Product Attribute Templates by Category

The hardest part of building a product attribute schema isn't the concept — it's the blank page. You know you need fields per category, tiered by importance, but staring at an empty spreadsheet for "apparel" or "supplements" doesn't tell you where to start. This article is the starting point: four working templates, one per representative category, built the way we build them with clients — starting from the question a shopper (or a shopping agent acting for one) actually asks.

Why start from a template instead of a blank schema

We covered the design process in full in designing a product attribute schema AI can use: pull 90 days of shopper questions, cluster them into facts, tier each attribute, and pressure-test on hero SKUs. That process works from zero, but it's slower than it needs to be when the category is a common one. Apparel, beauty, home goods, and supplements together cover a large share of mid-market catalogs, and the top questions in each barely vary from merchant to merchant — a shopper asking about a t-shirt wants to know the same handful of things whether they're buying from you or a competitor.

So instead of starting from nothing, start from a template calibrated to your category, then run the pressure test to adapt it. That's the honest use of everything below: a first draft, not a finished contract. Your catalog will have quirks — a supplement brand selling only capsules doesn't need a scoop size field; a home goods merchant selling only rugs doesn't need assembly required. Delete what doesn't apply and add what your own return reasons and support tickets surface.

Apparel: what product attributes does an apparel product need

Apparel questions cluster almost entirely around one theme: will this fit and hold up the way I expect. Shoppers can't touch the fabric or try it on, so every attribute in the template exists to substitute for that missing sense of touch and sizing intuition. This is also the category where compatibility and companion questions show up hardest — "does this run small," "what do I wear it with" — because sizing is the single biggest driver of returns in the category.

Two fields carry more weight than the rest for AI shopping specifically. fit_type and true_to_size are what let a Brand Agent or an off-site assistant answer the question that drives most apparel returns without guessing — and guessing here is expensive, because a wrong sizing answer costs a return, a shipping label, and a shopper who now trusts the agent less. If you enrich only two fields in this category this quarter, make them these two.

Compatibility data belongs here too, even though it doesn't fit neatly into a single-row attribute: "pairs with," "part of a set," "matches the [product] in [color]." We treat that relationship data as its own layer rather than a column — see compatibility data for AI agents for how to structure it.

Beauty: what attributes matter most for AI shopping in beauty

Beauty questions cluster around formulation and suitability — what's in it, what it does, and whether it's safe for a specific skin type, hair type, or sensitivity. This is a category where prose has historically carried almost all of the information ("a lightweight, non-greasy formula that absorbs instantly"), and where an agent needs structured fields to answer the questions that actually gate a purchase: ingredients, skin type fit, and claims that can be verified rather than merely asserted.

full_ingredient_list and skin_type are the two an agent leans on hardest, because they're the two most likely to appear verbatim in a shopper's prompt — "a niacinamide serum for oily skin" is a completely ordinary query for an off-site assistant to receive and needs both fields structured to answer well. Beauty is also where getting a field wrong costs more than leaving it blank: a fragrance_free flag set incorrectly, or a missing allergen, can mean an agent confidently recommends a product that causes a reaction. Treat every field in this category's ingredient and claims group as belonging to the required tier, not recommended — see the tiering model in the schema design article for how required-tier enforcement works.

Home goods: dimensions, materials, and the compatibility question

Home goods questions cluster around will it fit in my space and does it work with what I already own. Unlike apparel or beauty, the core anxiety isn't about the shopper's own body — it's about a room, a doorway, an outlet, or an existing piece of furniture. That makes dimensional precision and compatibility relationships the two heaviest-weighted groups in this template.

dimensions_lwh_cm and works_with do the most work here. Dimensions are the single field most likely to be stored inconsistently across a home goods catalog — some SKUs in inches, some in centimeters, some as a single string like "32 x 18 x 40" that no filter can parse — which is exactly the failure mode the schema design article warns about under unit discipline. Store each dimension as its own numeric field, pick one unit catalog-wide, and convert on entry. The works_with relationship is what lets an agent answer "does this fit my [existing product]" instead of declining — it's the highest-effort field in this template and the highest-value one.

