What Data Types Are Supported as Predictors in Einstein Discovery?

Quick Answer

Keeping Customer_Status__c as a text field and letting Einstein Discovery treat it as a categorical predictor is correct because that's exactly how the platform is designed to handle string-valued fields like this one; values such as Active, Inactive, and Churned represent distinct categories rather than numeric quantities, and Einstein Discovery automatically recognizes and encodes text fields this way without requiring the data scientist to manually convert them first. Manually transforming the field, for example into separate numeric flags or a custom encoding, would be unnecessary extra work and risks losing the semantic meaning and full set of possible values the field naturally carries, since a manual transformation could inadvertently drop or misrepresent a category the platform would have otherwise handled correctly. Native support for categorical predictors is one of the conveniences that makes Einstein Discovery approachable for data prep, because it removes a step that's often manual and error-prone in traditional modeling workflows. The general principle to carry forward is that when a field in Einstein Discovery holds a limited set of distinct text values representing states or categories, rather than a continuous number, the right move is almost always to leave it as-is and trust the platform's native categorical handling rather than reaching for manual preprocessing.