Web and product analytics
Cohort analysis explained, with GA4 steps
The short answer
A cohort analysis groups customers by when they started (their first visit, signup or purchase) and tracks what share are still active in each later period. Read down a column to compare cohorts at the same age, and along a row to see one cohort decay. In GA4, use Explore, Cohort exploration, with first touch as inclusion and weekly or monthly granularity.
On this page 6 sections
Why averages mislead
A business can show flat overall retention while every new cohort is worse than the last, because older loyal customers prop up the average. Or a pricing change can quietly improve retention for everyone who joined after it, invisible in a blended number. Cohorts separate the effect of time-since-joining from calendar time.
How to read the triangle
| Cohort (joined) | Users | Month 1 | Month 2 | Month 3 |
|---|---|---|---|---|
| July | 400 | 42% | 31% | 27% |
| August | 520 | 38% | 29% | - |
| September | 610 | 45% | - | - |
- Read down a column to compare cohorts at the same age. September's month-1 retention (45%) beats August's (38%): something that changed in September may be working.
- Read along a row to see one cohort decay. Look for the month the curve flattens; customers still active there are your durable base.
- The diagonal is calendar time. A dip across a diagonal points to an outage, a season or a price change hitting everyone at once.
- Weight by size. A cohort of 40 swings far more than one of 600.
Standard, rolling and cumulative
- Standard: users who met the return criteria in that period only. The usual retention view.
- Rolling: users who returned in that period and every period before it. Strict; shows truly habitual use.
- Cumulative: users who returned in any period up to that point. Useful for revenue per user over time.
Build it in GA4
- Explore, then Cohort exploration.
- Cohort inclusion: first touch (acquisition date) for new users, or a specific event such as sign_up or purchase.
- Return criteria: any event, any transaction, any key event, or a specific event (for a SaaS, a core action rather than any page view).
- Granularity: daily, weekly (Sunday to Saturday) or monthly. Weekly suits most small sites.
- Calculation: standard, rolling or cumulative.
- Optionally add a breakdown such as first user default channel group to compare retention by acquisition channel. GA4 shows up to the top 15 breakdown values and up to 60 cohorts.
GA4's cohorts are user-based and depend on cookies, so they understate retention for people who switch devices or decline consent. For customers and revenue, build cohorts from your billing or store data, or from the BigQuery export. Paste your own numbers into the cohort retention calculator for a table and average curve, and connect retention to value with the churn rate and LTV to CAC calculators.
Questions
- What is a cohort in analytics?
- A cohort is a group of users who share a starting characteristic, most often the week or month they were acquired or first purchased. Cohort analysis tracks what share of each group stays active over time.
- What is the difference between standard and rolling retention in GA4?
- Standard retention counts users who returned in a given period. Rolling retention counts only users who returned in that period and in every previous period, so it is always equal or lower.
Sources
Next step
Try the cohort retention calculator
Related guides
SaaS metrics: LTV, CAC, payback and churn explained
How to calculate customer lifetime value, customer acquisition cost, CAC payback and churn for a SaaS or subscription business, and what the ratios mean.
GA4 BigQuery export: set-up, limits and costs
How the free GA4 to BigQuery export works: daily vs streaming export, the 1 million events a day limit on standard properties, table names, sandbox, and what it costs.
How to build a funnel exploration in GA4
Build a GA4 funnel exploration step by step: open vs closed funnels, directly vs indirectly followed by, elapsed time, next action and breakdowns, with a lead-gen example.
What is predictive analytics for a small business?
Predictive analytics in plain terms: what it predicts (churn, lifetime value, lead score, demand), the data you need, and how a smaller business starts.