How to Optimize a Product Taxonomy With Attributes

A category called “Circuit Breakers” with three hundred products in it is technically a taxonomy. It’s not a useful one. A customer looking for a 2-pole, 16A, C-curve breaker has to open products one at a time to find it, because the category itself carries no information beyond “these are all breakers.” The taxonomy did the easy 20% of the work — grouping by product type — and left the useful 80% — letting someone actually narrow down to the right item — undone.

The difference between a category tree and an attribute-driven structure

A pure category hierarchy answers “what kind of thing is this.” An attribute-driven structure answers “which specific one do I need,” by exposing the attributes that actually distinguish products within a category as filters: pole count, current rating, trip curve, mounting type. The category gets you to the right aisle. The attributes get you to the right shelf.

Most catalogs have the first without the second, usually because attributes were collected inconsistently — some products have current rating recorded, others don’t; some record it as “16A,” others as “16 A” or “16.0” — which makes filtering unreliable even where the data technically exists.

What optimization actually requires

Identify the attributes that matter per category, not a generic set applied everywhere. The attributes that distinguish breakers (pole count, trip curve) aren’t the attributes that distinguish cable glands (thread size, IP rating). This is category-specific work, often informed by an existing classification standard like ETIM, which already defines relevant feature sets per class.

Normalize those attributes across every product in the category before exposing them as filters. A filter built on inconsistent units or formats produces wrong results — a customer filtering for “16A” won’t see a product recorded as “16 A” if the system treats them as different strings.

Fill the gaps, don’t just filter on what happens to be populated. If 40% of products in a category are missing the attribute customers most want to filter by, the filter is unreliable for nearly half the catalog. Enrichment to fill those gaps is part of taxonomy optimization, not a separate project.

Before and after

Category pages that are really just long, unfiltered lists? Book a 30-minute call.

Related reading

• Which classification standard should you use?
• What is data normalization?

FAQ

What's the difference between a category tree and an attribute-driven taxonomy? +

A category tree groups products by type. An attribute-driven structure additionally exposes the specific attributes that distinguish products within a category — like current rating or thread size — as filters, letting customers narrow to an exact product rather than browsing a list.

Why do attribute filters often not work well even when the data exists? +

Inconsistent formatting and units across products — “16A” versus “16 A” — make filters unreliable even when the underlying attribute is technically recorded, because the system treats inconsistent values as different.