https://www.growthbook.io/ logo
#announcements
Title
# announcements
w

worried-waiter-18787

02/09/2023, 10:58 AM
Hi all, looking for some advice on how to implement a metric. I’ve got a custom denominator on my metric, and the timestamp of the events in the denominator occur after the numerator timestamps. What I’m seeing is that the count in the numerator is lower because the timestamp comparisons that Growthbook adds as part of the query it generates look for numerator timestamp >= denominator timestamp (+/- conversion delay) and numerator timestamp <= denominator timestamp + conversion window. I’ve set a large conversion window for both numerator and denominator. I’ve also explored adding a negative value for conversion delay, which does include more of the numerator events, but unless I set it to align with the experiment start time it’s including events from before the experiment too, which is not what I want. Do you have any guidance on how I should implement the metric such that I have a count of all events for the numerator and denominator after experiment start for this setup? Or am I doing something incorrectly which is making this trickier?
f

fresh-football-47124

02/09/2023, 1:45 PM
We have metric place holders
template variables:
that should allow you to use the experiment start date
w

worried-waiter-18787

02/09/2023, 1:59 PM
awesome thanks Graham, I'll give these a try but they sound like they should help me out
just been investigating this further and I'm seeing that when I use '{{ startDate }}' as the filter and there's a conversion delay set, the value thats inserted where the template variable is, is the 'adjusted' value. For example experiment start 7th January 2023, metric has a conversion delay of -24h, it appears as thats inserted as
timestamp >= '2023-01-06'
so it kind of just moves the issue. Is there a way to stop startDate from being affected by conversion window/delay?
10 Views