Thanks for the link, I've actually read your article before and thought that it was pretty damn cool. With that being said I believe that we're dealing with two separate issues, unless I am failing to see a way to utilize your solution for mixed granularity to my problem of mixed dimensionality.
In my case I have two separate dimensions. Transaction Channel dimension is completely logically separate from Location dimension, so the problem is not of mixed granularity, but of mixed dimensionality. I'll try to re-state the problem
Our combined fact table has two fields:
We want the filtering selections of Location only be applicable when we use ExtNetDollars for our measures, we want selections of Transaction Channel to only be applicable when measure uses PlanNetDollars. Is there a way to model the data in such way that we would not have to resort to using set analysis similar to the one below:
Being new to QlikView I might be completely off base here, or might be missing the point you're trying to make. In this case, I am sorry. Could you elaborate a bit on the solution you're offering?
You can still use generic keys for solving this problem. Just assign the <ALL> symbol to the appropriate records.
This way a selection in "Location" will not remove any PlanNetDollars. See more on http://community.qlik.com/docs/DOC-3451