Every business that sells more than one thing has a catalogue, even if it is a price list in a spreadsheet or a services page on the website. And every catalogue has a shape. How many items there are. How they are split. Which differences become separate products and which become options. What comes in a bundle. How prices step up.
Most of that shape was never designed. It grew. A supplier's codes were copied in as they were. Someone added a variant as a new product because the system made that easier. A promotion became a permanent bundle. The result works for the people who built it and confuses everyone else.
The shape matters more now, because people are no longer the only readers of your catalogue. Search engines, marketplaces and AI shopping assistants read it too. If they cannot tell what you sell, they cannot recommend it.
The parts of a catalogue's shape
- The SKU. One sellable thing with its own code, price and stock. The hard part is deciding where one SKU ends and the next begins.
- Variants. Versions of the same product that differ along a few dimensions: size, colour, pack quantity. They belong under one product with options, so a customer sees one item and picks.
- Attributes. The facts people filter and compare on: dimensions, material, compatibility, certifications, lead time. Stored as fields, they can be filtered. Buried in a description, they are invisible.
- Bundles. Several SKUs sold together at one price. Good bundles match how people actually buy. Bad ones exist because of a promotion three years ago.
- Price points. The steps from the cheapest option to the most expensive, and the quantity breaks. They show a customer where they fit.
A distributor's catalogue, reshaped
Take a hypothetical food-service packaging distributor that sells takeout containers, lids, cups and bags to restaurants and caterers.
Its catalogue has 1,800 lines, copied over the years from supplier price lists. The same 32-ounce round container appears four times, under two supplier codes, each sold in cases of 150 and of 500. Lids are separate products with no link to the containers they fit. "Compostable" appears in some descriptions and not in others. A restaurant owner looking for a soup container with a matching lid has to know the supplier code.
Reshaped, the same stock looks different:
1. One product per thing a customer recognizes. "Round soup container, 32 oz", with variants for material and case size.
2. Attributes as fields. Capacity, material, compostable or not, microwave-safe, dimensions, case quantity. Now the store can filter, and an assistant can answer "compostable 32-ounce containers with lids, case of 150 or less."
3. Relationships. Each container links to the lids that fit it. The product page shows them together, and a reorder can suggest both.
4. Bundles that match real orders. If most new restaurant accounts order the same twelve items, offer a starter kit at one price.
5. Price steps that make sense. One unit price per case size, with quantity breaks a buyer can see before asking for a quote.
The warehouse is the same. The catalogue changed, and with it how easily customers, Google Shopping and an AI assistant can find and compare what is inside.
Why agents care about shape
An AI agent, whether it works for your customer or for you, reads a catalogue the way a careful clerk would. It looks for one record per product, clear attributes and prices it can trust. A clean shape lets an assistant answer an exact question and lets a buying agent place a correct order. A messy one forces it to guess, and a guess on a quote or an order is expensive.
The same goes inside the business. A reorder agent can only suggest the matching lid if the catalogue says which lid matches. A quoting agent can only apply the right price break if the breaks are stored as data.
Shape changes with size
The Decision OS under Humonix is called Allometry, after the branch of biology that studies how an organism's shape changes as it grows. Galileo noticed that a larger animal needs proportionally thicker bones. Catalogues follow a version of the same rule.
At 40 items, a flat list works and everyone knows everything. At 400, you need categories and a few attributes. At 4,000, you need a real product model: variants, attribute standards, relationships, and rules for when something becomes a new SKU. A catalogue that keeps its 40-item shape at 4,000 items stops working, and the cost shows up as wrong orders, slow quotes and customers who cannot find what you carry.
So check two things. Is the catalogue clean? And does its shape fit your size today, and the size you expect next year?
Where to start
- Pick your top 100 by revenue. Fix their shape first: one product per recognizable thing, variants under it, attributes as fields.
- Write the rules down. When is something a new SKU, and when is it a variant? Which attributes are required in each category? Who approves a new product?
- Use the customer's words. Name products the way buyers search for them, and keep supplier codes as a field.
- Check it with a question. Ask an AI assistant to find something specific you sell. If it cannot, look at what it would have needed.
Humonix B2B e-commerce work starts with this. The catalogue goes into the Brain as clean records, and the store, the product feeds and the agents all read from that one shape. Price changes still wait for a person's approval.