GA4: Five Months Left, and Your Historical Data Is on the Clock

On July 1, 2023, Universal Analytics stops processing hits. Not "enters a deprecation period." Stops. Google has been saying this plainly for almost a year, and yet half the properties we audit in new-client discovery still have GA4 installed as an afterthought: default config, no events mapped, nobody looking at it.

Here is the part that makes this urgent rather than merely important: GA4 does not import your Universal Analytics history. Your year-over-year comparisons will only ever go back as far as your GA4 property has been collecting good data. Every week of delay in 2023 is a hole in your 2024 reporting. The deadline is July, but the real deadline was months ago, and the second-best time is this week.

What "migrated" actually means

Having a GA4 tag firing is not a migration. A real migration has five parts:

• An event taxonomy designed on purpose. GA4's event model is genuinely different from UA's pageviews-and-sessions world. Map your business questions to events first, then implement. Do not recreate your old goals one-to-one; half of them were cargo cult anyway.
• Conversions defined and tested. Including the ones feeding Google Ads bidding. If you run paid and your conversions are still importing from UA, July 1 breaks your bidding signal, which is a much more expensive problem than broken dashboards.
• Data retention settings changed from the default. GA4 ships with 2 months of user-level retention. Set it to 14 and do it today, because the setting is not retroactive.
• BigQuery export enabled. It is free to link on standard properties now, and it is your insurance policy against every GA4 interface limitation. Even if nobody on the team writes SQL yet, start the export. Future you will be grateful for the history.
• An export of your UA history. Decide what reporting you actually revisit, pull it into flat files or BigQuery via the API, and archive it before the interface access eventually goes away too.

Yes, GA4 is worse at some things. Migrate anyway.

Let us not pretend. The GA4 interface is slower, several beloved reports are gone, and marketers who could navigate UA blindfolded feel like they are wearing mittens. The complaints are legitimate. They are also irrelevant, because the alternative on July 1 is nothing. You can be annoyed and migrated at the same time; we recommend it.

There is also real upside once you stop grieving. The event model matches how people actually use products across web and app. BigQuery export at no cost was an enterprise feature two years ago. And GA4's modeling approach is at least an honest attempt at a world with less third-party data, which is the world we are all entering whether we migrate gracefully or not.

The two-week version for teams starting late

If you are reading this with nothing in place, here is the triage order: create the property, implement base tagging plus your five most business-critical events, define conversions, fix retention, link BigQuery, connect Google Ads. That is achievable in two focused weeks for most sites. Parallel-run against UA through the spring, expect the numbers not to match exactly, and write down why so nobody panics in a Monday meeting.

Then put a recurring event on the calendar for June: verify nothing still depends on UA. Audiences, imported conversions, Looker Studio dashboards, that one automated email report the CFO secretly relies on. July 1 is a hard cutover, and hard cutovers reward people who rehearse. Five months sounds like a long time. In analytics-implementation time, it is barely enough. Go.