do i need to join in the users from the experiment into the query so i only get users from one of the variants into the guardrail metric, or what is the best practice

the exposure queries and metric queries should be written generically without referencing a specific experiment or variation.