Product Data Structure

The data model everything else sits on.

Product schemas, variant hierarchies, attribute taxonomies, media management, category mapping, localisation structure, and the PIM decisions that make or break multi-channel selling. Get this right once — everything downstream becomes easier.

Product schema · TK-047

6 linked entities

Parent

Merino Runner Sock

TK-047

Variants

27 SKUs

Attributes

84 fields

Media

18 files

Get these right up-front — save rebuilds later.

Parent/variant hierarchy

What's a variant (size/colour) vs what's a separate product? Wrong choice here leads to exploded catalogues or lumped listings that Amazon rejects.

Attribute types

Text, numeric, boolean, enum, multi-select, controlled vocabulary. Every attribute should have a defined type and allowed values — prevents garbage-in errors later.

Mandatory vs optional vs computed

Some fields mandatory (title, price), some optional (gift message), some computed from others (GTIN check digit, slug). Classifying these up-front prevents rework.

Category taxonomies

Your internal category structure vs Amazon's vs Google's vs Etsy's. Map once, use everywhere. Don't maintain 4 parallel categorisations manually.

Media structure

Image roles (hero, lifestyle, detail, swatch, infographic), sort order, per-channel variants, alt text at each language. A non-trivial schema all by itself.

Pricing hierarchy

Base price → currency conversion → regional markup → tax handling → promotional override. Price needs to be a small data model, not a single field.

Localisation structure

Per-locale fields (title, description, bullets, SEO) vs shared fields (dimensions, GTIN, images). Getting this wrong = translation nightmare.

Do you need a PIM?

The answer is "it depends" — here's how to decide.

A PIM (Product Information Management) system adds complexity. Worth it only when it removes more complexity than it adds. The thresholds matter.

You probably need a PIM if

100+ SKUs with frequent adds/changes

3+ sales channels (your site + 2 marketplaces)

Multi-language (2+ locales)

Complex variant structures (size × colour × material)

Multiple content contributors (marketing + merch + translators)

Frequent bulk updates (supplier feed imports, pricing rounds)

Shopify + a spreadsheet is fine if

Under 100 SKUs with stable catalogue

Single or dual channel (your site + Amazon maybe)

Single language, single market

Simple variants (size only, or colour only)

One or two people managing listings

Low update frequency (rare catalogue changes)

Our default recommendation: start with Shopify + well-structured metafields. Graduate to Akeneo, Plytix, or a custom Postgres-backed PIM when you cross 200 SKUs or 3 channels.

What clean structure unlocks

One data change, six downstream outputs updated.

Your own website

Shopify/Woo/headless store reflects latest product data. No manual sync.

Marketplace listings

Amazon, eBay, Etsy listings update when source data updates. Change once.

Google Shopping feed

XML feed regenerates with every change. Always current, always valid.

Meta Commerce catalogue

Dynamic product ads always serve the real price and availability.

Email marketing templates

Klaviyo/Mailchimp product blocks pull from the same source — stays consistent.

Print catalogues and point-of-sale

If you publish to print or run retail POS, data structure feeds those too.

Stacks we typically build on

Three patterns depending on scale.

Small

Shopify metafields + Channable

Under 200 SKUs, 1–3 channels, single language

Shopify tier + ~€100/mo

Mid

Plytix or Akeneo PIM + feed middleware

200–3000 SKUs, 3–6 channels, multi-language

~€500–€2000/mo tools

Large

Custom Node.js + Postgres PIM + built APIs

3000+ SKUs, 6+ channels, complex enrichment rules

Build project + low ongoing infra

Frequently asked

Data structure questions.

For under 200 SKUs, single-language, maybe two channels, Shopify's native metafields + product model can be sufficient. Beyond that, the lack of governance (anyone can add any metafield), no versioning, and weak media management starts to bite. We migrate from Shopify-as-source to PIM-as-source when that threshold is crossed.

Yes. We scope 4–8 weeks for a mid-scale PIM implementation depending on data migration complexity. Includes data modelling, initial data import, integration with your store and marketplaces, and user training for your team.

Product data structure is the source; Product Feed Optimisation is the delivery. We treat them as two ends of the same pipeline — getting the structure right at the top of the pipeline makes feed optimisation far easier and more maintainable.

For image variants (thumbnails, channel-specific sizes) yes — handled in pipeline. For full Digital Asset Management (video, brand collateral, marketing assets) we typically recommend Bynder or Brandfolder for that side and integrate them with the PIM.

Related services

The full marketplace stack

Product Feed Optimisation

Downstream delivery

Product Catalogue

Ecommerce catalogue mgmt

Marketplace Listings

Per-platform listing work

Get your data foundation right.

Share a sample of your current product data (export from Shopify, a spreadsheet, even a manual listing) and we'll return a data model assessment with specific structural improvements. Free diagnostic.