Should experimenters and optimizers be worried about GA4?
This interview is part of Kameleoon's Expert FAQs series, where we interview leading experts in data-driven CX optimization and experimentation. Dana DiTomaso is the President and Partner at marketing agency Kick Point.
Q: You’ve written a lot lately about setting up GA4 as painlessly as possible. Why do you think organizations and marketers are leaving this to the last minute? What are the risks of putting this off?
GA4 is a huge change from Universal Analytics, and while the interface has some familiar words, it’s difficult to jump right from Universal Analytics into GA4 without a lot of “where the heck is that” poking around. Even right now, I’m trying to make GA4 my default Google Analytics choice and I keep thinking how much easier it is to find some things in Universal Analytics.
I also don’t think that the improvements have been communicated in a way that speaks to the needs of marketers. Plus the roll-out has clearly been rushed—the July 1, 2023 date caught a lot of people by surprise and now everyone is scrambling to switch to a product which frankly, still feels like it’s in development.
Q: What might product owners and marketers who’ve been comfortable with UA to analyze their optimization efforts not realize about GA4?
GA4 uses an event-based model, and if you’re used to events in Universal Analytics being this little add-on with the category, action, label, you wouldn’t realize exactly how useful this model actually is! GA4 is far more flexible, and since you can multiple parameters per event, you can capture so much more information than you did before, particularly when it comes to understand how visitors are using your website or app.
Q: What’s your process when working with brands that are less savvy with data analytics? What steps do you take so they can start gaining and applying customer insights?
We start by talking about goals instead of analytics. One of the first things we work on with clients is a goal charter—either formally or informally—which helps the client sharpen their focus on what matters most. If we’re working with an in-house team, we may also ask which KPIs or metrics they’ll be evaluated against when it comes to a performance review because we want to be sure that our work reflects well on them.
We’ll also ask for past reports and what kinds of questions people ask them when they’re looking for analytics information. With some clients we’ll also ask them to toss everything they know about analytics out the window and tell us what they’d want to know when it comes to analytics. That can be really illuminating because sometimes not being able to measure something isn’t a failure of analytics, but rather a lack of knowledge that something is actually possible.
Q: You’ve spoken before about how page views can be a deceiving metric. How should you use GA4’s capabilities to assess whether your site solves users’ problems?
Page views can be deceiving because people tend to leave tabs open forever. When you reactivate your browser or view your tabs on a mobile device, that will send a page view off to Google Analytics, even though the page wasn’t actually viewed. GA4 has a new metric called user engagement which shows the time that your web page was actually in the focus, not just hiding in the background. There’s also an engaged sessions metric which measures the number of sessions that lasted a certain period of time (10 seconds is the default), or converted, or viewed 2 or more pages.
Q: One problem for CROs is that experimentation platforms & tools may show different data than GA. How should organizations ensure a single source of truth when A/B testing?
It’s important to realize that no individual product is going to give you perfect data — there’s just too much that can get in the way. People use ad blockers, a Chrome extension could interfere, or they could be using a privacy-focused browser. That’s just a few examples. Instead of focusing on absolutes, treat the data like it is a trend.
Q: What are the most common mistakes organizations make when trying to use data to understand their customers? How would you avoid them?
What I just said about experiences is also true here. You can’t measure everything, and in fact you’re already likely making decisions based on incomplete or inaccurate information. First off, understand where your data is flawed. For example, what percentage of your website audience uses a browser with Intelligent Tracking Prevention (ITP)? You cannot trust multi-touch attribution for ITP browsers, and on top of that, multi-touch attribution also assumes that people are using the same devices or are conveniently signed into a Google account on all their devices. A perfect attribution path rarely exists! Also, make sure that you understand what the metrics you’re reporting on truly mean. I’ve worked with clients who were reporting on New Users and Bounce Rate in Universal Analytics for years, but then were horrified to learn what these metrics were actually telling them. Take the time to learn the platform before reporting on metrics from it.
Q: Is there anything else not covered above that optimizers need to know about GA4?
I have three last thoughts:
1. I’m seeing a lot of marketers re-using the same Category, Action, and Label structure for their GA4 events. Don’t do that! You can structure them however you want and you can have the event parameters make more sense than just carrying over the restricted configuration from Universal Analytics.
2. Get GA4 running on your sites ASAP. Even if you don’t do anything to add extra events, just get the code on your site so you can start collecting data. July 1, 2023 is coming sooner than you think and you’ll want that year over year data ready.
3. Take a course on GA4. As I said earlier, Google hasn’t done a great job of education. I have two courses out on LinkedIn Learning that cover Migrating from Universal Analytics to GA4 and Advanced GA4. Krista Seiden also has a great set of courses, including a technical deep-dive which is a must if you’re doing anything with ecommerce.
