PEGACPDS88V1 practice question 16 of 25

The purpose of regular inspection is to detect factors that negatively influence the performance of the adaptive models and the success rate of the actions. Which two issues should be discussed with the business? (Choose Two)

Full set: 140 questions
Answer options
A: Predictors with a low performance_________
B: Actions that have a low number of responses
C: Actions that are offered so often that they dominate other actions Correct answer
D: Predictors that are never used
E: Actions for which the model is not predictive Correct answer
Correct Answer:
C. Actions that are offered so often that they dominate other actions; E. Actions for which the model is not predictive
Explanation

Regular reviews with business stakeholders should focus on strategic issues that impact business outcomes and may require changes to the business strategy. An action that dominates others (C) indicates an unbalanced portfolio. This can lead to customer fatigue and prevents other, potentially valuable, actions from being presented, thus starving their models of learning opportunities. This is a strategic issue requiring business input on contact policies and arbitration. An action with a non-predictive model (E) represents a failed strategy. The system cannot intelligently determine the right audience for the action, leading to wasted impressions and poor ROI. This requires a business decision to either retire or redesign the action.

Why Incorrect

A. Predictors with a low performance: This is a technical finding for a data scientist to address, typically by removing the predictor. The business is concerned with the overall model outcome, not the performance of its individual components.
B. Actions that have a low number of responses: While a critical issue, this is often a direct cause for a model being non-predictive (E). Discussing the non-predictive model (the ultimate outcome of the problem) is a more direct way to address the failure with the business.
D. Predictors that are never used: This is a technical data hygiene issue. It should be handled by the data science or administration team as part of model maintenance, not as a strategic business discussion point.

References

1. Pega Academy, Certified Pega Decisioning Consultant (CPDC) 8.8 Course: The module "Monitoring adaptive models with Prediction Studio" explicitly instructs users to monitor the Performance (AUC) of models. A model that is not predictive has an AUC value around 0.5, which is a primary indicator of failure requiring attention. This supports option E.

2. Pega Platform 8.8 Documentation, "Monitoring adaptive models in Prediction Studio": This official documentation states, "The Models landing page provides an overview of all the adaptive models in the system. From this page, you can identify any models that require attention, for example, models with low performance..." This directly supports option E as a key issue to identify.

3. Pega Platform 8.8 Documentation, "Monitoring and analyzing Next-Best-Action": This documentation emphasizes the need to monitor the distribution of actions. It states, "Monitor the distribution of actions to ensure that one action is not dominating the others, which can cause customer fatigue and prevent other models from learning." This directly supports option C as a critical issue for discussion.

Scroll to Top