This message was deleted.

Hi Ahmed, One, we don't recommend having ids that are that invariant. Two, metrics define what identifiers they return (you set this). If an experiment is set to do analysis on an identifier that is different from one that the metric is returning, then we'll use the idenfitier joins, which is defined on the data source page

Hi Graham, thank you for your response. I think we need at least two identifiers (for guests and logged-in users). The third one was needed for some targeting conditions. You were right I can set the identifiers with metrics, and I realized that this should be the only correct way to conduct analysis. The issue arises, however, from having metrics that uses two identifier, but the the analysis is done on a third identifier. Then, as you said, you will use the identifier joins and here we have inconsistency, it uses one metric identifier, e.g. exp_id <-> GA4_id, with one metric and uses another identifier, e.g. exp_id <-> user_id with another metric. I agree with you now that this is considered a misuse of the identifiers, but wanted to clarify it here anyway. Thanks again and have a great day!