cohentrixCohentrix product team · Reviewed 29 October 2026 · Independent worked examples
What is a product taxonomy?
A product taxonomy is a defined hierarchy that organizes product records into categories. It gives different source systems a shared structure for describing and comparing products.
One practical hierarchy is category → subcategory → product → variant. The correct levels depend on the decisions the business needs to make; a hierarchy designed for menu-mix analysis may differ from one used in a retail catalog.
Classification, categorization, and normalization.
| Term | Practical meaning |
|---|---|
| Classification | Assigning a record to defined classes or structured fields. |
| Categorization | Placing the record in a category; often used as a narrower part of classification. |
| Normalization | Making naming or representation consistent while preserving meaning. |
| Taxonomy mapping | Connecting a source record or source category to a target taxonomy. |
A simple example.
“Thé glacé citron 0,5 L” and “Iced citrus tea 500ml” could be mapped to Beverages → Tea → Iced citrus tea → 500 ml. That mapping should be treated as a proposal until a reviewer confirms the flavor, composition, and size are equivalent.
Why review matters.
Similar words do not always mean the same product. A “regular” portion may have different sizes across markets. A generic beverage label may omit a flavor or preparation method. Reviewers should preserve these distinctions instead of forcing every ambiguous record into a convenient answer.
How to evaluate a classification pilot.
- Representative sample: include categories, markets, languages, and difficult cases.
- Reference labels: agree how reviewers establish the expected mapping and resolve disagreements.
- Coverage: measure which records receive a useful proposal.
- Completeness: count required fields that remain blank or invalid.
- Reviewer agreement: compare proposed mappings with the agreed reference.
- Correction rate: measure how often and where reviewers edit proposals.
- Error analysis: inspect results by market, category, ambiguity, and input quality.
What a taxonomy does not establish.
A consistent structure does not establish that sales data is complete, that two restaurants are comparable, or that a campaign caused an uplift. Those questions require their own data validation and analytical methods.