Cohort Analysis

Cohort analysis groups users by something they share — almost always the week they first arrived — and follows that group over time. Instead of asking "how many people visited last month," it asks "of the people who first visited in the second week of May, how many came back in week three? Week four?" GA4 builds this in the Explore section as a cohort exploration.

Why it matters

Most analytics numbers are a single snapshot. Cohort analysis adds the dimension that snapshots hide: whether the people you attract actually stick around. Two sites can have identical weekly traffic, but if one keeps bringing people back and the other doesn't, only a cohort view will show it. For a content-driven business, that difference is the whole story.

A concrete example

Suppose you publish a new article and 400 people read it in its first week. A cohort report lets you watch that specific group: 90 of them return the next week, 40 the week after. Compare that to a cohort from a viral social post — 2,000 readers, almost none of whom ever come back. Same traffic chart, completely different value. The cohort tells you which audience is real.

The common misreading

The trap is reading a young, thin cohort as a trend. A cohort that's only two weeks old, built on a few dozen users, will swing wildly week to week — and it's tempting to declare "retention is collapsing" or "we've cracked it" off that noise. A cohort needs enough people and enough weeks before its shape means anything. Early movement is usually just small numbers, not a direction.