abundant-air-78515
02/10/2024, 11:09 AMhelpful-application-7107
02/12/2024, 4:54 PMIn both cases, I apply a delimiter to evaluate the total mass.Denominators (if that is what you mean) are different for cont and binomial metrics. What are you choosing as your denominator?
helpful-application-7107
02/12/2024, 4:55 PMabundant-air-78515
02/12/2024, 7:11 PMabundant-air-78515
02/12/2024, 7:11 PMabundant-air-78515
02/12/2024, 7:12 PMabundant-air-78515
02/12/2024, 7:12 PMhelpful-application-7107
02/12/2024, 7:36 PMIn both cases, I apply a delimiter to evaluate the total massWhat do you mean by this?
helpful-application-7107
02/12/2024, 7:38 PMabundant-air-78515
02/12/2024, 7:42 PMabundant-air-78515
02/12/2024, 7:43 PMabundant-air-78515
02/12/2024, 7:44 PMabundant-air-78515
02/12/2024, 7:46 PMhelpful-application-7107
02/12/2024, 7:49 PMvalue
column. What are you returning as the value
column?helpful-application-7107
02/12/2024, 7:50 PMbinomial
metrics the numerator is 0 or 1 for users, but for count
metrics we sum the value
column. It seems possible from your above queries that you could end up with multiple rows per user_id
which would result in different numerator values.abundant-air-78515
02/12/2024, 8:58 PMabundant-air-78515
02/12/2024, 9:00 PMabundant-air-78515
02/12/2024, 9:01 PMhelpful-application-7107
02/12/2024, 9:14 PMuser_id
level (or whatever identifier type you are using for your experiment). For Count we SUM(value)
and for binomial we just look for any row for a user and count that as 1
. So ideally you shouldn't change your count definition to match the binomial and can just rely on the binomial metric.
If you want to make them match, you could use the "Custom Aggregation" field of simply 1
rather than the default SUM(value)
and then the results should be similar (the statistics won't be, because we model the two metrics differently, but they should be close, and the uplift percents should be the same).abundant-air-78515
02/13/2024, 8:04 AMabundant-air-78515
02/13/2024, 8:05 AMabundant-air-78515
02/13/2024, 8:08 AMabundant-air-78515
02/14/2024, 9:03 AMhelpful-application-7107
02/14/2024, 4:31 PM1
as the User Value Aggregation.helpful-application-7107
02/14/2024, 4:32 PMabundant-air-78515
02/15/2024, 11:55 AMabundant-air-78515
02/15/2024, 11:55 AMabundant-air-78515
02/15/2024, 2:24 PMabundant-air-78515
02/15/2024, 2:24 PMhelpful-application-7107
02/15/2024, 6:45 PM1
as the User Value Aggregation.helpful-application-7107
02/15/2024, 6:45 PMvalue
column in the query can honestly be anything.