I have set up measurement for enough companies now to see the same pattern. Someone decides they need better tracking. They hire someone to install GA4. Tags go in. Events fire. The dashboard fills with numbers. And three months later, nobody trusts any of it.
The problem is almost never the tool. GA4 works fine. GTM works fine. The problem is that most implementations are built without deciding what decision the data is supposed to inform.
I see this most often with e-commerce companies. They want to know their ROAS. So they set up purchase tracking. But their “purchase” event fires before they have confirmed the order is not fraudulent. Their “add to cart” event fires on page load, not on actual click. Their attribution window is thirty days because nobody changed the default. The numbers look clean. They are not clean.
The second common problem is ownership. The tracking was set up by an agency two years ago. That agency is gone. The tags are still firing, but nobody knows what half of them measure. There is a custom event called gtm.click.3 that has been recording forty percent of all conversions. Nobody knows what it is. Nobody wants to delete it just in case.
The third problem is that measurement is treated as a technical project, not a business one. The developer installs the tags. The marketer looks at the reports. Neither of them sits down together to agree on what “conversion” actually means for this business. Is it a lead? Is it a qualified lead? Is it a closed deal? If you do not answer that first, everything downstream is noise.
Three questions before any implementation
A proper tracking implementation starts with three questions. What decisions will this data inform? What needs to be true for us to trust this number? Who owns this when something breaks?
Everything else is just configuration.
I have rebuilt measurement stacks for companies that were making budget decisions based on attribution models that had not been validated in six months. The rebuild is never technically difficult. It is just a matter of slowing down, agreeing on definitions, and building the thing properly.
Infrastructure, not a project
The companies that do this well treat tracking as infrastructure, not as a project with an end date. They audit quarterly. They document what every event measures. They know exactly which numbers are trustworthy and which ones are approximate.
That is the difference between measurement that drives decisions and measurement that decorates slides.