywhere.