How to Build a Category Attribute Template Suppliers Can Fill

The fastest way to make a tidy spreadsheet is to copy last month’s supplier file. It also locks that supplier’s names, units and blind spots into your catalog. A category attribute template should describe the decisions your buyers make, then accept many suppliers’ ways of expressing them.

If you want a starting point rather than a blank sheet, the Attribute Library has candidate templates for over 100 categories, each downloadable as CSV. Use this guide to adapt one to your own range.

Start from ten real products

Pick about ten products that represent the category:

• a common, high-volume item
• a near duplicate of that item from another supplier
• a variant family (sizes, colors or ratings of one base product)
• a kit or assembly
• a private-label item
• a product with no GTIN
• two or three items that are technical, regulated or often returned

For each one, ask what would make a buyer choose the wrong part. That question tends to expose a mounting type, pressure rating, thread, material, compatibility note or pack quantity that the current spreadsheet leaves out.

Build four layers

1. 1
   Identity

   Link the supplier SKU, manufacturer part number, GTIN where one exists, variant and pack level. Nothing else in the record can be trusted until these point at one item. See GTIN vs MPN vs SKU for how the keys relate.

2. 2
   Technical specifications

   Store each measured or rated value as a number with a separate unit, plus the conditions it was measured under. “65 kA at 240 V” is two facts, not one string.

3. 3
   Selection and fit

   Record what separates adjacent products: type, material, application, and what the item mounts to or replaces. Fit claims need a source; a similar-looking part is not a verified fit.

4. 4
   Evidence

   For listings, certifications, origin and other claims, store the source document, its revision, the page, and the approval state. A claim without a document is a question for the supplier.

Define each field once

For every field, write down:

• a plain label and a stable key (interrupting_rating_ka)
• the value type: number, controlled list, identifier or text
• the unit, or the list of allowed values
• whether it belongs to the product, the variant or the pack
• the rule for a missing or conflicting value

Keep the raw supplier text beside the normalized value. “3/4 in” and “DN20” are related but not interchangeable, and should not be merged because they look close. For safety- or compliance-related attributes, send uncertain values to a specialist; a model’s confidence score is not a certificate.

Test against a second supplier

Load a second supplier’s file into the draft and count four things:

1. fields that can be read directly
2. fields that need a unit or vocabulary conversion
3. fields that need a supporting document such as a datasheet
4. fields that are still unknown

A usable template leaves room for unknowns. It does not turn them into guessed values. If most fields land in the third bucket, the template is fine and the work is extraction from documents. That is the gap a source-backed enrichment workflow closes.

The CSV columns to start with

Every template in the Attribute Library downloads with these columns:

Column	What goes in it
field_key, label	The stable key and the display name
group, priority	Identity, technical, selection, fit or evidence; core, buyer decision or channel
value_type, unit	How the value is stored and its unit
example_format	What a well-formed value looks like
raw_value, normalized_value	What the supplier sent, and your governed version
source_uri, source_page_or_row	Exactly where the value came from
confidence, review_state, last_verified	How sure you are, who approved it, and when it was last checked

The last seven columns are empty on purpose. Filling them per product is what turns a field list into a record you can defend.

What is a category attribute template? +

It is the agreed list of fields for one product category: each field’s label, key, data type, unit or allowed values, whether it belongs to the product, variant or pack, and the rule for a missing or conflicting value. It describes the decisions buyers make, so it can accept many suppliers’ ways of expressing the same facts.

Why not just copy the best supplier's spreadsheet? +

A supplier spreadsheet carries that supplier’s names, units and blind spots. If you copy it, every later supplier has to be forced into one vendor’s view of the category, and the fields that supplier never published stay missing in your catalog too.

How many products do I need to design a template? +

Start with about ten that represent the range: a common item, a near duplicate, a variant family, a kit, a private-label item, a product with no GTIN, and a few technical or regulated items. Then test the draft against a second supplier’s file before you roll it out.

Should normalized values replace the supplier's raw text? +

No. Keep the raw supplier text beside the normalized value. It lets a reviewer see what the source actually said, and it stops values such as 3/4 in and DN20 from being merged just because they look related.