How GA4 decides someone is returning
Every visit arrives with an identifier: a first-party cookie on the web, an app instance identifier in an app, or your own User-ID if the person is logged in. When GA4 finds a record it already holds for that identifier, the visitor is counted as returning rather than as new. Nothing else is checked — not the login, not the device, not the intent.
That makes returning the mirror image of new, and it inherits the same weakness. Recognition depends entirely on the identifier surviving. Clear the browser storage, switch from a phone to a laptop, open a private window, or decline tracking, and a loyal customer appears as a first-time new user instead. The count therefore understates loyalty rather than exaggerating it.
One quirk catches people out in reports: new and returning users can add up to more than your total users for the period, because somebody whose very first visit falls inside the date range and who then comes back inside it belongs to both.
Why returning users matter
The split tells you which half of the job your marketing is doing. A business whose traffic is almost entirely new is buying attention repeatedly and keeping very little of it. One whose traffic is almost entirely returning has a loyal base and a reach problem. Both are fixable, and the fixes are opposite.
Returning visitors also behave differently, and averaging them together with first-timers hides that. They convert on different pages, need less introduction and respond to different offers, which is exactly the logic behind remarketing campaigns that treat the two groups separately.
Common mistakes with returning users
The first is reading the count as a measure of loyalty. Any device or privacy change erodes recognition, so a falling returning share can reflect browser behaviour rather than a falling audience. Compare it with something you own, such as repeat orders or repeat enquiries, before drawing a conclusion.
The second is judging campaigns on it without context. An awareness campaign should raise new users; a nurture email should raise returning ones. Marking either result as a failure because the other number did not move is a reporting error, not a performance problem.
The third is comparing date ranges of different lengths. A longer window gives people more chances to come back, so the returning share rises with the size of the range on its own.
How to act on it
Segment before you conclude. Look at the new and returning split by channel, by landing page and by device, and see where the returning visitors actually go: if they head straight to pricing or contact, your job is to remove friction, and if they re-read the same guide, your job is to answer the next question.
Improve recognition where you legitimately can — first-party cookies, cross-domain measurement between your own sites, and a User-ID for logged-in customers. Then tie the picture back to real repeat business from your CRM or booking system, because that is the number the analytics figure is only ever approximating.
