Question
Cognizant Technology Solutions
GB
Last activity: 10 Sep 2021 9:26 EDT
Offer prioritization in CDH
Hi All -
I am seeing a very unique issue in a CDH implementation for last few weeks . We normally pick the top 3 offer based on Priority . Action level propensity is enabled and the application is in production for 5-6 months now. Because of low uptake of certain offers we decided to boost the Action level weights (Action weighting) massively for this offers so that if they are eligible then by P*C*L*V calculation even with low Propensity this offers will swim to the top . Overriding the propensity trend by this artificial boost was a calculated gamble at this stage.
What we are seeing now is that few customers are bombarded with the preferred set of offers but for quite a few (~ 50%) they are still receiving the ones with low priority and high propensity even after all this boost . If i check from strategy the result is different with priority the key differentiator but run time the picture is different.
Any help will be highly appreciated.
We are using CDH 8.4 for communications .
Pegasystems Inc.
GB
Hi, bit tricky this one to diagnose from that post, few thoughts
You say that run time results are different from 'strategy result'. What is 'strategy result'? Are these tests via the test run panel? A distribution test or something else? What environment are you doing this test in?
It is possible that there is something real time influencing your results, a real time context perhaps? Are you inbound? Maybe call reasons or web page that is boosting certain actions that you miss in the strategy test. How about Contact policies, are these possibly a culprit as reliant on the historical interactions.
I would say maybe your engagement policies could be filtering the boosted actions, but again this would not explain the different results.
There are some other things that could come into play but likely the above is enough for now. To help further it would be good to understand what you mean about the differing results from strategy and run time.
Cognizant Technology Solutions
GB
Hi ,Thanks for your reply .
By strategy result I was actually referring to test via the test panel .
I was also suspecting that real time might have a influence on the result . To mimic it , I put all the contextual data on the data flow while running the test however the output was as expected and what we were hoping for. The contact policies would be triggered even on a test from test panel using data flow when we are testing end to end.
This is indeed on inbound channel and we are using Container REST services to provide the NBAs to calling systems. I will run some more test today and will let you know my observations .
Pegasystems Inc.
GB
@SanBan Ok, that's good, so you are testing the top level strategy via the test run panel, using contexts populated by a data flow in the same environment. The reason I raised contact policies is that sometimes people test in a different environment where the data is not reflective.
Have a go today with your testing and if still no luck read this article and see if model maturity / control groups could be influencing
Pegasystems Inc.
US
@BURTA Bear in mind that model maturity is only applied for Outbound interactions and so should not come into play here if your tests are for Inbound scenarios.