Appendix A: Target Scenarios
To get accurate predictions, the accounts need to have at least twenty-four months of historical data. Historical data should be set up in dense cube views. Sparse data is not recommended as Predictive Analytics cannot use sparse data to make time series statistical predictions.
This solution is meant to be used interactively and not in batch mode. Predictive Analytics can only run one template at a time and cannot run multiple templates at once in a batch. There is no built-in functionality to automate running templates for Predictive Analytics.
NOTE: It is recommended that users not exceed 300 targets in a Predictive Analytics template, as going beyond this value will result in long prediction run times.
-
One Sales account by one Entity by one UD1 member by one UD2 member across twenty-four time periods is one target.
-
One Sales account by five Entities by one UD1 member by one UD2 member across twenty-four time periods is five targets (five Entities)
-
One Sales account by five Entities by five UD1 members by one UD2 member across twenty-four time periods is twenty-five targets (five Entities * five UD1 members)
-
One Sales account by five Entities by five UD1 members by five UD2 members across twenty-four time periods is 125 targets (five Entities * five UD1 members * five UD2 Members)
An example use case is written and shown below:
-
One Sales account by five UD2 members by three UD4 members across 12 time periods is 15 targets (5 UD2 members *3 UD4 members)
In this scenario, Operating Sales is the account, the type of ball is the UD2 member, and the store is the UD4 member. The final row is the column total and is not included in the target number.
NOTE: If your use case calls for making automated predictions or utilizing larger volumes of data (greater than 300 targets), it is suggested that you consider Sensible Machine Learning for your predictive analytic use cases. Sensible Machine Learning has the ability to handle a larger amount of targets along with a number of other features.
NOTE: The example shown is from one of the GolfStream applications. Your results will differ depending on your historical results and the configuration of your prediction template.
