> ## Documentation Index
> Fetch the complete documentation index at: https://docs.getclaro.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Taxonomy

> Hierarchical, versioned categorization — first-class, not a free-text label.

Taxonomies in Claro are managed objects, not strings on a record. Every taxonomy has a structure, a version, an owner, and assignment history per record.

***

## What a taxonomy is

A taxonomy is a tree (or forest) of nodes. Each node has:

* **Name** — the human-readable label.
* **Path** — the full path from root, e.g. `Electronics > Audio > Headphones > Over-Ear`.
* **Synonyms / aliases** — alternate spellings or supplier-specific names.
* **Description** — optional definition used by AI assignment.
* **Properties** — typed attributes inherited by every record assigned to this node.

Taxonomies are versioned. Each published version is immutable; changes happen on a draft until you publish.

***

## Where taxonomies come from

You can populate a taxonomy in three ways:

1. **Generate from data** — run the **Generate Taxonomy** operation under Modules → Analyse. Claro proposes a hierarchy from your records, which you curate before publishing.
2. **Import a standard** — bring in GS1, Google Product, UNSPSC, or your internal taxonomy as a starting point.
3. **Hand-edit** — build or refine the tree manually in the Taxonomy surface.

Most teams combine all three: import a starting standard, generate proposals for the gaps, and curate by hand.

***

## Assigning records to taxonomy nodes

Assignment is itself a tracked operation with confidence scores.

* **Bulk Enrichment** with the taxonomy as a target attribute will propose assignments for unclassified records.
* Above-threshold proposals are auto-applied; below-threshold proposals queue in [Notifications](/notifications) for review.
* Assignments are versioned alongside the taxonomy: when you publish a new version, Claro shows you what would change and lets you re-classify in bulk.

***

## Working with the Taxonomy surface

The Taxonomy page lets you:

* Browse the tree and inspect node properties.
* See record counts per leaf and per branch.
* Drill into a node to view assigned records, with confidence and provenance.
* Edit nodes — rename, merge, split, move, or deprecate.
* Compare versions side-by-side.
* Publish a new version, with a preview of which records would be re-assigned.

***

## Versioning and rollback

Each published version has a version number and an effective date. Every record's taxonomy assignment is linked to a specific version. When a new version is published:

* Existing assignments stay valid.
* Records assigned to deprecated nodes are flagged for re-assignment.
* A bulk re-classification job can be queued automatically.

You can roll back to a prior version at any time; assignments revert to their state under that version.

***

## Tips

* Keep node descriptions specific and unambiguous — they directly improve AI assignment accuracy.
* Use synonyms aggressively: every supplier-specific term you add reduces manual review later.
* Don't model attributes as taxonomy nodes. *Color: Red* is an attribute, not a category.
* Generate Taxonomy works best on catalogues that already have decent free-text descriptions or category hints.