Hi folks. I posted this a while back but please can anyone help me confirm the following? 🙏 Binomial count: is this UNIQUE users as determined by user_id and/or anonymous_id? So if 10 records are found by the query but they representing 5 users, the count would be 5? Count sum: is this for ALL users? Is the value derived from ALL records even if unique users are duplicated in these? Count average: is the n value for this again based on UNIQUE users? Apologies if already in the docs, I couldn't see quite what I was looking for. I am writing custom queries resembling those used by GA4, which means it essential to understand the above.

Binomial count: is this UNIQUE users as determined by user_id and/or anonymous_id? So if 10 records are found by the query but they representing 5 users, the count would be 5?

I wonder if the language in the Fact Tables is clearer for you, but yes, in this case the numerator would be

5

and the denominator would be the total number of unique users in the experiment. In this case the numerator will necessarily be <= the denominator.

Count sum: is this for ALL users? Is the value derived from ALL records even if unique users are duplicated in these?

Count average: is the n value for this again based on UNIQUE users?

I'm not sure what the distinction between these is in your mind, but our

Count

metric will

SUM

the

value

column in your metric query for each user. Then, in each variation, we

SUM

the user-level sums, and use that as the numerator, and the total number of users in the experiment is in the denominator. So in your example above, if there were 10 records and each record had

1 as value

in the query, then the total numerator value would be

10

.

@helpful-application-7107 thanks so much for the clarification on this. That is as I thought for the binomial case. For count metrics, maybe it's easiest to explain with an example. Say I have a metric that produces multiple values per user: user_id value 1 10 1 5 2 10 2 5 In the results of the metric, would the total be 30 and the number of users be 2 (unique users, as above), and therefore the average 15? I know that in experiments this totalling gets split between the users depending on assignments, but it is just the principle that multiple values per user get summed?

In the results of the metric, would the total be 30 and the number of users be 2 (unique users, as above), and therefore the average 15?

Yes.

Thank you!