The Double Lift challenge solution

Document created by ggallo Employee on Sep 5, 2017
Version 1Show Document
  • View in full screen mode


A discussion on the traditional challenge to propagate information input at aggregate levels belonging to the same hierarchy, down to the granular level.


1   The Challenge

1.1     What is the Double Lift Issue?

Let’s consider the following challenge: a Plan by Channel and Salesman should drive a Plan by Customer. So users would plan at a higher level and then BOARD would allocate and propagate down to the lower level the plan.

Channel and Salesman are both aggregations of the entity Customer.

The Cube A (by Salesman and Channel) must be propagated to the Cube B (by Customer).



1.2     Version Compatibility

BOARD 10.1 and following manage the above situation under the following conditions:

  • The dataflow should be a simple copy of the amounts from the aggregated level to the granular level (b=a)
  • The dataflow should be performed using the method HBMP or HBMP+


for earlier versions than BOARD 10.1 the solution proposed requires the definition of a [R] Replicated Entity and the process below


1.3     DATAFLOW Schema

Let’s consider this Data Flow Schema and how it drives the Data Allocation



1.3.1 Step 1

Thanks to the E/R Model each Customer is given the correct value, but the salesman[R] dimension isn’t recognized as belonging to the Customer Hierarchy. Therefore, per each Customer, all the salesman[R] items are populated. The criteria is given by the relationship between the Customer and the Channel.


The adoption of an [R] Entity is only to guarantee the item alignment with the Hierarchical Entity (Salesman). A standalone entity could be used as well.


1.3.2 Step 2

Now the Diagonal Matrix (Cube C), thanks to the Relationship with the Salesman Dimension, identifies which item of the Salesman[R] dimension should be kept per each Customer


1.3.3 Step 3

Finally, you can copy the Cube D into a cube by customer only (the salesman [R] dimension is collapsed) as per our initial requirement.




Attached you can find an example illustrating the process needed for BOARD versions earlier than BOARD 10.1

3 people found this helpful