hey there. was wondering if you all have any advic...
# ask-questionsf
fancy-shoe-71869
09/13/2024, 4:39 PMhey there. was wondering if you all have any advice on adding LTV metrics to Growthbook? we want to be able to keep track of what happens to users even after the experiment concludes. my understanding is that the event needs to happen between the assignment and the experiment end date.
one idea was to create an "event" on the dbt side that fires right after the user assignment. it'd be like handing each user a cardboard box after their first experiment assignment and then continuing to fill that box with metrics even after the experiment concludes. we could then compute metrics like 30D, 60D, 90D LTV and understand potential long term impacts.
f
fresh-football-47124
09/13/2024, 7:25 PM
You can have events that happen after exposure
fresh-football-47124
09/13/2024, 7:26 PM
you can adjust the window
fresh-football-47124
09/13/2024, 7:26 PM
even after the experiment has ended
fresh-football-47124
09/13/2024, 7:28 PM
Screenshot 2024-09-13 at 3.27.57 PM.png
fresh-football-47124
09/13/2024, 7:29 PM
you can even override them per experiment
f
fancy-shoe-71869
09/13/2024, 8:01 PMinteresting. wasn't aware of the conversion window grabbing events from after the experiment end date. looking at the queries, it seems like it filters for timestamps prior to the experiment end date even after overriding the metric window
f
fresh-football-47124
09/13/2024, 8:03 PM
Is that for the experiment exposure events? Or the metric events?
f
fancy-shoe-71869
09/13/2024, 10:03 PMthat's for
__metricf
fresh-football-47124
09/17/2024, 8:08 PM
we have two solutions:
1. Create a custom report (ad-hoc report) where you delete the end date (so that the experiment metadata isn't screwed up).
2. Keep the experiment report running (long running experiment - you can stop assigning into it if you need to stop)
fresh-football-47124
09/17/2024, 8:15 PM
Three options, probably in order or preference:
• Use a very long conversion window. This will work well, but of course we need to scan the full metric source over the full window, so if you do this for 90days and the metric source is large, that query just gets bigger and bigger.
• Do you suggestion. This will also work well, but it can be harder to interrogate where the data are actually coming from once it gets in the "cardboard box"
• Use an ad-hoc/custom report and delete the end date of the experiment (this will change the window for all metrics if they don't have conversion windows) and give you long term effects for all your metrics.
f
fancy-shoe-71869
09/17/2024, 8:25 PMoh that's interesting. didn't realize you could delete the end date on a custom report. this is helpful. thank you! cc @modern-waiter-38896
38 Views