This all depends on the structure of your data. Any chance you could share a sample?
Basically I've got one giant table with many columns. The only two that are relevant to the question are "user_id" and "timestamp".
Maybe a simpler way to present the question is by days only (don't worry about hours), and a constant step increment. So something like:
For x in (1-30) , Plot xy, where y = (count of distinct user_id's with timestamp within x days)
Does that make sense? Not sure the most clear way to phrase what I'm trying to do.