How Universal Analytics worked
Universal Analytics was built around the session. A visit opened a session, everything the person did was recorded as a hit inside it — a pageview, an event, a transaction — and the session closed after a period of inactivity or at midnight. Reporting was organised the same way: sessions, bounce rate, pages per session, goal completions.
Structurally it was an account holding properties, and each property holding views, where a view was a filtered copy of the data. Views were the standard way to separate internal traffic, staging traffic and spam. Goals were configured in the interface, and ecommerce was a separate data layer arrangement bolted alongside the main tracking.
Google retired it in favour of GA4, first stopping the properties from processing new data and later removing access to the historical reports and the reporting API. Nothing of a Universal Analytics property is reachable now unless somebody exported the data while it was still available.
Why Universal Analytics still matters
It matters because the vocabulary survived it. Owners and agencies still quote bounce rate, sessions and goals as though they meant what they used to, and they do not. GA4 counts events rather than hits, so its idea of engagement is defined differently and its figures will not line up with anything remembered from before.
It also matters when older reports resurface. Historical decks, agency benchmarks and internal targets set under the old model are still circulating, and comparing them with current figures produces conclusions that are simply wrong. A drop that looks alarming is often just two different definitions sitting next to each other.
Where it goes wrong
The most damaging mistake was assuming the data would always be there. Many businesses never exported, and their trading history before the migration no longer exists in any form they can query. That is not recoverable, and it is worth stating plainly rather than hunting for a workaround.
The second mistake is porting old goals and definitions across unchanged. A goal that made sense as a session-level completion may need rethinking as an event, and copying it mechanically produces a metric with a familiar name and unfamiliar behaviour.
What to do about it
If you still hold an export from the old property, keep it somewhere safe and label it clearly as belonging to a different measurement model, so nobody charts it against current data. If you do not, start your comparison history from the GA4 data and say so; an honest short history beats a long misleading one.
Then check the current setup on its own terms rather than trusting a migration done in a hurry. Conversions, channel groupings and internal traffic exclusions all behave differently now, and a careful GA4 setup review usually finds at least one definition that was carried over rather than rebuilt.