How it Works

The not_constant test from dbt-utils™ validates that a column contains more than one distinct value. A column that is fully constant — where every row has the same value — is often a sign of a broken join, a hardcoded default, or an error in the transformation logic.

Steps and Conditions

1. Column Selection: Identify the column to validate.
2. Execution: The number of distinct values is computed.
3. Outcome: Pass if more than one distinct value exists; fail if the column is entirely uniform.

Example Usage: Event Tracking

An event tracking model should contain multiple distinct event types. If every row has the same event_type, it likely means a filter was accidentally hardcoded.

models:
  - name: user_events
    columns:
      - name: event_type
        tests:
          - dbt_utils.not_constant

A constant event_type column would indicate only one type of event is being captured, which would severely limit the usefulness of the model for analytics.