3 Replies Latest reply: Jun 18, 2013 5:54 AM by Henric Cronström RSS

    Best way to deal with mixed dimensionality

      What is the recommended way to deal with mixed dimensionality in QV? Presuming I have a Location, and Transaction channel dimension. Now I have a combined "Facts" table that contains both, Plan, and Sales data. Sales data is slicible by Location, but not by Transaction channel, Plan data is slicible by Transaction Channel but not by the location. Next, I define two metrics, one for Plan and another one for Sales, but I don't want filtering selections of Location dimension to be affecting Plan metric, and I don't want Transaction Channel filtering selections to be affecting Sales.

       

      I know that I can achieve desired results by using set analysis and negating any selection made for an inappropriate dimension and that's easy enough to achieve (if you only have to metrics with mixed dimensionality). What's concerning to me is that intuitively, I sense that there's a CPU load and cache footprint overhead associated with putting this logic inside of set analysis. Is there a better way of dealing with said scenario?

       

      P.S. Example above is put together for demonstration purposes only, actual application we have uses tens of metrics with various dimensionlities.

       

      Thanks for taking the time!