Sparsity advisor AWM 10.2.0.3

Question

I have a cube that is created with 6 dimensions. It used to finish with the complete build and refresh in about 20 minutes.
I used the sparsity advisor on it and based on the current data it gave me a different dimensionality and asked me to change one dimension from sparse to dense. I did recreate the cube with those recommendations.
Now the cube build takes more than an hour an counting. It never came back. I had to kill it.
Anyone had this kind of experience? Or is it me only?
The cube has one measure and six dimensions. the cube data source has approximately 1 MN rows. If anyone want more details I can provide. But I would really like to know what am I (or the Sparsity advisor) doing wrong here

Answer 1

The sparsity advisor is certainly not omniscient; it follows a set of heuristics and gives its best guess, but there are certainly adjustments that can be made to it. I can't promise anything, but if there are data sets it does a poor job on you should feel free to forward me information about them and I can consider what sort of enhancements could be made in the future. At a minimum, turning on the parameters
olapcontinuous_trace_file=TRUE
olappoutlog_echo_to_eventlog=TRUE
and mailing the resultant trace file (It should have lots of lines with the string '[SparseAdvise]' in it) to me would be useful. Anything else like a description of the model, source tables, data patterns and especially the dimensional sparsity you chose and why would be extra helpful. If they're not too big I would also be happy to take dump files of any sample schemas which you think illustrate problems.
Being just some code, it's never going to be as good as an experienced OLAP professional in figuring stuff out but we would like it to be good enough that in the vast majority of cases it gives advise which isn't totally out of whack (which it certainly is if it's sending aggregations off into la la land).