Tabular Model Both Single Grain (cross filter) vs Multiple Grain (M:1)best chosen based on User Query
Modeling Kimball conformed stars into one semantic model requires all facts to dimension joins be M:1. even the Junk/drill to detail dimension has to be M:1 from fact to dim. This works exactly as expected to support drill across. Single grain model for one data mart can support cross filter for everything so the use can choose attributes without any metric and is very user friendly. Could we model each data mart as single grain cross filter AND as part of the one larger semantic model and then at run time, PBI uses the single grain when only one measure group metrics are in the report and use the Larger semantic model when a user has multiple Measure Groups involved. best of both worlds user experience. Still have one unified semantic model with dimensions conformed across all facts but provide the user friendlier query experience when only a single grain is in scope.
This would make building reports so much easier.
We run into this scenario daily and would appreciate consideration your ability to support this model