Supplements: dosage, form, and the fields an agent must never guess

Supplements share beauty's ingredient sensitivity but add a second layer: dosage and form questions that are effectively health questions in disguise. "How much do I take and when" and "is this safe with what I'm already taking" are the two questions that dominate this category, and they're also the two where a confidently wrong agent answer carries real consequences, not just a return.

Every field in this template's top half — active_ingredients_and_dosage, allergen_flags, third_party_tested — belongs in the required tier without debate. This is the clearest case in any of the four categories for the rule we lay out in the schema design article: if a wrong answer here is worse than no answer, the field is required, not optional. An agent that declines to state a dosage because the field is empty is doing its job; an agent that states a wrong dosage because someone typed a placeholder value to get past a lax gate is a liability. Supplements is also the category where structured data for AI shopping matters most for off-site visibility — assistants like ChatGPT and Perplexity are especially conservative about recommending health-adjacent products without clear, structured label data to cite.

How the four templates compare

Laid side by side, the four categories don't just have different fields — they weight the same underlying question types differently. Seeing that pattern is more useful than memorizing any single template, because it's what lets you build a fifth template for a category not covered here.

The pattern holds across most categories we've enriched: physical products people wear or place lean on fit and dimensions; products people consume or apply lean on formulation and constraints; anything with accessories or a room around it leans on compatibility. When you're building a template for a category not covered here — furniture, electronics, pet food, whatever your catalog holds — start by asking which of those four buckets your category leans into hardest, then borrow the closest template above as a skeleton.

Adapting a template to your catalog

None of the four templates above should go into production unedited. Use them as a first draft and run the same pressure test we recommend for any schema: hand-fill the template for your ten best-selling SKUs in the category, and pay attention to every field that makes you hesitate, improvise, or leave blank. That hesitation is signal — either the field needs a clearer definition, or it doesn't actually apply to your catalog and should come out.

1. 1

   Pick the closest template

   Match your category to whichever of the four above shares the most DNA — a footwear catalog borrows from apparel; a candle catalog borrows from beauty's formulation fields; a pet bed catalog borrows from home goods.

2. 2

   Cut what doesn't apply

   Drop fields your catalog has no variation on. If every SKU in a category is fragrance-free, that field isn't adding decision-making value — though it may still be worth keeping as a fixed claim for GEO purposes.

3. 3

   Add what your data surfaces

   Pull your own 90 days of support tickets, search queries, and return reasons per the schema design process and add any recurring question the template missed.

4. 4

   Re-tier for your business

   A field that's optional in the template above might be required for your specific catalog if it drives your specific returns or your specific shopper questions. Tiers are a judgment call, not a fixed property of the field.

5. 5

   Pressure-test on hero SKUs

   Hand-fill the adapted template for your ten best sellers before rolling it out catalog-wide. Fix ambiguity now, while it affects ten products instead of two thousand.

Where these templates fit in the bigger enrichment effort

A category template solves one piece of a larger problem. It tells you what fields to define; it doesn't fill them, govern them, or connect them to the systems that read them. The catalog enrichment playbook covers the full sequence — audit coverage, enrich hero SKUs first, then high-traffic categories, then the long tail — and the schema design article covers the governance layer that keeps a schema from decaying once the enrichment project ends. If your catalog spans several categories, category-by-category taxonomy decisions also matter for how these attribute sets nest under your navigation and facets — that's covered in product taxonomy design for AI.

Used together, the three pieces cover the whole discipline: taxonomy decides how categories are organized, the schema decides what each category's products must say, and templates like the four above give you a fast, calibrated starting point instead of a blank page. None of it is exotic work — it's closer to a merchandising exercise than an engineering one — but it's the exercise that determines whether a Brand Agent, an on-site search bar, or an off-site assistant like ChatGPT can actually answer a shopper's question about your product instead of guessing or declining.