topic Re: % split node (after optimization) in SAS Customer Intelligence

% split node (after optimization)

Gerda — Thu, 25 May 2017 07:39:10 GMT

Good morning .. Please help

In my diagram, have an available contact base of 272,251 customers

and I'm using a percentage split node 34% and 66%

In my campaign group, I have specified a required count on campaign level of 60,000

I also have a cellsize constraint of 60,000 in my scenario on campaign_cd

I need an end result (after optimization) with a total of 60,000 customer that is split accordingly to the above specifications ?

Re: % split node (after optimization)

LinusH — Thu, 25 May 2017 10:54:04 GMT

Are you BY any chance using CI Studio or any other of product in the CI/MA offerings?
Also, what have you tried so far?

Re: % split node (after optimization)

Gerda — Thu, 25 May 2017 11:44:55 GMT

hello .. yes, i am using Customer Intelligence Studio, Version 6.3

I have tried to limit my campaign's overall cellsize to 60000 in the Campaign Group

I have also added a Contraint in my senario but it doesn't seem to work

I have attached a document (first message)

Thanks for your help 🙂

Gerda

Re: % split node (after optimization)

Dmitry_Alergant — Mon, 29 May 2017 17:12:25 GMT

What exactly are you trying to achieve - do you want these 60000 to be split between these two cells as 1:2? Since you know the total count, you also know that this will mean 20.000 and 40.000 counts in the cells. You can enable two cell-level constraints instead of one campaign-level aggregate constraint.

Re: % split node (after optimization)

shill — Tue, 30 May 2017 04:39:11 GMT

This functionality is being considered for a future release, but for now the only reliable way to accomplish it is to have a post-process node handle it in code, or put a linkable cell node after the optimized communication node and do your split by percentage in a follow-up (non-optimized) campaign (link > split > cell/cell > comm).

If you try to force the logic into your optimization scenario, you'll either:

have potentially biased results (since the higher value customers will likely go into the larger cell, since that will result in a higher objective value)
get an undesired result (if the input size is < 60,000 total and you specify 40,000 and 20,000 respectively as maximum constraints) since there's no guarantee the split cells will have the correct ratio
have infeasible constraints (if your input size is < 60,000 and you try and specify 40,000 and 20,000 respectively as minimum constraints for two other cells)
have to frequently maintain the results (if the "main" cell size changes from 60,000)

Re: % split node (after optimization)

Gerda — Tue, 30 May 2017 06:11:58 GMT

Thanks 🙂