July 1, 2023: RIP Universal Analytics

Written in 2026, looking back at the GA4 deadline, now that GA4 is just the water we swim in.

On July 1, 2023, Universal Analytics stopped processing new data. Google had warned everyone for a year, put the date on a banner inside the product, sent the emails. And still, when the day came, a meaningful share of the accounts I touch were not properly ready, including a couple I’d personally sworn were fine. There is nothing like a hard deadline to reveal which of your good intentions were real.

This one isn’t a story about AI. It’s a story about how migrations actually go, which is never the way the migration guide says they will.

The war story part

Universal Analytics and GA4 don’t just look different, they think differently. The old one counted sessions and pageviews the way we’d all built our brains around for a decade. The new one counts events, and a lot of the numbers you thought you understood simply don’t line up across the two. So for months, every client conversation had the same awkward middle: yes, the traffic looks lower, no, the traffic isn’t lower, it’s counted differently, please don’t panic, here is a chart explaining the chart.

The scramble had a texture I remember well. Clients who’d ignored every warning suddenly wanting it done yesterday. Historical data that didn’t automatically carry over, so anyone who hadn’t exported was about to lose their own past. Goals that had to be rebuilt as events from scratch. Reports that had to be re-taught to people who’d finally, after years, gotten comfortable with the old ones. I spent a good chunk of that spring not doing SEO at all, just moving houses so the analytics wouldn’t go dark.

The quiet tragedy was the historical data. GA4 didn’t inherit your Universal Analytics history. If you hadn’t exported it before the shutdown, your years of clean before-and-after comparison were just gone. I lost some for accounts that came to me late, and it still bothers me, because so much of my job is being able to say here’s what changed, and you can’t say that without a yesterday to point at.

The practical retrospective part

A few things I’d tell myself in early 2023, now that I’ve lived on the other side.

Export everything, earlier than feels necessary. The history you don’t save is history you can’t ever get back, and you always want it eighteen months later for a comparison you didn’t foresee.

Don’t trust the automatic setup. The assisted migration Google offered was fine for a rough start and wrong in the details, and the details are where reporting lives. Every account needed hands on it.

Reset expectations before the numbers do it for you. The single most useful thing I did was warn clients in advance that the figures would move and why, so the drop read as a measurement change and not a disaster. The clients I warned stayed calm. The one or two I forgot to warn did not, and that was my fault, not GA4’s.

What it turned out to mean

At the time the whole thing felt like busywork, a forced march with no upside, Google reorganising my furniture for its own reasons. And honestly, a lot of it was.

But GA4’s event model, annoying as it was to adopt, turned out to fit the world that was coming better than the old session model did. As traffic got weirder, as clicks and rankings drifted apart, as visits started happening in places analytics can’t see, the rigid old pageview picture would have flattered me with numbers that meant less and less. I didn’t appreciate that in July 2023. I was too busy rebuilding goals and apologising for charts. The migration I resented was quietly better preparation for the mess ahead than the comfortable thing it replaced.