A sound product taxonomy separates three layers: an internal master classification that drives attributes, channel-specific navigation trees for shoppers, and mappings to external standards such as GS1 GPC, ETIM, ECLASS and Google's product taxonomy. Each layer has its own owner and change process, and the PIM maintains the links between them.
A product taxonomy works best as three linked layers. The first is an internal master classification that decides which attributes each product needs. The second is a set of navigation trees built for shoppers on each channel. The third is a set of mappings to external standards such as GS1 GPC, ETIM, ECLASS and the Google product taxonomy. A PIM (product information management) system holds all three layers and the links between them, which lets each layer change without breaking the others.
What is a product taxonomy?
A product taxonomy is a hierarchy that classifies products into categories, from broad to specific, such as Lighting > Indoor lighting > LED panels. A classification scheme is the formal version of that idea: a controlled set of classes, often with codes, definitions and attributes attached to each class.
In a PIM, taxonomy serves three purposes. It determines the attribute set through the link to product families. It drives navigation and filtering on websites and in print catalogues. It also supplies the class codes that retailers, wholesalers and marketplaces require.
Which external classification standards matter?
Each standard serves a different trading context, and their structures differ. The facts below come from the standards bodies' own documentation.
| Standard | Typical use | Structure | Notes |
|---|---|---|---|
| GS1 GPC (Global Product Classification) | Consumer goods, GDSN data synchronisation | Segment > family > class > brick | Published twice a year and date-stamped rather than versioned. A placeholder brick of 99999999 is used where no brick fits yet. |
| ETIM | Electrical, HVAC, plumbing, tools, building products | Product group > product class, with typed features | Classes have language-independent IDs and local-language synonyms. ETIM 11.0 is planned for 1 December 2026. |
| ECLASS | Industrial procurement and engineering | Segment > main group > group > commodity class | Eight-digit codes, two digits per level. Properties attach only at level four. |
| Google product taxonomy | Google Shopping and Merchant Center feeds | Multi-level category paths | Optional attribute. Google auto-assigns a category, and merchants can submit an ID or a full path to override it. |
Our guides to ETIM, BMEcat and ECLASS and to GS1, GTINs and GDSN cover each standard in more depth.
How should the master classification be designed?
The master classification is the internal backbone. Its job is to assign the right attribute set and the right owner to every product. Design criteria that hold up in practice:
- Mutually exclusive leaves: every product belongs to exactly one master class.
- Attribute-driven splits: create a new leaf only when the products need a different attribute set or a different owner.
- Stable codes: class codes never change. Labels can be renamed and translated.
- Definitions and examples: each class has a written scope note with included and excluded examples.
- Depth guided by data: typically three to five levels, deeper only where the attributes justify it.
- External mappings: map each leaf to GPC, ETIM, ECLASS or Google classes. Adopt one external standard as the master only when a single trading context dominates.
How do navigation trees differ from the master classification?
Navigation trees serve shoppers. They group products by intent ("garden lighting", "gifts under 50 euros"), may repeat products across branches, and change with seasons and campaigns. Typical shopper-facing trees stay at three or four levels.
Keep navigation separate from the master classification in the PIM, or in the commerce platform, with rule-based assignment where possible. A rule such as "family = LED panel and IP rating ≥ 44 → Outdoor lighting" keeps the tree current without manual tagging.
How should taxonomy changes be governed?
Taxonomies drift without a process. A practical change process includes:
- A named taxonomy owner, usually within the product data or merchandising team.
- A change request log with rationale, affected products and affected channels.
- Impact analysis on attributes, completeness scores and outbound mappings before approval.
- Scheduled updates aligned with external releases, such as the twice-yearly GPC publications and ETIM major versions.
- Regression checks on feeds after each change.
These steps belong in the wider product data governance model, and they protect product data syndication feeds from silent breakage.
Key takeaways
- Treat taxonomy as three layers: master classification, channel navigation trees and external standard mappings.
- Split master classes only when the attribute set or the owner changes, and keep class codes stable.
- GS1 GPC uses segment, family, class and brick. ECLASS uses four coded levels. ETIM uses groups and classes with typed features.
- Navigation trees should follow shopper intent and can be rule-based on attribute values.
- Govern taxonomy changes with a named owner, impact analysis and a calendar aligned to GPC and ETIM releases.
Frequently asked questions
What is the difference between a taxonomy and a product family?
A taxonomy is a hierarchy of categories used to classify and find products, such as Tools > Power tools > Drills. A product family is a data structure that defines which attributes a product has. One family can appear in several taxonomy branches, and one category can contain several families, so the two are usually maintained separately in a PIM.
Which classification standard should a manufacturer use?
It depends on the trading partners. Consumer goods suppliers that synchronise data through GDSN need a GS1 GPC brick for each item. Electrical, installation and building products wholesalers in Europe commonly require ETIM. Industrial procurement platforms often use ECLASS. Many manufacturers maintain mappings to two or three standards from one internal classification.
How deep should a product category tree be?
There is no single correct depth. Shopper-facing trees typically work best with three or four levels, so that most products are reachable in a few clicks. Internal master classifications can go deeper when each leaf genuinely needs a different attribute set. Test any tree with real search and navigation data before finalising it.
For guidance on structuring a PIM selection, see the independent resource PIMselection.com.