GA4: Why Google Gives Away the Measurement Layer

Context

Google Analytics is the most widely deployed analytics tool on the planet, and it has been free for nearly its entire life. Free, on a product Google poured two decades of engineering into, from a company that monetizes almost everything it touches. The instinct is to call this generosity or a loss leader. It is neither. GA is free because measurement is not the product Google is selling. Measurement is the moat around the product Google is selling, which is the ads auction.

What I noticed

Follow the value. When you install GA on your site, you instrument every page, every event, every conversion, and you hand Google a structured map of what users do after they click. Google does not charge you for that map. It does not need to. The map makes Google’s advertising machine smarter, and the advertising machine is where the money is. The tag on your site that you think of as your analytics is also a sensor feeding the conversion-optimization models that decide what your competitors pay per click.

This reframes every product decision in GA4. The questions the product optimizes for are not always the analyst’s questions. They are Google’s questions: did this user convert, how do I attribute that conversion back to an ad, how do I model the conversions I can no longer observe directly. I work on the data behind a DSP and an exchange, and I can tell you that whoever owns the conversion event owns the optimization loop. Google understood that owning the default measurement layer of the open web is worth more than any licence fee, because the conversion signal is the raw material of the auction, and the auction is a money printer.

The forced migration

The clearest evidence that GA serves Google first is the Universal Analytics to GA4 migration. Google deprecated UA on a hard deadline, stopped data collection, and forced millions of sites onto GA4 whether they wanted it or not. The backlash was loud and justified. GA4 was not a friendlier UA. It was a different data model entirely.

UA was session-and-pageview based, the mental model analysts had used for fifteen years. GA4 threw that out and rebuilt everything as events. Every interaction became an event with parameters, reports analysts relied on vanished or moved, the interface got harder, and historical data did not cleanly carry over. From the analyst’s chair this looked like a downgrade dressed as an upgrade.

But look at it from Google’s chair and it is obviously correct. An event-based model is exactly the schema you want if your real goal is to track conversions across web and app in one stream, feed them into machine-learning attribution, and pipe them into Google Ads. GA4 was not remodelled for the analyst. It was remodelled for the optimization loop. The session model was a relic of a measurement-first product; the event model is the native language of an ads-first product. The migration pain was the cost of realigning GA with what Google actually needs it to do.

Consent mode and modelling

The second tell is how GA4 handles the privacy collapse. As consent banners and tracking restrictions knocked out a growing share of observable conversions, Google did not just report less. It introduced consent mode and conversion modelling, which fill the gaps in observed conversions with modelled estimates so the ads auction keeps getting a usable conversion signal even when the user said no to tracking.

Think about whose problem that solves. An analyst who wants ground truth is uneasy about modelled conversions, because a model is not a measurement. But Google’s bidding algorithms do not need ground truth, they need a continuous, dense signal to optimize against, and a modelled conversion is good enough to keep the auction liquid. Consent mode is a measurement compromise and an advertising rescue at the same time, and GA4 ships it as a headline feature because the advertising side is the side that pays.

The product decision and the opening

Here is the decision laid bare: Google optimizes GA4 for Google’s needs, ads integration and conversion modelling, ahead of the analyst’s needs, clean data and ergonomic exploration. That is a defensible call for a company whose analytics tool is a sensor for its ad business. It is also exactly why third-party analytics keeps finding oxygen.

Every place GA4 underserves the analyst is a wedge. Tools that promise first-party-owned data, simpler product analytics, server-side control, privacy-by-design, or just an interface that respects how analysts actually think have a market precisely because GA4 was never really built for the analyst. The free tool with the conflicted owner leaves a gap, and the gap is large enough that paid, neutral analytics is a thriving category despite competing with free. You cannot out-price free. You can out-serve a free product that is secretly optimizing for someone else.

The PM lesson

When a powerful company gives away a category-defining product, ask what it is actually selling. GA is free because the asset is not the software, it is the conversion signal the software collects, and that signal feeds a market far larger than analytics ever could be. The strategic risk Google accepts in return is that a measurement layer optimized for the auction will always serve its owner before its user, and that gap is a permanent invitation to every competitor willing to simply build for the analyst instead.