Hey, what does a manual data source mean in experiment set up? We have a use case that we would like to try. Instead of having an existing experiment_id that links to the data source, we have a set of user_id that are in the buckets and want to see the experiment result based on the manual bucketing <- is there a way to analyse a test with this sort of set up?
f
future-teacher-7046
02/18/2022, 3:04 PM
With a manual data source, it will ask you to input the count, mean, and stddev for all the metrics manually instead of pulling them from your data source.
future-teacher-7046
02/18/2022, 3:05 PM
Probably not what you want for that use case. Is it possible to insert those user ids into the experiment table so it looks like a real experiment? Or is that not feasible to do?
g
gorgeous-caravan-72440
02/18/2022, 4:16 PM
Not directly to the experiment table, but we could potentially find a workaround, like rather than writing directly to the experiment table, we can have a union query on another "pretend to be real experiment" table that have these user_id
gorgeous-caravan-72440
02/18/2022, 4:17 PM
Just wondering if there's a cleaner way of doing it instead of a workaround
f
future-teacher-7046
02/18/2022, 4:57 PM
how many user ids would it be? Like would it be feasible to have them in a giant
IN(...)
clause in a SQL query?
future-teacher-7046
02/18/2022, 5:00 PM
Adding a union to the main experiments query would work, but would definitely be a little ugly.