How reporting identity works
Every analytics tool has to answer one awkward question: when the same person turns up twice, how does it know? Reporting identity is where you tell GA4 which pieces of evidence it is allowed to use. It sits in the property’s admin settings and it changes how reports are counted, not what is collected.
GA4 offers three settings. Device-based uses only the identifier stored in the browser, so a phone and a laptop are always two users. Observed adds the User-ID you send when someone logs in, and the signed-in evidence from Google signals if that is switched on. Blended starts with the observed evidence and then models the gaps, estimating the missing part of the picture where consent or cookies are absent.
The important mechanical detail is that the setting is applied when a report is drawn, not when data arrives. Switch it and yesterday’s report redraws with the new rule. Nothing is destroyed and nothing is recovered — you are choosing a lens, and you can put a different one on tomorrow.
Why reporting identity matters
It quietly decides whether your user count is believable. A site with a login and a busy mobile audience will report far more users on the device-based setting than on observed, because every device counts separately. Conversion rates move with it, since the denominator changes while the number of enquiries does not.
It also decides how much of your reporting is measured and how much is estimated. Blended is the most complete-looking, and it is the least literal. If you are presenting numbers to a client, a board or a lender, that distinction is worth knowing before somebody else points it out.
Common mistakes with reporting identity
The most damaging is changing it mid-year and then comparing periods as though nothing happened. The reports will happily draw a step change that has no cause in the real world. Any switch needs a note in the reporting pack and a redrawn baseline.
The second is expecting observed or blended to work when nothing feeds them. Without a login sending a User-ID, and without signals enabled, all three settings resolve to roughly the same device-level counting. The third is comparing GA4’s user count to a figure from an advertising platform and treating the difference as an error, when the two are simply counting under different rules.
How to act on it
Pick one setting, write down why, and leave it alone. For most small businesses without a login, device-based is the honest default: it says what was seen and claims nothing more. If you run a membership site, a portal or an ecommerce account area, observed is usually the better fit because the identity is real rather than modelled.
Choose blended only when you understand that part of the number is an estimate and you are willing to say so in the report. Whichever you choose, record it alongside the property’s other settings during GA4 setup, so the next person to open the account knows why the numbers look the way they do.
