Hey bit of a weird question but is it possible to analyse A/B experiments on request level? Scenario: we have our own A/B experiments set up via traffic shifting but instead of sessions or users we solely run experiments on request level, would it be save to substitute user_id with request_id in that case if we only want to onboard growthbook for experiment analysis?

In the data source setting, you can define your own identifier types. By default we use

user_id

, but you can change that to

request_id

instead if you want. Then, update your assignment queries and metric queries to reference the request id and it should work as expected

Ty for answering @future-teacher-7046 Other than the renaming, is there anything obvious which we have to look out for or would lead to wrong evaluation if we'd adopt in that way on request level?