My first thought would be to calculate the average or median external to metabase and put the result in a table in your database that metabase points to. Although it would be great if you could just do it on the fly somehow.
@sheki — Did you find the answer to trailing / rolling / moving averages? Just strapped up Metabase to our DB and trying to do the exact same thing for our acquisition, activation, conversation, and retention rates. Will be working on it tonight and can update if I figure it out. I’m using the “question” interface, not SQL.
The windowing stuff will not be ready for .33, while sum-where & a couple of other new aggregations and custom expression functions will be.
As for the original question: we now have trendlines which in some cases might be good enough smoothing, else if you want something like weekly daily average, first create a question where you group by day, then a nested question off of that where you group the days by week and average the daily averages. Of course that will give you pseudo-smoothing only in the case of avg and distinct.