Statistics Cube Overfeeding Issues. Alternative Solutions?

[michaelc99]

Good Morning/Afternoon All,

I am still fairly new to TM1/PA, but I am attempting to create a Stats Cube to replace Excel-based reports which have added calculations, like DPO, DSO, etc., that were previously not in TM1. My initial thought process was to keep these calculations separate from our main GL MAIN cube structure. So, I created a Statistics Cube that mirrors our GL Main cube, but using a blank Account dimension called "Account Stats" that now contains about 50-specific members and expected to grow over time.

I initially went the Feeders route, but I facing the classic problem of underfeeding v. grossly overfeeding. In some attempts, I found an 8-10 GB feeder file, which takes 8-12 hours to calculate. I searched for overfeeding tests and found the below solution to evaluate overfeeding. Basically, duplicate the overfed cube and add the code (shown below.)

Code: Select all

# This Code is located in our Stats Rule file

[] = DB('STATISTICS - Overfeeding',!VERSION MAIN,!REGION MAIN,!ENTITY MAIN,!CURRENCY MAIN,!DEPARTMENT MAIN,!COST CENTER MAIN,!CHANNEL MAIN,!PRODUCT CENTER MAIN,!SITE MAIN,!ACCOUNT STATS,!PERIOD MAIN,!GL MAIN MEASURE);

# The Below code is located in our Overfeeding Cube

SKIPCHECK;

[] = N: IF(
          DB('STATISTICS',!VERSION MAIN,!REGION MAIN,!ENTITY MAIN,!CURRENCY MAIN,!DEPARTMENT MAIN,!COST CENTER MAIN,!CHANNEL MAIN,!PRODUCT CENTER MAIN,!SITE MAIN,!ACCOUNT STATS,!PERIOD MAIN,!GL MAIN MEASURE)= 0

,1
        ,0);

FEEDERS; 

So, while I wait for the above code to run its' course, I am pondering a few questions:

1) Would it be adventitious to build "Stats" only dimensions for the remaining dimensions to narrow down the members?
2) Would a TI process be more efficient (v. Feeder rules) to copy balances from GL MAIN to STATS?
3) What is the risk of using GL Main to build the custom Reports? Would this be a better solution?
4) Am I missing a better solution?

Thanks in advance!
Michael

[lotsaram]

Google "feederless rules". This is a good solution for this use case.

Basically you make the KPI a C element and the children are the elements used in the calculation. You don't need to feed C rules as they are naturally "fed" by theunderlying data. As almost all such KPIs involve ratio calculations they are perfect for this approach.

And no don't use a separate cube. Just add the KPI elements to your account dimension.

[michaelc99]

Google "feederless rules". This is a good solution for this use case.

Basically you make the KPI a C element and the children are the elements used in the calculation. You don't need to feed C rules as they are naturally "fed" by theunderlying data. As almost all such KPIs involve ratio calculations they are perfect for this approach.

And no don't use a separate cube. Just add the KPI elements to your account dimension.

Thank you! The feederless solution resolved all of the challenges that I was faced. Instead of 32 Feeder Rules from GL Main to Stats, and the several hundred lines of code in Stats, I am now down to 40 lines of code.

Thank you again for your help!

Michael