Adaptive Model, Weighting of Outcomes?

General
, , ,
1

We can set a response or outcome as Postive or Negative against a given Adaptive model.

But, we have noticed that different responses are of a different importance to us.

Example:

Open of the Offer helps with notifications if our treatment selected is getting a response, helping select that treatment again
Actually completing (converting) the offer is of much more importance on this.

So, looking at the above, Open is of lesser importance, but still has importance than a Conversion which is much more important.

Looking at the data in Pega, there isn’t a weighting on outcomes that I can see. Is there a way to say an Open is .2 and a conversion is 1 for impact to models?

I’m not talking about the current Arbitration for selecting certain actions based on outcomes, this is to train the model.

2

Hi @RobertC57,

You can have more than one Prediction Rule to predict the different outcomes like below

Prediction details with different outcomes are below:

  • Propensity to Click
  • Propensity to Accept
  • Propensity to Conversion

Once you the learning from the Prediction models use one of the Pega CDG OOB extension rules **ActionLevelPropensityExt to implement the custom prioritization (**This is an extension rule for Arbitrate/Augment models rules and defines your own propensities by combining/giving different weightage to the different model outcome).

Hope this helps!

Thanks,

Nanjundan Chinnasamy | Pega Lead Decision Architect | 1:1 Customer Engagement

3

@Nanjundan Chinnasamy actually this does help. That’s a very interesting approach and we’ll look at that one!