My question is about Calculated fields and GA4 data attribution. I'll paint the case first - it might be a simple example that might be solved in a simple way, but the problem is virtually unsolvable in more complicated cases.
Let's say we want to get results for sessions containing certain URLs as Page Paths.
As a simple example, to do this, we want to group URLs into 2 categories: CatA and CatB with a calculated field, let's say 15 URLs in CatA and 15 other URLs in CatB.
We would do a CASE statement with the dimension of Page Path (event-based) - when Page Path is the first 15 URLs, it returns CatA, when Page Path is the other 15 URLs, it returns CatB. Everything works so far.
Now we want to calculate the amount of sessions containing CatA URLs and CatB URLs. So we create two statistics summary visualizations and create two filters - one for "Calculated field equals CatA" and one for "Calculated field equals CatB". We plug these filters into the summaries and voila - we have the sessions. But the results are incorrect.
How do we know that? When we create two filters that have direct regexes with Page Paths for the first 15 URLs or the other 15 URLs instead of using calculated fields, the results are different. If we recreate the same logic in GA4, the results are consistent with these direct filters results and not those where we used calculated fields.
The problem in my opinion is scope, and in this case, sessions recalculation. If we use "Landing Page" instead of "Page Path" in the calculated field, the sessions calculate just fine, because the scope is sessions vs. sessions. So because our calculated field was event-based and we want sessions amount, Looker takes the resulting rows from the calculated field in a unspecified way (I have some ideas, but not very certain of any of them) and returns a false result.
In this simple case, of course we can use direct filters. But in more complicated examples or dynamic filters, we encounter a real issue.
So my question is - is this problem something that the Looker Studio team is aware of and willing to work on? Or is this more of a problem of GA4 attribution? This is directly a conflict of Looker's calculated fields logic and the GA4 connector so I'm not sure where to ask about it.
Any comments, suggestions or updates are welcome!
LinkedIn
Twitter
