Hi.I'm tring the metric function. Is that mean I need mantain a viewed_experiment table to record user hit which variation on what experiment?

If I understand you—that's right. Usually, in your database, you'll have a table that maps a user to the experiment and variation they saw. This is often referred to as an "experiment assignment table" or "experiment exposure table". Is that what you're asking?

Hi @strong-mouse-55694, yes, I think I should record these infomation. But, users hit on experiment's variations are based on rules(hash user id and map to variation id is also a type of rule) we defined in gb. Why not make it transparent to GB users? That I don't need to record these relations in business code(log and extract to HIVE).

I'm not sure I understand what you're getting at. Can you say more?

are you talking about pre-bucketing or assigning users?

Nope. I think the "experiment assignment table" or "experiment exposure table" @strong-mouse-55694 mentioned should be record by GB rather than users. All the rules/exp setup and changes are recorded in GB, and the traffic hit on which exp and which variation is known(because is calculated).

Thanks for the feedback. Was there anything else you had a question about?