Configure Measures, Predictions, and Exports
Measures
A measure is a calculation over unified profiles and activities, defined with a name, the entity it applies to, and a formula using functions and relationships. Measures produce per-customer values used in segments and analytics. Know that measures can be recalculated on a schedule and that dependencies matter-a measure referencing another measure must run after it.
Predictions
Predictions use machine learning to score customers. Understand the typical prediction types: customer churn, customer lifetime value, and product or subscription propensity. Each prediction requires training data (historical outcomes), a defined prediction window, and sufficient volume and quality of data. Predictions produce scores attached to unified profiles, which can then drive segments such as "high churn risk." Know that predictions must be trained, evaluated for quality, and then published before scores appear.
Exports
Exports push segments or entities to destinations such as Azure Blob Storage, Dataverse, or other connectors. Know that exports run on a schedule, require configured connections, and can be triggered manually. Understand that export configuration includes which segments or entities to send and the target.
Traps
Candidates often confuse measures with segments: measures compute values, segments select populations. With predictions, a common trap is assuming a model works without enough historical data or with a poorly chosen outcome column. For exports, remember that a segment must exist and be refreshed before it can be exported.
Rule of thumb
Measures answer "how much," segments answer "who," predictions answer "what next." Validate data volume and outcome quality before training any model.