Hi, I’m just about getting started with GrowthBook to evaluate it, mainly the analysis part.
I have a question right away about the Experiments Query;
Do we really need to have the timestamp in there? As in; if we have all metric data with user-ids. AND we have a table with user id’s and experiment and variation id, why would we need to have ALL occurances of that user in the test, rather than just know that the user was in the test and analyse the associated metrics?
In our case this will blow up this table,
10/21/2021, 7:53 AM
I believe only one occurrence of the user in the experiment exposure table is all that's needed, but the timestamp is used as the experiment has phases and we need the count per phase (which is time based).
10/21/2021, 7:54 AM
ah great, so ‘first occurance’ or something should work. (I appreciate that you state to not worry about duplicates, but queries/data is getting huge quite quickly ;))