Introducing Endless AI and Endless MCP. Operate from anywhere.

1.1 · 3 min read

The Item Primitive

Why the item should be the record every commerce workflow references, and what goes wrong when each system keeps its own copy.

Contents

Some business software becomes infrastructure by organizing everything around one record that other workflows depend on. Salesforce organized sales and service around the customer record. Rippling organized payroll, benefits, and IT access around the employee record. In both cases, one shared record lets a single product handle many jobs without each job keeping its own version of the customer or the employee.

Commerce doesn’t have that record yet. We think it should be the item.

Commerce is built around transactions

Most commerce software is organized around orders, invoices, and shipments. Items appear on every one of those documents, and each system keeps its own copy of the item.

For a growing brand, that means the same product is described in Shopify, in the ERP, in a spreadsheet for the 3PL, and in a portal for each retailer. When a case pack or a carton dimension changes, someone has to update every copy, and the one they miss is the one that generates a chargeback. GTINs, internal SKUs, and retailer item numbers get mapped by hand for each new retailer.

Each category of software covers part of the problem. EDI providers carry item data between trading partners. PIMs manage product content for sales channels. ERPs handle accounting, purchasing, and inventory, and many of them struggle with retailer-specific attributes like case packs, compliance labels, and channel item numbers. PIMs can hold an authoritative product record, but purchasing, warehouse work, and retailer compliance still run on data kept somewhere else.

The item is the primitive

Purchasing, inventory, fulfillment, pricing, and EDI all depend on the item. A purchase order lists items with quantities, costs, and delivery terms. Inventory counts items by location. A pick list tells a warehouse which items to pull and how to pack them. Wholesale pricing sets a cost for each item and customer. An EDI document can pass every syntax check and still fail because its item identifiers or pack details don’t match what the retailer has on file.

An item record has to answer five questions:

  • What it is: attributes, descriptions, images, and identifiers like GTINs
  • Where it comes from: suppliers and the costs and lead times they quote
  • Where it goes: channels, retailer item numbers, and pricing by customer
  • How it moves: case packs, carton dimensions, and each retailer’s shipping and labeling requirements
  • What happened to it: orders, receipts, inventory adjustments, and returns

The first four correspond to the four roles described in The Four Horsemen of the Commerce Apocalypse: the brand defines the product, the manufacturer makes it, the retailer sells it, and the logistics provider moves it. The fifth is the history all of them add to.

Why now

AI makes inconsistent item data more expensive. An AI agent deciding whether an order can ship needs current stock, the right case pack, and that retailer’s routing rules. If those facts live in three systems that disagree, the agent will act on whichever version it reads, across every order in the queue.

Wholesale is also asking more of brands. Retailers expect EDI, accurate item setup, and compliant cartons from suppliers of every size, including brands that started direct-to-consumer and never planned for wholesale. Most of those requirements attach to the item, and a brand that keeps its item data in five places pays for the gaps in chargebacks.

What we believe

The item should be the authoritative record for what a product is and how it’s identified. Every workflow should reference that record instead of keeping its own copy.

Transactions should check against the item record. A purchase order with the wrong case pack should be caught at entry, before the retailer rejects the shipment.

Retailer requirements should be stored once, against the retailer and the items they apply to, so the next order uses them automatically.

Changes should apply going forward without rewriting history. A new case pack gets its own record, and a new case GTIN if the case carries one, while the old stock sells through. Each order keeps the configuration it was placed with.

We started with purchase orders and EDI because retailer orders are where bad item data gets caught, and charged for, fastest. Inventory, fulfillment, pricing, and sourcing work from the same item record.