Tightening the Screws on Your GA4 Infrastructure
Tell me if you’ve heard this story before: an organization was bad at data, so they rallied the troops, found the budget, invested in a new setup, generated excitement, used the data…and after months of using the data, they found out that their data was broken.
It’s happened to me, and I’ve seen it happen to others many times. It’s awful. Not only can this cause frustration (“we’ve reported on inaccurate data”), it can lead to a devastating false-belief that we just can’t get analytics right at our organization (“we’ve been burned before and we’ll be burned again”).
Part of working with web analytics is coming to an acceptance that data won’t be perfect and that things occasionally will break. There are so many things that can impact your setup that naming them all would be its own blog post. Some of the many examples that we’ve seen include…
- Google Tag Manager (GTM) getting dropped from the site
- A consent management platform (CMP) being added
- A page launching with a new URL structure
- A new video player being embedded
- A redirect being employed and breaking attribution
…the list goes on! The website evolves, and sometimes your analytics setup doesn’t evolve at the same rate–and that’s when your data can break. It’s incredibly annoying, but it’s salvageable, as long as you catch it quickly enough.
Musicians regularly tune their instruments and replace heavily used parts like strings or valves. A garden doesn’t stay beautiful in perpetuity—it is weeded, mowed, fertilized, and sprayed for pests. I send my bike to the shop to be serviced every year. We go to the doctor for annual physical checkups. When I worked with IKEA at my previous job and complained to my German collaborator about the furniture always falling apart, she incredulously asked, “Well you tighten the screws every three months, right?” (I had no idea that was a thing. Do any Americans do this?)
Most things we rely on require periodic upkeep, and so does analytics. Your GA4 solution cannot follow the “set it and forget it” model.
