OK I´ve learnt that the fact tables should have measures all of the same
granularity. It's just that in the real world, the measures are not all of
the same granularity. I've f.ex. measures that are on the shop/product/day
level and other measures that are of shop/season granularity. How should I
go about designing such solutions?

Designing one fact for each set of granularities, creating cubes on thop of
them and joining them through virtual cubes?

Making aggregations of one base fact table, creating cubes on top of those?

Or heck, why not just keep on repeating the values of the measures of higher
granularity, and agregate them by the max value in the cube...

Well, I'm sure you MD modelers out there have taken decisions already on
these matters so all I'd like is some advice or sharing of your experience.
Thanks a million.

Tore