Curious how product classification runs on your data? We'll show you on one real file.
See it on your data →Classification decides attributes, filters and compliance downstream
Classification is the layer that makes a catalog searchable, filterable and comparable — and it is usually the first thing to fall behind as a catalog grows. New products pile up unclassified because taxonomy work is specialist labor that does not scale with volume, and the specialists you do have will not agree with each other on every edge case. The cost shows up downstream: a filter that returns nothing, a search that misses the obvious result, a category page missing half its range. Claro reads the full record — not just the title — places it, and states the reasoning, so a reviewer can check the call in seconds instead of redoing the work.
Why two people classify the same product two different ways
New products pile up unclassified because taxonomy work does not scale with catalog growth.
Two people classify the same product two different ways.
Downstream search and filtering are only as good as the classification underneath.
How automated product classification works
The full product record, not just the title.
ETIM, eCl@ss, or a taxonomy you define.
A stated reason for every node chosen.
Low-confidence calls flagged for a human.
Who classifies at this scale
What goes in, what comes back
Claro writes back through files and APIs rather than certified connectors, so this list is a guide, not a limit.
Product classification: common questions
Which classification standards does Claro support?
ETIM, ECLASS, UNSPSC and GPC, plus any internal or customer-specific taxonomy you define. The tree is an input, not something hard-coded into the product.
Should we use ETIM, ECLASS or UNSPSC?
It depends on who consumes your data. ETIM dominates electrical and building materials in Europe, ECLASS is broader and strong in industrial procurement, UNSPSC is common in spend analysis. We wrote a comparison that walks through the trade-offs.
How accurate is classification across all four levels?
In one industrial-marketplace engagement, zero-shot classification of 100,000 products reached over 70% accuracy across all four ECLASS levels in high-confidence cases, with no category-specific training. Deeper levels are always harder than the top one, which is why confidence is reported per level rather than as a single number.
Does it need training data from our catalog?
No. It runs zero-shot against the taxonomy definition and the product evidence. Your corrections improve later runs, but nothing has to be labelled before you start.
Can it classify into our own internal taxonomy?
Yes — that is the common case. Give Claro the tree and the rules you apply, and it places products against yours rather than a public standard.
See it work on your own catalog.
Bring one supplier file and we'll run product classification on your real data — matched, classified and reviewable.