Semi-additive measures, multiple measure groups and time dimensions
My last post was about the effects of semi-additive measures on dimensions other than the time dimension. Now let's focus on the situation of having more than one dimension of type 'time' in our cube.
As we've seen from many posts this may result in our aggregations are not what we would expect. The effect becomes clearly visible with semi-additive measures (although you should always set the type of your time dimension in order to provide useful information to the OLAP-clients). Only the first dimension is treated as a time-dimension while all following time dimensions are treated as normal dimensions. If, for example, a specific measure is set to aggregate as 'average of children' this will only consider the first time dimension as the one to compute the average for.
Tags: design, time dimension