Cohort Exploration

Cohort Exploration groups users by a shared start date, then tracks a metric for each group over the days, weeks, or months that follow.

You set the inclusion criterion, usually first touch, and the return criterion, usually any active event or a specific one like purchase. GA4 builds a triangular grid: each row is a cohort defined by when it started, and each column is a period after that start, showing how the metric holds up. It is the standard way to read retention.

You build weekly cohorts of new users and set the return criterion to any activity. Week zero is a hundred percent by definition, and you watch how fast each row decays. The cohort that started the week you changed onboarding retains noticeably better in week two, which is the signal the change worked.

Cohort tables have a data problem in the newest rows. The most recent cohorts have not had time to age, so their later columns are empty or thin, and early numbers there jump around on small samples. Read the mature cohorts for the real trend, not this week's row, which will keep changing as data lands.

Retention is the one metric that separates a growing product from a leaking one.