https://www.growthbook.io/ logo
b

billowy-horse-43368

03/14/2023, 8:33 PM
How could I add a metric/setup an experiment to understand D1 or Dx retention for a specific metric?
h

helpful-application-7107

03/14/2023, 9:05 PM
It depends on how you want to measure retention. Currently in GrowthBook you can do either of the following: 1. Check for "retention X days after first experiment exposure". This can be done by having a metric that stores the date and user_id for each day that a user was "retained". Then in GB you set up a
binomial
metric with a conversion delay of
24*X
and conversion window of
24
hours. So D0 would be "was the user retained as per this metric in the first 24 hours after the experiment". D1 would be "was the user retained as per this metric between 24 and 48 hours after the experiment". And so on. 2. You could easily generalize the above to check for ranges of time after a delay. For example, you could look at retention any time in the second week after exposure, by setting the delay to
24*7=168
and the window to
168
hours. Then you could say "what % of users in variation B were retained as per this metric at least once, one week after their first experiment exposure".
If you want retention defined as per some date other than the user's first exposure to the experiment, then you'll have to pre-compute this retention yourself and use a large conversion window to just capture all post-experiment data (although we're working to improve this experience so that you don't need to set a large conversion window).
b

billowy-horse-43368

03/17/2023, 10:16 AM
thanks @helpful-application-7107!
4 Views