Can You Sell Order and Inventory Data for AI Training?

Yes. Order, inventory and ERP records are among the company records buyers name most plainly: eight buyers of company records name them, and Mercor runs a listing for payments and commerce businesses that asks for years of transaction, order, inventory or shipment tables, at $80K to $900K a company. They sell as the record of how the business runs, with the customers taken out. A customer list is a different product, sold to a different buyer.

The store platform is rarely what stands in the way. The terms we read leave the records with the merchant and keep only a licence and an anonymised aggregate for the platform. The exception is selling through someone else's store: a business that sells on eBay, Amazon or Etsy holds its buyers' details under that site's rules, not its own.

This guide is for founders and finance or operations leads of online stores and direct-to-consumer brands, retailers, wholesalers and distributors, and manufacturers whose orders and stock run through an ERP (enterprise resource planning: the system that holds orders, inventory, purchasing and the books together). It covers what order and inventory data is, which buyers name it, why the operating record matters more than the customer list, what your store or ERP platform's terms say, what the records are worth, the customers and others in them, how to take the records out and prepare them, and two worked examples. It sits under what AI labs buy, beside the guides to accounting and finance records, CRM data and support tickets.

Contents

• What order and inventory data is
• Which buyers name it
• Why the operating record matters more than the customer list
• What your store or ERP platform's terms say
• What order and inventory data is worth
• The people in the records
• Taking the records out and preparing them
• Two worked examples
• Questions people ask
• What this means for you

What order and inventory data is

Every business that sells goods keeps a record of each sale and of the stock behind it, and most keep far more than they think. An online store's platform holds every order with its lines, discounts, taxes, payment status, shipping and refunds, beside the product catalogue and its price changes. A wholesaler's or manufacturer's ERP holds sales orders and quotes, purchase orders to suppliers, goods received, stock movements between locations, work orders, invoices and the approvals that moved each of them along. Around both sit the customer service tickets about orders, the emails with suppliers and the spreadsheets that planned the next season.

Taken together, these records show how a business decides: what it bought and when, how it priced, what it did when stock ran short or a shipment went wrong, which orders came back and why. That is the part buyers want. A list of who bought what is a smaller and riskier part of the same record.

The rows with nobody in them, and the rows where people can be taken out without losing the meaning, are the core of a licensable lot. The rows that are about people (a customer file, a supplier's contacts) need the most work and add the least.

Which buyers name it

Buyers of company records publish what they want on their own pages, and we read them for order, inventory, ERP and e-commerce records. The finding: eight buyers name them, and one publishes a price range for them.

Mercor lists order records, inventory logs and listings for retail and e-commerce among its target industries, and its listing for payments and commerce businesses asks for years of transaction, order, inventory or shipment tables. Polyshares takes e-commerce systems for orders, products and customers (Shopify, Amazon Seller Central, WooCommerce, BigCommerce) and accounting and ERP systems for ledgers, invoices, bills and inventory (NetSuite, Sage Intacct, Microsoft Dynamics, SAP). micro1 names fulfilment workflows and inventory management under operations and logistics. Scale AI lists DTC e-commerce among its industries of interest. Centific lists ERP and finance among the systems it takes, and wholesale and distribution among its industries. Handshake AI lists NetSuite among the sources it buys, and Nyne names NetSuite and the transaction history behind it.

The demand is moving this way. Troveo, which licenses company records, says buyers now ask for structured data more than text, with large databases and ERP systems in higher demand than email and Slack, and Grepped's own instant quote scores CRM, accounting and ERP above email and chat. Most buyers also have a size line: the guide to buyer programmes sets the floors side by side, and Mercor's commerce listing is one of the few that says it takes companies of any size.

Why the operating record matters more than the customer list

Two different markets meet in a retailer's records, and confusing them costs money.

The first is the customer list: names, addresses and purchase histories, bought by a business that wants those customers. It has a long public record, mostly from retailers that closed. Pier 1's brand, website and customer databases sold in its bankruptcy for $31 million, Circuit City's e-commerce business and customer information for $14 million plus a revenue share, Zulily's brand and customer lists for $4.9 million, and three British fashion brands with their customer data and stock for £25.2 million. In sales like these the buyer is usually the closing company's closest competitor, and buyers' own accounts book the customer data at a small share of the lot: the brand carried the price. Selling customers' personal data to another business is also closely regulated, as the guide to selling customer data in bankruptcy shows.

The second is the operating record: how orders, stock, prices, suppliers and problems moved through the business, year after year. AI buyers want it to teach models how work is done inside a company, which public text cannot show; as SuperAnnotate says of company records, the reasoning matters more than the finished output. For that purpose the customers' names add nothing, and taking them out is what makes the record licensable at all.

Three things raise the value of an operating record, and they are the ones buyers name:

1. History. Several years of orders and stock show seasons, price changes, stock-outs and how the business reacted. A single year shows little.
2. Connection. Orders joined to their returns, support tickets, shipments and invoices tell a story a flat export does not.
3. Decisions. Purchase approvals, price overrides, write-offs and exceptions handled by hand are the judgment buyers pay for, and the part a flat export of orders loses first.

What your store or ERP platform's terms say

Before a business offers its records, it should know what its platform's terms say about who holds them and what the platform does with them. We read the published terms of three platforms. The finding: a store platform leaves the records with the merchant, and eBay sets rules on buyers' information.

