Ok, i've changed the model to have intersction as fact, and now i get "The specified criteria didn't result in any data" when doing the simplest analysis (count by type)

Hmm, that could work, will try this approach.

Hmm, the only fact i'm actually interested is the count of activities for a given combinations of dimension. I'm not sure i can get this count with S_ACT_EMP as a fact table ?