Cohort dimension
The cohort dimension groups users by a shared starting point — typically their acquisition date — so you can follow each group's behaviour across subsequent days, weeks, or months. GA4 builds cohorts in the Cohort exploration from a first-touch criterion and a return criterion. It is the backbone of retention analysis, but small cohorts and identity loss can make later-period values unstable, so trends matter more than single cells.
What this means
A cohort is a set of users who share an inclusion criterion — most often the date of their first session. GA4's Cohort exploration then tracks a return criterion (any activity, a transaction, a specific event) across later time buckets, producing the classic retention triangle.
Reading rows shows how one acquisition group decays; reading columns compares the same period across cohorts.
Interpretation pitfalls
Later periods in a cohort are cumulative-attrition views and naturally thin out. Small cohorts swing wildly cell to cell, so judge the curve, not one number. Identity loss reassigns returning users into fresh cohorts, systematically understating retention. Define one return criterion and hold it constant, or cohorts across reports will not be comparable.
- Inclusion criterion is usually acquisition date
- Return criterion must be held constant
- Identity loss understates measured retention
How it appears in analytics and logs
A cohort cell is the count or rate of users from one acquisition group active in a later period. Thin later cells reflect attrition and identity loss, not necessarily a product problem.
Diagnostic use case
Use cohorts to measure retention — what fraction of users acquired in a given week return and act in the weeks that follow.
What WebmasterID can help detect
WebmasterID can frame retention from first-party acquisition signals while being clear that cleared identity reshapes cohorts and dampens measured return rates.
Common mistakes
- Over-reading a single thin late-period cell.
- Changing the return criterion between reports.
- Ignoring identity loss when retention looks low.
Privacy and accuracy notes
Cohorts group by first-party acquisition timing, not cross-site identity. Identity resets move users into new cohorts, which can understate true retention.
Related pages
- New vs established user dimension
The new vs established user dimension classifies a user as 'new' or 'established' based on whether GA4 had recorded prior activity for them before the reporting window. It is user-scoped and derived from the user's first-seen timestamp. This differs from the session-scoped new-vs-returning split, which classifies each visit; conflating the two produces mismatched user and session counts.
- Days since last session dimension
The days since last session dimension reports how many days have elapsed since the user's previous session. GA4 computes it from the stored last-session timestamp on the current identity. It supports recency and re-engagement analysis, but it can only be calculated when GA4 still recognises the user — if the identifier was cleared, the prior session is invisible and the return is counted as new, so the gap is undercounted.
- Predicted LTV bucket dimension
The predicted LTV bucket dimension groups users by GA4's modelled estimate of their future revenue, banding a continuous prediction into segments for audience building. GA4 generates predictive metrics like predicted revenue only when the property meets minimum data thresholds and has the required purchase events. These are model outputs, not observed facts, so they carry uncertainty and should never be reported as actual lifetime value.
- Attribution analytics
Retention by acquisition cohort, first-party.
Sources and verification notes
Last reviewed 2026-06-24. Facts are checked against primary/official sources where available; uncertain specifics are marked “Data not yet verified” rather than guessed.