I've been interested in cohort analyses for a while, but it's always felt like the "meta" coding that I don't have time for. I know, I know. For those of you w

I've been interested in cohort analyses for a while, but it's always felt like the "meta" coding that I don't have time for. I know, I know. For those of you who run a cohort analysis on your startup, do you generate your cohorts dynamically every time you run the analysis (using created_at dates to group them into months or biweekly segments)? Or do you use some sort of data warehousing to create static data sets at the end of every day / week / month?