Pega CPDC key terms

The single action, offer, or treatment that the decision engine selects as most valuable for a specific customer in a specific context, balancing customer needs with business objectives.

The discipline of using data, predictive analytics, and business rules to automatically choose and deliver the most appropriate action to each customer at each interaction point.

A business offer or message that can be presented to a customer, such as a product recommendation, retention offer, or service action, defined independently of how it is delivered.

A configured instance of a proposition tied to a specific channel and direction, such as an outbound email offer or an inbound web banner, that can be selected during decisioning.

A specific variant of how an action is presented, used for testing and optimization, such as different creative content or wording for the same underlying offer.

The sequence of interactions and stages a customer moves through over time, which decisioning uses to determine contextually appropriate actions at each stage.

The set of circumstances describing a customer and interaction, such as channel, time, tenure, and current journey stage, used to filter and prioritize eligible actions.

The use of statistical models trained on historical data to estimate the likelihood of future customer behaviors, such as accepting an offer or churning.

A named collection of customer attributes and properties, such as profile data, interaction history, or aggregates, that decisioning reads when evaluating strategies and models.

The consolidated record of a customer's attributes, history, and behavior stored in a data set, serving as the primary input for segmentation, eligibility, and model scoring.

A configuration that defines how and when data sets are refreshed or loaded, including source, schedule, and mapping, so decisioning works with current customer information.

A reusable building block of a decision strategy, such as a filter, group-by, prioritization, or propensity component, that transforms a candidate population into a ranked result.

A rule-based logic flow assembled from strategy components that evaluates candidates and produces a prioritized list of actions for a given context.

The visual layout of a decision strategy, showing the sequence and branching of components from input population through filters and prioritization to final output.

A condition that determines whether an action, proposition, or treatment is eligible for a given customer and context, evaluated before scoring and arbitration.

The set of rules and criteria, such as product ownership or regulatory constraints, that must be satisfied for a customer to be considered for an action.

The division of a customer population into groups sharing characteristics, used to target strategies, set eligibility, and tailor actions to relevant audiences.

A repository optimized for storing and querying the large volumes of interaction and model data that decisioning and adaptive models rely on at runtime.

The configuration environment where business users define engagement policies, actions, treatments, and arbitration settings that together produce the decision strategy hierarchy.

The set of rules governing when and how actions may be presented, including applicability, eligibility, suitability, and contact limits across channels and customers.

A policy check that ensures an action is appropriate for a customer's financial situation, risk profile, or preferences, beyond simple eligibility.

Rules that limit how often and through which channels a customer may be contacted, preventing over-solicitation and respecting customer preferences and regulations.

The interaction medium through which an action is delivered, such as web, mobile, email, call center, or branch, each with its own constraints and timing.

A named collection of actions or propositions organized for shared eligibility and arbitration treatment within Next-Best-Action Designer, such as a product family.

The process of comparing eligible candidate actions and selecting the best one or few, using prioritization weights, business rules, and predictive scores.

A strategy component that ranks candidate actions by a computed value, often combining business weight, propensity, and value to determine which action wins arbitration.

A configurable numeric factor expressing the relative business importance of an action or group, used alongside predictive scores during arbitration.

The predicted probability that a customer will take a desired action, such as accepting an offer, produced by a predictive or adaptive model.

The estimated business benefit of a customer taking an action, often combined with propensity to compute expected value used in prioritization.

A machine learning model that continuously updates its predictions as new outcome data arrives, learning customer response patterns without manual retraining.

An input variable, such as customer attribute or context field, that an adaptive model uses to estimate the likelihood of a particular outcome.

The observed result of presenting an action, such as acceptance or rejection, recorded as feedback that trains adaptive models over time.

The mechanism that captures actual customer responses and routes them back to adaptive models so predictions improve with each interaction.

A measure of how well an adaptive model predicts outcomes, typically expressed through metrics like AUC or log loss computed from scored responses.

A testing capability that runs strategies against historical or synthetic data to compare outcomes and validate decision logic before production deployment.

A testing approach that runs an existing strategy (champion) against a new candidate strategy (challenger) to compare performance on the same population.

A subset of customers excluded from a treatment or strategy so results can be compared against normal decisioning to measure true impact.

The defined population of customers a simulation or strategy test runs against, often drawn from historical interaction data or a sample data set.