For a story created using multiple data sources (Salesforce, CSV, and external data), what is...

A supplemental dataset is appropriate when an Einstein Discovery model requires predictor information that is not available directly on the Salesforce object used for scoring. During model deployment, CRM Analytics allows supplemental datasets to provide external data, computed fields, or other required model variables. The supplemental dataset is associated with the Salesforce object through corresponding record-ID fields, allowing the additional predictor values to be supplied when predictions are generated. This directly addresses the missing-field situation shown in the exhibit. Deploying without connecting to a Salesforce object is intended primarily when the model does not depend on a Salesforce object, while actionable variables determine which variables Einstein can recommend changing to improve an outcome; they do not supply missing predictor data. Therefore, creating and associating a supplemental dataset is the appropriate solution.