Certified Pega Data Scientist key terms

A predictive model that updates its learned behavior as new case outcomes arrive, so predictions track changing patterns without a full manual rebuild. Pega uses it for churn, propensity, and prioritization where behavior drifts over time.

A statistical model trained on historical case data to score a numeric or categorical outcome, such as likelihood to accept an offer. It is built in Prediction Studio and deployed through a decision strategy or data flow.

The Pega authoring environment where data scientists create, monitor, and remediate predictions. It shows performance, alerts on degraded models, and hosts adaptive and predictive model lifecycles, plus the prediction data inventory.

A transparent, points-based model where each predictor contributes weighted points to a total score mapped to an outcome band. Pega supports scorecards as an alternative to machine-learned models when explainability is required.

Grouping a continuous or high-cardinality predictor into ranges or buckets so a model can treat each group separately. It reduces noise, handles outliers, and is common in scorecards and strategy conditions.

An input variable used by a model to explain or forecast an outcome, such as tenure, product count, or prior contact volume. Predictors must exist and be populated in the customer or case data at scoring time.

The mechanism that records the actual result of a decision, such as accepted or rejected, so models can learn from it. Without reliable outcome capture, adaptive learning and model performance measurement cannot work.

The scheduled or manual retraining of a predictive model with newer data, followed by comparison and promotion if performance improves. It keeps predictions aligned with current population behavior and is tracked in Prediction Studio.

A ranking of how much each predictor influences a model's output. It supports explainability, helps spot leakage or proxy variables, and guides which data to invest in collecting.

The degree to which a model's logic and drivers can be explained to stakeholders, regulators, and customers. Scorecards favor transparency; complex machine-learned models need supporting explanations and documented rationale.

A Pega rule that combines data, predictions, and business logic to produce a decision result. It chains components such as strategies, models, and scorecards and is the core artifact of Decision Management.

A rule that maps combinations of input conditions to output values in a tabular grid, evaluated top-down or by priority. It is readable by business users and often used for eligibility, pricing, and routing rules.

A rule that branches on successive condition tests to reach a result, similar to a flowchart of if-then paths. Pega offers it as an authoring style for logic that is easier to express as a hierarchy than a grid.

A reusable building block inside a decision strategy, such as a shape that applies a model, filters, aggregates, or sets properties. Components are wired together in a data flow of logic.

An offer, action, or message presented to a customer, with properties such as product, channel, and treatment. Propositions are the candidates that a decision strategy ranks and selects.

The set of properties describing each proposition, including eligibility conditions, applicability, and value. Decision strategies use these properties to filter, score, and prioritize offers.

The decisioning outcome that selects the most valuable and appropriate action for a customer or case given context, eligibility, and business objectives. It is produced by combining propositions, predictions, and arbitration.

The step that resolves competing propositions into a final set, applying business rules such as priority, caps, and suppression. It ensures the chosen actions respect contact limits and business constraints.

A declarative Pega rule that expresses policy, calculation, or logic independently of a process flow, such as a decision table or when condition. Rules are versioned, inherited, and reused across cases and strategies.

Pega's mechanism for choosing which rule version applies at runtime by evaluating class inheritance, ruleset and version, and circumstance. It explains why the same rule name can behave differently across contexts.

A container for rules that is versioned and deployed together, controlling which rule versions are active in an application. Rulesets are organized into layers that determine availability at runtime.

A condition attached to a rule so that a specialized version applies only when the condition is true, such as a specific channel or customer segment. It is evaluated during rule resolution.

The definition of a business transaction in Pega, including its data, stages, processes, and policies. Case types structure work from creation through resolution and carry the context used by decisioning and AI.

A major phase in a case lifecycle, such as intake, review, or resolution, that groups related processes and steps. Stages give visibility into progress and can trigger decisions at transitions.

The end-to-end path a case follows across stages and processes, from creation to resolution. Designing the lifecycle shapes where automation, decisions, and human approvals occur.

A sequence of steps within a stage that performs work, such as assignments, automations, and decisions. Processes define who or what acts at each point in the case.

A unit of work routed to a user or work queue for action, such as review or approval. Assignments create the human-in-the-loop points where case data and decisions are confirmed.

A shared pool of assignments that eligible users can pick up, used for team-based or skill-based routing. Queues support load balancing and are often driven by decision logic.

A rule that loads and caches data for use across the application, with configurable scope and refresh behavior. Data pages supply case and customer context to strategies and predictions.

A Pega rule that maps source data into a target structure, applying conversions, defaults, and calculations. It prepares inputs for models, strategies, and integrations.

A Pega construct that moves and processes data in batches or streams, often used to compute and store decision results. Data flows support model scoring at scale and feed outcome data back for learning.

Pega's capability set that applies AI to work orchestration, such as suggesting case actions, predicting case outcomes, and automating routine steps. It aims to improve efficiency and consistency across case work.

Using predictions and rules to send work to the best-suited person or queue based on skills, workload, and case context. It reduces manual triage and improves handling time.

Ongoing tracking of a deployed model's performance, data quality, and drift, with alerts when thresholds are breached. Monitoring triggers remediation such as retraining or retiring a model.

A testing pattern where the current model (champion) competes with an alternative (challenger) on live traffic. Results decide whether the challenger replaces the champion, supporting evidence-based model changes.

Running a decision strategy against sample or historical data to compare outcomes before deployment. It reveals how rule or model changes affect volumes, eligibility, and value without touching production.

A repository that holds decision-related data, such as propositions and interaction history, for fast access by strategies. It supports real-time decisioning and analytics on decision outcomes.

A record of past customer contacts, offers, and responses used to inform future decisions and model training. It underpins contact limits, arbitration, and adaptive learning inputs.