GA4 Cohort Analysis: Build and Read a Weekly Retention Table

GA4 cohort analysis groups users who first arrived in the same period and compares, in a table, how many of them come back days or weeks later. The Google Analytics Help page describes a cohort as "a group of users who share a common characteristic." In the Korean interface, it appears under the name "donggil jipdan" (homogeneous group). This article covers the steps to build the table, what each setting means, how to read the table, and what to check when values do not appear.

GA4 cohort analysis compares users grouped by sign-up period

Suppose visitors are growing but revenue stays flat. The overall average mixes new visitors with loyal long-term ones, so you cannot see which side has the problem. Think of a school. Looking at the dropout rate of students who enrolled in March and those who enrolled in September separately is the same idea. The average attendance of the whole school will not tell you which enrollment period starts losing students. Splitting by enrollment period shows it right away. GA4 groups users by a criterion such as the date they first arrived and follows them the same way. Note that a small group can swing a lot from the actions of just a few people, so always read the rate together with the user count.

The same cohort exploration help page builds a cohort exploration from three parts. The inclusion criteria decide which users go into one group. The return criteria decide what counts as "came back." The granularity sets the table interval (daily, weekly, or monthly). For example, set inclusion to "first touch (acquisition date)," return to "any event," and granularity to "weekly." One cell then shows how many users who first arrived in the first week of January were still using the site two weeks later.

Cohort analysis looks at a different place than an ad performance report. An ad report gives you the total of clicks and conversions. A cohort tells you whether those converted people are still around afterward. We explain why growth experiments look at this return stage separately in our guide to the 5 AARRR stages of growth hacking.

How to create a GA4 cohort analysis exploration

According to the GA4 cohort exploration help page, you start from a template. A template fills in the settings for you, so it is easier at first than building a blank exploration yourself.

1. Open Explore in the left menu of GA4.
2. Next to "Start a new exploration," click Template gallery and choose Cohort exploration.
3. Under Cohort inclusion, choose the grouping criterion. At first, use "First touch (acquisition date)."
4. Under Return criteria, choose the action that counts as coming back. For using a service, pick "Any event." If purchase is the goal, pick "Any transaction" or a specific key event.
5. Under Cohort granularity, choose daily, weekly, or monthly, and then set the Calculation method.

After you build it, first check that the table is filled in. If values appear from the period right after sign-up and each cell has a user count, it is working. The most recent cohort has not had time to pass yet, so cells on the right may be empty. Do not read that as an error.

GA4 cohort settings: inclusion, return criteria, and calculation

Each setting changes the numbers in the table. The values from the Google help page are collected in one table below.

Setting	Available values	Meaning
Inclusion criteria	First touch (acquisition date), event, transaction, conversion, other events	Decides which action date is used to place a user in a group.
Return criteria	Event, transaction, conversion, other events	A user counts as "returned" if this action happens at least once in the exploration period.
Granularity	Daily, weekly, monthly	Weekly runs Sunday to Saturday, and monthly runs from the 1st to the last day of each month. It does not cut the data into rolling 7-day blocks.
Calculation: Standard	-	Counts only users who met the return criteria in each period.
Calculation: Rolling	-	Counts users who met the return criteria in that period and in all previous periods.
Calculation: Cumulative	-	Counts users cumulatively across the whole exploration period.

Starting with "Standard" is easier to read. Standard counts only the people who actually returned in week 3, so you can see right away which week dropped suddenly. Cumulative cell values keep rising. It fits better when you want to see how many people came back at least once as time passes, rather than find the point where they left.

Know the display limits too. According to the Google help page, up to 60 cohorts are shown, and when you break down by a dimension, only the top 15 values appear. For demographic dimensions such as age and gender, groups with few users that could identify individuals may be left out.

How to read a cohort table: retention calculation and what to judge

Each row of the table is one cohort, and the cells further right are later points in time. The retention overview report example in the Google help page shows how to read it. If 4 users first arrived on October 1 and 3 returned by the 2nd, day 1 retention is 75%. If 2 are active by the 8th, day 7 is 50%. If 1 is active by the 31st, day 30 is 25%.

Check three things in order when reading. First, check that the first cell of each row (right after sign-up) has enough users. Second, compare values at the same elapsed time vertically. If week 2 of the March cohort is lower than the February cohort, you have a reason to suspect the effect of an ad or a redesign that started that month. Third, look horizontally at where the curve bends. The same help page reads a cohort whose retention falls close to 0% as a sign that you are not holding on to users successfully.

Do not conclude the cause from the table alone. A cohort table tells you only "which group shrank and when." You must check separately whether the ad creative changed or whether the sign-up path changed. To split by traffic source, you need to tag each link with its source as described in our guide to UTM parameters in GA4. Only then can you compare them in the cohort table.

To set a target number, first decide which metric to target. Follow the order in what a KPI is and how to set marketing KPIs. Add return rate as a KPI candidate, and use the day 7 and day 30 values in the cohort table as your baseline.

The retention overview report differs from cohort exploration, and what to do when values are missing

GA4 also has a report that opens without any setup. According to the Google help page, the retention overview report under Reports > Life cycle > Retention shows day 1, day 7, and day 30 retention for cohorts grouped by acquisition date, with up to 3 lines. It also shows the daily return rate for the first 42 days and the average revenue (lifetime value) of new users over their first 120 days. You cannot change the inclusion criteria or return criteria.

Item	Retention overview report	Cohort exploration
Location	Reports > Life cycle > Retention	Explore > Template gallery
Cohort criteria	Fixed to the date the user was acquired	Choose from first touch, event, transaction, conversion
What you can see	Day 1, 7, 30 retention, 42-day curve, 120-day lifetime value	Daily, weekly, monthly tables, with your own return action
When to use	To get a rough view of retention first	To look only at purchasers or users who triggered a specific event

If the exploration table is empty or past cohorts do not appear, check the data retention setting. The Google data retention help page says event-level data can be kept for 2 months or 14 months, and this setting affects only explorations and funnel reports. If the report period is longer than the retention period, the extra range is not shown. Standard reports are not affected. An Editor changes the setting under Admin > Data settings > Data retention. The help page does not say whether periods that already passed come back after you change the retention period. If you plan to do cohort analysis, it is safer to set it to 14 months first.

The drop-off range you find in cohort analysis is easier to interpret when you view it alongside the stages customers go through. How to lay out the actions at each stage is covered in our guide to how to build a customer journey map.

You have two things to do today. First, check the retention period under Admin > Data settings > Data retention. Then open the cohort exploration template under Explore and build one table with first touch, any event, weekly, and standard. Write down the user count in the first row and the week 2 value. They become the baseline to compare with next month's cohort.

GA4 cohort analysis FAQ

Is one week in a weekly cohort the last 7 days?

No. According to the Google help page, weekly runs from Sunday to Saturday, and monthly runs from the 1st to the last day of each month.

What happens if there are more than 60 cohorts?

One table shows at most 60 cohorts. Narrow the date range or change the granularity to weekly or monthly to reduce the number of cohorts.

Do I still need cohort exploration if I have the retention overview report?

The retention overview is fixed to cohorts based on acquisition date. To group only users who purchased or triggered a specific event, you need to change the inclusion criteria in cohort exploration.