Skip to Content
0
Former Member
Sep 22, 2005 at 12:36 AM

Follow-up 3: Hmm! An issue ... u0093How to handle Forecasting when modeling

18 Views

Any comment on my Summary:

-


My understanding is that, in dealing with forecasting during modeling

a) One approach will be to create a new Cube, infobjects for Forecast data in addition to the cube for the actual data. That makes it 2 Cubes.

Advantage:

if you have huge amount of data having several cubes may serve as a kind of physical partitioning - improved performance.

Disadvantage

To compare the data in the actual and forecast, the two Cubes will have to be united with a Multiprovider; more DB space because of redundant storage of characteristics in the similar cubes.

b ) a Second solution will be to keep to a single Cube, i.e combine the actual and forecast, and differentiate the actual and forecast with a new characteristic (like 0Version) and set different values for actual, plan, forecasted data.

(By the way, will we have to create a new dimension for this additional characteristic)?

3

Also on this message “...in other words, if you have a lot of data, go for a multiprovider, otherwise you can manage only one provider with a char that aims to separate different views...”

Does “go for a multiprovider” mean create 2 Cubes and use multiprovider to unite them?