discover-labs/dictum

Rolling aggregations

roveo opened this issue · 0 comments

roveo commented

Support metrics defined as rolling aggregations, e.g. there's "Number of new registered users" per Month vs. "Total userbase" per Month, meaning number of users in the database at the end of the month.

Without the time dimension, they work the same way as total table transform. With the time dimension, they are calculated along the time dimension, partitioned by any other dimensions.

Is this even needed? Counterintuitive? Maybe just support the necessary transforms?
Generally it looks like Tableau's LOD expressions, but the problem is that Tableau workbooks are specific and metrics are generic. And even then LODs are a mindfuck.