Why does everything look stable at first glance?
I’ve seen many setups where GA4 looks calm and confident. Conversions are there, attribution is nicely distributed, trends make sense. BigQuery is connected - but barely used.
“Everything matches anyway.” (LOL)
Until someone asks a simple question: *what exactly are we optimizing here?*
Where does reality quietly drift?
GA4 modeling hides a lot:
• missing events
• duplicated users
• reconstructed paths
Without checking raw events, you don’t know which numbers are facts and which are guesses made by the algorithm. Decisions end up based on a reconstructed version of reality, not on what actually happened.
What’s easy to misunderstand?
Modeled data is fine for orientation. It’s dangerous when treated as ground truth. Without raw-level validation, you never see where facts end and assumptions begin.
How I usually handle this?
I regularly work with raw events in BigQuery: volumes, timestamps, event chains. Then I lock rules in writing - which metrics can rely on modeling, and which must be based on factual events only.
Want to get all my top Linkedin content? I regularly upload it to one Notion doc.
Go here to download it for FREE.
