Google Analytics 4: What it promises, and what that really means
Hey Google: 2016 called. It wants its analytics back.
If you’ve been following Heap at all, you probably know we’ve often been critical of Google Analytics. While it’s great for tracking ad spend and web behavior, when it comes to seeing what users actually do once they reach your site, we’ve long seen it as an inferior solution.
Apparently Google’s been listening! With Google Analytics 4, Google is changing, well, nearly everything. For the first time, GA users can track more than sessions and pageviews. GA offers more views, new engagement metrics, and more detail across the board. Above all, GA4 is switching to an event-based model, a major change from GA’s earlier sessions-based model.
From our perspective, the announcement of GA4 shows that Google is recognizing what we’ve always believed: that delivering insights on user behavior is the best way to improve your site or product. In short, GA4 is recognizing what people need and attempting to be a serious player in the analytics game. That’s all great news. But … once you start looking under the hood, it’s not so hard to see that the analytics paradigm GA4 delivers is based on features that were innovative in 2016, but not today.
Companies across the analytics world have since developed features that far exceed what Google offers. GA4’s “new” features may have been table stakes 6 years ago, but today fall pretty far short.
Above all, Google is promising "automated insights about the user journey, and easy activation to improve digital marketing ROI." Let's look at some of these promises and see what they really mean.
Codeless and Automatic tracking
With GA4, Google promises something called “codeless tracking.” This turns out to be a kind of manual tracking, albeit one that doesn’t require engineers to set up. This can save significant time. It’s certainly better than the standard manual tracking offered by companies like Amplitude and Mixpanel.
However, codeless tracking requires using both GA4 and Google Tag Manager together. This makes things too complex for most business users to be able to self-serve.
And, more importantly, with codeless tracking you still have to decide what events you want to track in advance. With this method, you’ll still end up with an incomplete, biased dataset. GA4 doesn’t offer retroactive data capability, so you’re limited to collecting data on events from the moment you decide to track them. Forget to track something? Too bad!
The biggest weakness here is that because you have to decide what to look at, GA4 offers no capabilities that surface things you’re not already tracking. Alternate user paths? Unseen moments of friction? User backtracking? With GA4, there’s no way to see these things. You’re left working with only user behavior that you know about, not user behavior as it actually happens.
GA4 does provide a list of things it tracks automatically. These automatically-tracked events DO NOT include clicks on links or specific elements in your site, such as clicking an “add to cart” button or a link to view a testimonial.
There’s also a list of “enhanced measurement” events, which need to be manually configured up front, but once enabled, they behave like additional auto-captured events. These include when users see ads, interact with embedded videos, or click links that take them out of your site or app.
The weakness here is the lack of flexible targeting. When users click around a page, the thing you care about — the user action, like adding an item to cart — doesn’t always correspond to the micro-event that’s being tracked. If there’s a box with a button inside of it, you may not care if the user clicks the box, the button, the text in the button, or something else. From your perspective, those all correspond to adding an item to cart. With GA4, you can’t combine these. This makes it difficult to query the actions you care about.
Integrations and customizations in GA4 are limited. Period. Users will always have to interact with middleware. In GA4’s case this is BigQuery, Google’s data warehouse platform.
While a free connection to BiqQuery will be included, you still have to pay for the data you use.
This is a problem if you’re trying to make your data useful. We think it’s far more useful for companies to have endless direct integrations, without limitations on their data.
AI and Predictive Insights
To this one, we have to drop a big “really?” We suggest you do too.
We spent some time playing with GA4. Here’s what we found. GA’s “AI and predictive metrics” are basically alerts. You can set up events and parameters ahead of time, and GA will notify you about changes. You still have to manually define the conditions to detect trends that you deem important or critical.
Here’s what GA4 doesn’t do: it doesn’t analyze hidden behaviors to find new insights in your data. Ever. To be fair, GA’s results are automated, so they may save you some time. But they don’t really reach the level of true insights.
We’ve long believed that the future of insights rests in data science that looks through your data to find things you didn’t already know. That’s not the direction GA is taking.
The modern digital world increasingly runs on massive data streams. We were hoping that GA4 would acknowledge this and build in robust governance protocols. Unfortunately, that wasn’t the case.
GA4’s governance tools are built for a simpler era. There are no automatic alerts when events go inactive, no dedicated workflows for repairing events, and no enforcement of naming conventions, which can lead to a giant mess of overlapping events. When a product change causes an event to break, you lose data in the downtime.
In short, GA4’s governance capabilities assume you’ll only be handling a small number of events. We think that this will be a problem for most companies. The future is very clearly moving towards a model where companies collect large amounts of data and sift through it to find the information they need. For this, GA4 can’t really help.
Here’s the really bad news: WITH GA4, YOU HAVE TO START OVER FROM SCRATCH.
While Google is advertising GA4 as an upgrade, it’s really a whole new architecture.
Because of this, switching to GA4 isn’t really a migration. It’s a new, difficult, and time-consuming installation. For most companies, this will be a costly endeavor. SAdmins and teams will need to live in both applications until GA3 is discontinued—setting up dual tracking, choosing which events are important, then organizing and re-tagging everything on your site. Some businesses estimate this could take years.
So it’s worth considering whether GA4 is worth this time and effort. Making the transition means you first have to set up GA4, then set up Google Tag Manager, then tag (or retag) every event on your site…and then wait for new data to be collected…and then cross your fingers and hope you’ve tagged the right events.
If you value time and effort, we’d recommend a solution that does all this work for you, ideally by automatically capturing all the data from your site.
Even we can admit that GA4 is better than Universal Analytics. And we’re happy to see Google validate the approach we’ve been taking for years.
But ultimately, we think GA4 is the right product for the wrong era. Had it come out seven years ago, we could see how many teams might have gotten excited. Now, however, it’s a relic of an old, misguided approach to insights.
To learn more about how Heap helps teams get the insights they need to move their businesses forward, feel free to schedule a demo anytime.
To learn more about GA4, read our complete guide.