Data Prep Terminology for Einstein Discovery

Variables

In Einstein Discovery, a variable is a category of data. Variables correspond to columns in CRM Analytics datasets and fields in Salesforce objects. Data scientists sometimes refer to variables as attributes or features. The variable you want to analyze or predict is called the outcome variable. The other variables in the dataset are called explanatory variables (in descriptive analytics) or predictor variables (in predictive and prescriptive analytics).

Observations

In Einstein Discovery, an observation represents an instance of the data you want to analyze. Observations correspond to rows in CRM Analytics datasets and records in Salesforce objects. When building a model, Einstein Discovery considers only those observations with a known outcome (the outcome variable for that observation contains a value).

Data Types

Data types are described differently for Einstein Discovery, CRM Analytics datasets, and Salesforce objects:

• In Einstein Discovery, a numerical variable represents number values you can perform math on. Numerical variables correspond to measures in CRM Analytics datasets and number fields in Salesforce objects.
• In Einstein Discovery, a categorical variable represents text values. Categorical variables correspond to dimensions in CRM Analytics datasets and text fields in Salesforce objects. Categorical values are represented as structured text. In contrast, unstructured text values represent free-form text content, such as customer email and text communications, written survey responses, product reviews, social media posts, and so on.
• In Einstein Discovery, a date variable represents date/time values. Categorical variables correspond to dates in CRM Analytics datasets and date fields in Salesforce objects.