This message was deleted.

I'm not sure I fully understand. By "percentage metric", are you trying to do something like "99th percentile latency"? Or something like "percent of users who did X"?

Iam having a experiment where i have the goal metric on number-of-contact-request-sent by unique employer_account_id . In my experiment i want to setup a guardrail metric becuase this experiment can have a implication on the grade of request being accepted by the candidates. So in my world this needs to by a % metric. Now i have setup a guardrailmetric by a sql where i have ainterval in time from the experiment started, to now. the value metric is (number of sent request / accepted request) and i also group by employer_id which is user_id så i get the percentage for each employer account så i can track from which test it bellongs.. Is this correct and best practice for this? The timestamp is gonna be N/A so i write MAX at this line becuase in this interval there are many timestamps..