Skip to main content
Version: Next

Hierarchies

The Hierarchies tab shows dimensional hierarchies extracted from this data source during semantic extraction. Hierarchies define logical drill-down paths through data, enabling the AI to group and aggregate results in meaningful ways.

What Is a Hierarchy​

A hierarchy is an ordered set of columns that form a drill-down path. For example:

  • Year → Quarter → Month → Day
  • Country → Region → City
  • Category → Subcategory → Product

When a user asks "show revenue by month" or "break down by product category", the AI uses hierarchies to understand how to group and aggregate the data at the right level.

How Hierarchies Are Generated​

Hierarchies are generated automatically during semantic extraction. The extraction process analyzes the schema, column names, and data types to infer likely hierarchical relationships.

Managing Hierarchies​

Enabling and Disabling​

Each hierarchy can be enabled or disabled. Disabling a hierarchy removes it from the semantic layer — the AI will not use it for query generation. Use this to remove hierarchies that are incorrect or not relevant for your use case.

Adding Custom Hierarchies​

You can define additional hierarchies manually:

  1. Click Add Hierarchy.
  2. Select the columns that form the hierarchy levels, in order from the broadest level to the most granular.
  3. Save.

Locking​

Lock a hierarchy to prevent it from being overwritten or removed during future semantic extractions. Use this for hierarchies you have manually defined or confirmed as correct.

Best Practices​

  • Review auto-generated hierarchies after extraction and disable any that are incorrect.
  • Add hierarchies that the AI did not detect, especially for business-specific drill-down paths.
  • Lock hierarchies that drive core analytical workflows so they persist across re-extractions.