The only analogy I have on my end is a "questionnaire_id". It has one or more questions, and each has one or more choices. So when someone goes through, they may answer one or two or 20 different things. So I have an event for the choice selected. If I just do "total sessions", it picks up frequency per user, but if the user has done it multiple times (separate questionnaires but multiple choice selections across them, it "inflates" the frequency per user calc. So if i do an aggregate on the distinct count of hour per day, it gets a closer to true answer bc the odds of a user doing it multiple times in the same hour is less. I don't have a unique "query_id" being passed, otherwise I would just use that and would probably be the best option.