Taxonomy and Labelling
Date: 2026-08-17
The words used for categories, filters and attributes. It’s the cheapest high-leverage work in information architecture and the most neglected — a correct structure with your vocabulary rather than the customer’s fails exactly as badly as a wrong structure.
A taxonomy is the controlled set of terms used to classify things. Labelling is which of those terms the customer actually sees.
They’re separable, and separating them is the useful move. The internal taxonomy can be precise and systematic; the label can be the word people use.
INTERNAL CUSTOMER-FACING
SKU_CATEGORY_04 Cleansers
"Ambient" Doesn't need refrigerating
"P&P" Delivery
Where the vocabulary should come from
SITE SEARCH LOGS what they type, in
their words
← the best source you
already own
ZERO-RESULT SEARCHES terms you don't
recognise
SUPPORT TICKETS how they describe
problems
REVIEWS how they describe
products
CARD SORTING the names they give
their own groups
— Card Sorting and Tree Testing
SEARCH DEMAND what has volume
externally
See: Card Sorting and Tree Testing
Site search is the highest-value input and the most ignored. People typing “postage” when your label says “Delivery” is a finding handed to you for free, and it’s sitting in the search log — Search and Findability.
The internal-jargon problem
The most common failure, and it’s invisible from inside:
| Your word | Their word |
|---|---|
| Ambient | Doesn’t need the fridge |
| SKU | Product |
| Consumables | Things you use up |
| Modality | — |
| Range | Brand, or collection |
| Basket / bag / cart | Pick one and use it everywhere |
Everyone inside the business learns the internal vocabulary within a month and then cannot hear it. This is why the words have to come from data rather than from a meeting.
Rules that make labels work
- Use their word, even when yours is more accurate. If customers call it “postage”, the label is “Postage” — precision that isn’t understood is not precision
- Be specific. “Products”, “Solutions”, “Resources” are placeholders, not labels
- Front-load the distinguishing word. “Delivery information” and “Delivery costs” scan differently when the eye only reads the first word or two — Reading Behaviour Online
- Be consistent everywhere. The same thing must have the same name in the menu, the filter, the URL, the confirmation email and the packing slip
- Avoid clever. A named category nobody can decode costs findability for personality
- Don’t label by internal structure. Categories mirroring departments is the classic tell
Synonyms are the cheap win
You do not have to choose one word — you have to choose one label, and then accept every synonym in search.
LABEL Delivery
ACCEPTS postage · P&P · shipping ·
dispatch · when will it arrive
LABEL Sensitive skin
ACCEPTS reactive · eczema · allergy ·
hypoallergenic · gentle
Synonym mapping in site search is among the highest-return, lowest-effort changes available, and it turns a zero-result search into a sale — Search and Findability.
Attributes and facets
The taxonomy that matters most commercially is the attribute set, because it drives filtering:
NEEDS TO BE
consistent "500ml" not "0.5L"
on some products
complete a product missing a value
disappears from that filter
the right filter by "skin concern"
dimension not by "marketing category"
Incomplete attribute data is invisible and expensive. A product with no “skin type” value never appears when that filter is applied, and nothing surfaces the omission — the product simply stops selling — Faceted Filtering, Data Quality Monitoring.
Maintenance
Taxonomies rot. New ranges get filed under whatever existed, terms drift, and after two years the structure describes history rather than the catalogue.
QUARTERLY
zero-result searches → missing terms
filter usage → dead facets
attribute completeness by category
new products filed correctly?
Attribute completeness is the metric worth tracking, because it’s the one with a direct revenue consequence and no natural alarm — Data Quality Monitoring.