GA4 isn’t the future anymore. It’s the only option.
Google Analytics 4 was released in October 2020 as a complete overhaul of Universal Analytics. For three years the two ran side by side and migration was something you could plan around. That period is over.
Standard Universal Analytics properties stopped collecting data on 1 July 2023. On 1 July 2024 Google shut UA off entirely — 360 properties included. The interface went dark, the API stopped answering, Looker Studio connectors broke, and the historical data behind all of it was deleted. If you did not export it before that date, it is gone. There is no recovery path and no support ticket that brings it back.
So the question is no longer whether to migrate. It is whether the GA4 property you migrated to is actually configured properly — and in our audits, more often than not, it is not.
What genuinely changed
The data model. This is the real difference, and everything else follows from it. UA organized the world into sessions, pageviews and a fixed vocabulary of categories, actions and labels. GA4 records everything as an event with parameters. That is far more flexible — you can model almost any interaction — but it also means the reporting you inherit is only as good as the event taxonomy someone designed for you. A sloppy taxonomy produces a GA4 property that is technically collecting and practically useless.
Modeling, not magic. GA4 does apply machine learning, and it is worth being precise about what for: filling gaps left by consent refusals and cross-device journeys, and driving predictive audiences. It does not discover your conversions for you. Enhanced measurement picks up scrolls, outbound clicks, file downloads and site search automatically, and that is a genuinely useful starting point — but it is a starting point, not a measurement plan. Every conversion that matters to your business still has to be deliberately defined, tagged and tested.
BigQuery for everyone. Under UA, raw data export to BigQuery was a 360 feature with an enterprise price tag. In GA4 it is available on the free tier, and it is the most under-used feature in the product. It means any business can own its raw event data, query it properly, join it to CRM and cost data, and stop being limited by what the GA4 interface is willing to show. If you take one action after reading this, make it turning that export on — it is not retroactive, so every day you wait is a day of raw data you will never have.
Retention defaults. GA4 keeps user-level and event-level data for 14 months at most, and the default is shorter. Aggregated standard reports survive; the granular exploration data does not. Another argument for the BigQuery export, where retention is your own decision.
Privacy controls. GA4 was built after GDPR and CCPA rather than retrofitted to them. Which raises the thing that catches most organizations: since March 2024, Consent Mode v2 has been mandatory for EEA traffic if you use Google Ads or GA4 audiences. Non-compliance does not throw an error. It quietly degrades your remarketing lists and conversion modeling while the interface continues to look completely normal.
What this means in practice
The migration deadline did its job and nearly everyone moved. What far fewer organizations did was revisit the event taxonomy, wire up BigQuery, or verify their consent implementation. Those three are where the actual value sits, and all three are still open on most properties we look at.
If you are not sure which of them applies to you, that is exactly what an audit is for.