PEGACPDS24V1 Sample Questions & Answers
Checks your knowledge of adaptive model monitoring, the single biggest weight, plus model governance, Pega Process AI, Pega NLP text analytics, Customer Decision Hub predictions, prediction patterns, and MLOps for building and evaluating predictive models.
Launch the full PEGACPDS24V1 simulator →Showing 6 of 12 free samples.
- Question 1Intermediate
Adaptive Analytics · How adaptive models self-learn from customer responses
A data scientist is monitoring a newly launched adaptive model in Prediction Studio. The model is intended to predict the likelihood of a customer accepting a premium credit card offer. Because the model was launched with no historical data, it currently relies on self-learning. What underlying mathematical approach does the Adaptive Decision Manager (ADM) use to update the model's scoring dynamically as new customer responses arrive?
Show answer & explanation
Correct answer: B
Pega's Adaptive Decision Manager (ADM) fundamentally relies on a Naive Bayes algorithm and Bayesian scoring techniques. This approach is highly efficient for online learning, allowing the model to instantly update predictor bins and probabilities as each new customer response (accept/reject) is recorded.
- Question 2Advanced
Adaptive Analytics · Interpreting adaptive model reports and predictor performance
When reviewing the performance of an adaptive model in the Prediction Studio Bubble Chart, a data scientist notices that a specific model has a very high success rate (Y-axis) but a low model performance / AUC (X-axis). What is the most likely business implication of this scenario?
quadrantChart title Adaptive Model Bubble Chart Analysis x-axis "Low Performance (AUC)" --> "High Performance (AUC)" y-axis Low Success Rate --> High Success Rate quadrant-1 High Value / Needs Review quadrant-2 Optimal Models quadrant-3 Dormant / Poor Models quadrant-4 Niche / Target Refinement Current Model: [0.2, 0.8]Show answer & explanation
Correct answer: C
A high success rate combined with a low AUC (near 50) means that almost every customer is accepting the offer, making it impossible for the model to find differentiating predictors. This often happens when an offer is 'too good to be true' (e.g., free money), indicating the business might be giving away value unnecessarily.
- Question 3Beginner
Adaptive Analytics · How adaptive models self-learn from customer responses
A consultant is optimizing the predictors for an adaptive model in Pega Customer Decision Hub. They notice that the ADM automatically organizes similar predictors into clusters. What is the primary purpose of predictor grouping in Pega Adaptive Decision Manager?
Show answer & explanation
Correct answer: B
Predictor grouping in ADM identifies highly correlated predictors (e.g., 'Age' and 'Date of Birth') and groups them together. When calculating propensity, the model only uses the single most predictive active variable from each group. This prevents the Naive Bayes algorithm from double-counting correlated evidence, which would skew the propensity score.
- Question 4IntermediateSelect 2
Adaptive Analytics · How adaptive models self-learn from customer responses
When an adaptive model is created in Pega, it learns within a specific "model context." Which TWO of the following dimensions are standard components used to define the model context in Customer Decision Hub? (Select TWO)
Show answer & explanation
Correct answers: A, C
In CDH, an adaptive model's context is typically defined by the business hierarchy: Issue, Group, Name (Action), Channel, and Direction. 'Issue' is the top level of this hierarchy (e.g., Sales, Retention).
The 'Channel' (e.g., Web, Email, Call Center) is a critical part of the model context. A customer might have a high propensity to accept an offer on the Web but a low propensity via Email, so ADM creates separate model instances per channel.
- Question 5Intermediate
Adaptive Analytics · Exporting adaptive model data for offline analysis
A data science team needs to perform deep offline analysis on the predictor bins and raw learning data generated by their adaptive models over the past 6 months. To access this data, they must utilize the ADM data mart. The correct approach to extract this data for offline analysis is to export the ______ and ______ datasets from the ADM data mart.
Show answer & explanation
Correct answer: B
The ADM data mart specifically consists of two key datasets: Model Snapshots (containing overall model performance, success rate, and metadata at various points in time) and Predictor Binning (containing the detailed statistical bins, intervals, and behavior of every predictor). Exporting these allows for comprehensive offline analysis.
- Question 6Beginner
Adaptive Analytics · Interpreting adaptive model reports and predictor performance
A system administrator is reviewing the Adaptive Models monitoring tab in Prediction Studio. They notice several models are classified as 'Dormant'. What does this classification indicate about these specific models?
Show answer & explanation
Correct answer: B
In Prediction Studio, a model is flagged as 'Dormant' if it has not received any response data over a configured period of time. This usually indicates that the action associated with the model is no longer being presented to customers, or there is an issue with the response capture feedback loop.
Ready for the real thing?
The full PEGACPDS24V1 simulator has every exam-style question, timed mode, and instant scoring.