Retention in GA4: the metric and the setting that deletes your data

Retention is about customers who come back. Of the customers who bought in January, how many bought again in February? That share is your retention, and the rest is your churn.

Inside GA4 the same word points at something else: data retention, the period after which Google deletes your event data. That period defaults to two months.

Mixing the two up tends to surface at the worst moment. Usually in October, when you want last year's retention curve and the data is gone.

GA4 gives you two places for this. The Retention report under Lifecycle shows returning users, new users and a retention curve per cohort of users who arrived on the same day.

The serious work happens in Explore, in the cohort exploration. There you pick the inclusion criterion yourself (first visit, or an event such as purchase), the return criterion, and the cohort granularity: daily, weekly or monthly.

For an online shop, a monthly cohort built on purchase is usually the one worth reading. Per purchase month you see how many of those buyers came back in month 1, 2 and 3, and whether a cohort from after a campaign holds up better than the one before it.

One warning. GA4 counts users based on the identity it has. Without consent, without a user ID, and across devices, one returning customer splits into several users. Your GA4 retention will sit structurally below what your order system reports.

You find it under Admin > Data settings > Data retention. Event data retention has two options: 2 months or 14 months. New properties start at 2 months.

Set it to 14 months. There is barely a reason not to, and the choice costs you nothing. Analytics 360 offers longer periods.

Below it sits Reset user data on new activity. With that on, the retention clock for a user starts over whenever that same user does something again, so someone who returns monthly never drops out of your user data. With it off, the clock keeps running from the first touch.

Whatever has expired does not come back. Raise the setting from 2 to 14 months in May and the longer history builds from that point forward. What Google already deleted is gone.

This is where most of the confusion lives. The retention period applies to the underlying event and user data, not to the standard reports.

• Standard reports keep reaching further back. They are pre-aggregated, so sessions and users per channel for last year still show up.
• Explorations, cohorts, funnels, paths and segments do lose the data. They query raw events live. With the setting on 2 months, an exploration over last quarter comes back empty.
• Custom dimension combinations that fall outside a standard report hit the same wall.

The painful consequence is year-on-year comparison. A cohort analysis or an exploration covering the same period last year needs at least 14 months of data, and 14 months is the ceiling on a free property.

If you need more than 14 months at event level, the BigQuery export is the answer. Switch it on under Admin > Product links > BigQuery links. GA4 then writes a daily table of your raw events into your own project, where the GA4 retention period no longer applies.

What lands in BigQuery stays there as long as you want it to. You set the table expiration yourself, or you set none.

The export only fills from the day you enable it. Same rule as the retention setting: late is late. Turn it on before you need it.

• Go to Admin, Data settings, Data retention now and set event data to 14 months. Check the properties someone else created too.
• Enable the BigQuery export as soon as you know you will want to look back further than 14 months.
• Use the cohort exploration for retention as a metric, with purchase as the return criterion for ecommerce.
• Do not compare GA4 retention one to one with your order system. GA4 counts users, your backend counts customers.
• If an exploration comes back empty for an older period, check the retention setting before you start debugging your tagging.