Hi! I want to count the CR metric from start of an event to success (let’s say event_start and event_success). Numerator - users who clicked on event_success after (!) event_start. Here I created usual binomial metric using SQL query Denominator - users who clicked on event_start. Here I have proportion metric from Fact table I tried to add denominator to my numerator, but I don’t have this option. How can I divide Numerator by Denominator in my situation? Maybe I need to change my metrics’ setup somehow?

You need to set the denominator as a fact metric, not as a "classic" metric using the old flow. Go back to edit your Fact Metric and change the metric type from

Proportion

to

Ratio

If I change my denominator (which is users who clicked on event_start) from

Proportion

to

Ratio

, I'll have to create a denominator for it, so the logic of this metric will be lost Perhaps you mean that I should change my numerator from a "classic" metric to a fact? Then I don't think it's possible, because I wrote it in SQL and there's logic about the time of two events being one after the other

I see, you want to make sure the time is after the other and treat it like a funnel metric. Yeah, so this isn't well documented right now, but pure funnel metrics like this will need to be set up only using classic metrics, not using fact tables. In the fact table you can get the

# users with event_success / # users with event_start

but we haven't added the logic to have the numerator always follow the denominator in time,.

Got it about Fact metrics 👌 So I need to count people who had success after start divided by not all users in the experiment, but only those who clicked on start (this's not really correct in terms of a/b testing, but sometimes we can't split users on 0 and 1 variants on the desired page, so we split users on the most closest page we could). What can I do in this case?

In this case you need to create two classic metrics. First create a classic binomial metric for

event_start

with some conversion window. Then create a second classic binomial metric for

event_success

and pick that first metric as the denominator.