To date calculations with roling calendar

Question

Hello,I am relatively new to Essbase and we are in the middle of building an inventory analysis database. We have 5 dimensions:Accounts,Time,Version,Customer,ProductThe time dimension spans from Year-Season-Quarter-Month-Week. Includes annual totals for 4 years. I have a rolling calendar and I do not use a scenario member to designate the year. I also have an alternate roll-up representing a fiscal year.My time dim looks somthing like this:TIME..Y2001...Fall....Quarter1.....Jan01......WK 2003-10-13..Y2002...etc, etcI am trying to come up with a YTD, STD, and MTD number for base sales_unit data. I should be able to get that number no matter what period selected in a report. For example if I am looking at Week 10-13-2003, I need to see what my MTD, QTD, and YTD numbers are for that date-member, same goes for upper level members in time. I can't use DTS because it only (seems to) work for the generation it's assigned too. Furthermore with DTS, I need the date results for my average inventory calc too.So the only way I could get this to works was with a VERY long series of if statements for each time member.Here are a few examples: For Month: IF (@ISMBR(Month01-2001)) @SUMRANGE(Sales,WK2001-02-10:WK2001-03-03); ENDIFFor Quarter: IF (@ISMBR(Quarter1-2001)) @SUMRANGE(Sales,WK2001-02-10:WK2001-05-05); ENDIFFor Season: IF (@ISMBR(Fall-2001)) @SUMRANGE(Sales,WK2001-02-10:WK2001-08-04); ENDIFThis works but data retrieval times are too slow (16 minutes) for dynamic calcs and the calc script runs for more than 6 hours.It's obvious to me that this will give me the correct results but it will not work because of performance. Any suggestion on how to deal with a to date number with reasonable calc times using a rolling calendar would be graetly appreciated!Russell Krauss, VF Inc

Answer 1

I've run into similar problems and one thing I have given into is that there is no one solution. What I have been using is multiple alternate roll-ups using shared members. My Time dimension will have three or four sub-roll-ups (i.e. "By Month", "By Week", etc.) I do this by sharing the base level members (daily data). I'm not sure if it would be possible to do what you are looking to do in one roll-up because weeks will inevitably over lap months. I will check back however to see if anyone else has suggestions. One more thing to consider, if your time dimension is sparse (which is probably not a good idea in this case) but if it is, you could use attribute dimension to group some of your dates, particularly the seasons. Good luck!