Your monthly report says returning visitors are up 10%. Great news? Maybe. Or maybe you simply acquired a lot more people last quarter, and each new group is actually less loyal than the one before. Totals and averages blend old and new customers together, so they can hide a retention problem for months.
Cohort analysis fixes that. It groups people by a shared starting point, usually when they first visited or first bought, and then tracks what each group does over the following days, weeks or months. You can see whether January's customers come back more than June's, whether a new onboarding email improved repeat purchases, or whether a paid campaign brought in bargain hunters who never return.
This guide explains how cohort tables work, the main types of cohort, how to build a cohort exploration in Google Analytics 4 step by step, how to read the results and what to do with them.
Key Takeaways
- A cohort is a group of users who share a starting characteristic, such as their first-visit week or first-purchase month.
- Read cohort tables down the columns to compare groups at the same age, and across the rows to see how one group decays.
- GA4's Cohort exploration supports daily, weekly or monthly cohorts and standard, rolling or cumulative calculations.
- Explorations use your data retention setting, so set it to 14 months before you need long-range cohorts.
- Break cohorts down by first-user channel to see which marketing brings customers who stay.
What is cohort analysis?
Google's GA4 documentation gives a crisp definition:
“A cohort is a group of users who share a common characteristic that is identified in this report by an Analytics dimension.”
— Google Analytics Help, [GA4] Cohort exploration
In marketing, the most common characteristic is time of acquisition: everyone who first visited in week 12, or everyone who first purchased in March. You then measure a behaviour (returning, purchasing, revenue) for each cohort in each later period. Because each row starts from the same point, you compare like with like.
How to read a cohort table
Here's a simplified monthly retention table for an online shop. Each row is a cohort of first-time buyers; each column shows the share who purchased again in that month after their first order. The figures are illustrative, not benchmarks.
| First purchase | Customers | Month 1 | Month 2 | Month 3 |
|---|---|---|---|---|
| January | 400 | 18% | 12% | 10% |
| February | 520 | 15% | 10% | 8% |
| March (new welcome flow) | 480 | 22% | 15% | – |
| April | 610 | 21% | – | – |
- Read across a row to see how a single cohort decays. January falls from 18% to 10%.
- Read down a column to compare cohorts at the same age. Month 1 repeat rate rose from about 15% to 22% after the new welcome flow.
- Watch the diagonal (same calendar month across cohorts) to spot seasonal or site-wide effects, like a sale or outage.
Types of cohort analysis
| Cohort type | Grouped by | Good for |
|---|---|---|
| Acquisition cohort | Date of first visit or first purchase | Retention trends over time |
| Behavioural cohort | An action, e.g. used site search, joined newsletter | Finding behaviours linked to loyalty |
| Channel cohort | First-user channel or campaign | Judging marketing quality, not just volume |
| Product cohort | First product or category bought | Cross-sell and repeat purchase planning |
| Plan or segment cohort | Plan, location, device | Pricing and experience decisions |
How to build a cohort exploration in GA4
- Go to Explore and choose Cohort exploration (or start blank and select that technique).
- Cohort inclusion: choose how users enter a cohort, e.g. First touch (acquisition date), any transaction, or a specific event such as
purchase. - Return criteria: choose what counts as coming back, e.g. any event (active users) or any transaction (repeat buyers).
- Cohort granularity: pick daily, weekly or monthly. Weekly suits most content sites; monthly suits most shops.
- Calculation: pick standard, rolling or cumulative (see below).
- Breakdown: add a dimension such as First user default channel group to compare channels.
- Values: choose the metric, such as active users, transactions or purchase revenue, and whether to show totals or per-user values.
- Set the date range in the Variables panel and review the table. Name and save the exploration.
Standard, rolling or cumulative?
- Standard: a user counts in any period where they met the return criteria, regardless of other periods. Best for general retention.
- Rolling: a user counts only if they returned in that period and every earlier one. Best for habit-forming products.
- Cumulative: a running total across periods. Best for revenue per user over time, a useful proxy for customer lifetime value.
Two important caveats. Explorations only cover your data retention period; Google states the retention setting affects explorations and funnel reports, so set it to 14 months in Admin now. And very large queries may be sampled. For multi-year cohorts, export raw events to BigQuery; standard properties have a daily export limit of 1 million events.
Turning cohort insights into action
- Weak month-1 retention: improve onboarding emails, post-purchase follow-ups and first-visit experience. See our customer retention guide.
- One channel's cohorts decay fastest: shift budget or change targeting; cheap acquisition isn't cheap if nobody returns.
- Behavioural cohort retains better (e.g. newsletter subscribers): push more users toward that behaviour with email list building.
- A cohort spikes after a change: confirm with an A/B test before rolling it out everywhere.
- Retained users share traits: turn them into a GA4 audience for lookalike targeting and exclusions.
For simple checks, the standard Retention report in GA4's Life cycle collection shows new vs returning users and a basic cohort chart. Our GA4 reports guide shows where it sits; the exploration is where you get control.
Common mistakes in cohort analysis
- Comparing incomplete periods. The newest cohort hasn't had time to return; don't judge it against older ones on later columns.
- Cohorts that are too small. A daily cohort of 30 users swings wildly. Use weekly or monthly granularity.
- Wrong return criteria. “Any event” includes a single page view; for shops, repeat purchase is usually the real goal.
- Ignoring seasonality. November cohorts acquired by discounts often behave differently from March cohorts.
- Short retention settings. Two months of data can't show a 6-month retention curve.
- Analysis without action. Tie every cohort review to a decision in your data-driven marketing process.
Related Guides
- Data-Driven Marketing: A Practical Framework for Small and Mid-Sized Teams
- How to Build a Marketing Dashboard
- Google Consent Mode v2 Explained
Frequently Asked Questions
What is cohort analysis in simple terms?
Grouping customers or users by when or how they started, then tracking each group's behaviour over time so you can compare like with like.
Where is cohort analysis in GA4?
In Explore, choose Cohort exploration. The Retention report in the Life cycle collection offers a simpler view.
What is the difference between standard, rolling and cumulative cohorts?
Standard counts users active in a given period; rolling counts only users active in that period and every earlier one; cumulative shows a running total.
How far back can GA4 cohort explorations go?
As far as your data retention setting: 2 or 14 months on standard properties. Use BigQuery export for longer histories.
Which granularity should I use?
Weekly for most websites and content, monthly for ecommerce and longer buying cycles, daily only for high-volume apps.
Conclusion
Cohort analysis is the clearest way to see whether your marketing builds lasting customers or just traffic. Set GA4 retention to 14 months, build one monthly cohort exploration broken down by channel and review it every month alongside your acquisition numbers. If you'd like help connecting retention data to campaigns that grow revenue, explore our conversion optimisation services or request a quote.