Three things follow. First, a business selling through its own store (Shopify, BigCommerce, or a store it runs itself) holds its order history as its own, and no clause we read bars it from licensing a copy with the customers taken out. Second, the platform's aggregate is the platform's: a merchant can grant a buyer exclusive rights over its own records, not over what Shopify or BigCommerce already derived from them, and a Shopify merchant that wants to limit that can switch off Network Intelligence in its settings. Third, a merchant on eBay, Amazon or Etsy is in a different position. On eBay, using information about buyers without their consent is barred, so such a merchant's lot leaves buyers out entirely and offers what is its own: its catalogue, prices, stock, purchasing and operations. Amazon's and Etsy's terms for merchants set their own rules on buyer data; we could not read them by script, so a merchant there should read the terms in its own account before it offers anything that came from a buyer.

ERP vendors (NetSuite, SAP, Microsoft Dynamics, Sage) sell to businesses under negotiated or enterprise agreements, and the buyers above name them as sources they take. A business on an ERP should read its own agreement for who owns the data and how it is exported, as it would for any system.

What order and inventory data is worth

This is one of the few kinds of company record with a published price aimed at it. Mercor's listing for payments and commerce businesses, which asks for years of transaction, order, inventory or shipment tables, pays $80K to $900K a company, and its manufacturing listing, which asks for several years of work orders and production records, $100K to $1.5M. Its listing for consumer brands is for something else, product and campaign imagery the brand owns, which is worth knowing for a brand that holds both.

Across all sales of company records on record, the median is $72.5K per company a company, and published offers to holders run at a median of $277K per company. The buyers' own ranges are wide: Handshake AI says it has paid $100K to $4M+ per partnership, Mode (BizBuyData) $100K to $5M a company, Nyne commonly $100K or more, Scale AI $10K to $1M+, and micro1 sets out tiers per partnership. None of these is a price for orders alone: each buyer prices the company's whole record, with orders and stock as one part of it.

The customer-list sales above are not a guide to this price. They paid for a brand and the right to market to its customers, usually to a rival, and each came out of a bankruptcy, an administration or a wind-down. An operating record licensed to an AI buyer is priced on what it teaches, with the customers removed, and the company keeps its brand and its customers. The guide to what company data is worth sets out each factor, and the valuation applies them to what a business holds.

The people in the records

Order records are full of people, and most of them are customers who bought something and never imagined the record would train a model. A licence of order records is a licence of personal data until those people are taken out.

Customers. Under California's privacy law, licensing personal information to an AI buyer for money is a sale, for a business over the law's thresholds, and a business that sells must first tell consumers and honour their opt-outs. Records that are properly deidentified fall outside those rules, which is why de-identification is the usual route. In the UK, reusing customers' data for a new purpose is weighed under the compatibility test, as in the EU under Article 6(4), and customers must be told before their data is used for the new purpose. In Canada the privacy regulator expects consent for new purposes and disclosures to third parties. And using records in a way a privacy policy promised not to can be an unfair or deceptive practice.

Removing names is not enough. A delivery address, a rare product bought once, or a sequence of orders can point to one household. Canada's privacy regulator found that taking out names, emails and phone numbers did not meet the standard of no serious possibility of re-identification. In practice a lot keeps the order lines, the timing and the region, and replaces each customer with a stable code, so a buyer can see repeat purchases without knowing who made them. Card numbers and payment tokens come out entirely. The de-identification standard sets out how records are cleaned before they leave.

Sensitive purchases. What a person buys can reveal their health: supplements, medical devices, pregnancy tests, mobility aids. Washington's law reaches health data held by businesses that HIPAA does not cover, and selling identifiable consumer health data needs a signed authorization naming the buyer. A store that sells such products should leave those order lines out, or make sure no customer can be identified from them.

Business customers and suppliers. A wholesaler's orders name its client businesses and the prices it agreed with them; its purchase orders name its suppliers and their costs. These are business facts more than personal data, but they are often confidential under the contracts that set them. A business should read its supply and customer agreements before it puts negotiated prices in a lot, and code the names of the businesses on both sides.

Staff. The approvals, notes and stock movements are made by named staff, and in California the privacy regulator states that employees are consumers with rights under the law. Staff names are replaced with stable codes, so the trail of who approved what still holds together. The guide to whether it is legal to sell customer data covers the general rules.

Taking the records out and preparing them

Take the records out through the platform's own export, reports or API, while the account is live. Polyshares sets out how to count and export each system: for Shopify, CSV exports from Orders, Customers and Products; for Amazon Seller Central, the All Orders report or the Selling Partner API for history; for NetSuite, saved searches for record counts and a full CSV export. A business that is closing should export before it cancels its subscriptions, and the guide to how long to keep business records after closing covers what it must keep.

Before any order or inventory records go into a lot:

1. Name each system and its years. The store platform, the ERP, the warehouse and shipping tools, and from when to when. A business that moved platforms has two exports with different fields.
2. Read the terms in force for those years, for who holds the records, the platform's licence and aggregate, and any rules eBay, Amazon or Etsy set on buyers' data. Note what the platform already holds, so a buyer hears it from you.
3. Join the records into a history. Orders to their returns, support tickets and shipments; purchase orders to receipts and stock; work orders to their results. The joined record is the one buyers name.
4. Keep the decisions. Price changes, discounts approved, stock written off, orders held or cancelled by hand, with the notes that explain them.
5. Leave out what you cannot stand behind. Information about buyers on eBay, Amazon or Etsy, health-related purchases you cannot de-identify, card and payment details, and anything your privacy notice for those years did not cover.
6. De-identify the people. Customers, staff and the contacts at suppliers and business customers: names, emails, phone numbers and exact addresses come out or are generalised, and each gets a stable code so the history still holds together.
7. Describe it in a datasheet: the systems, the years, the number of orders, products and stock movements, what each record holds, what was removed and what was left out and why.

The guide to preparing data for sale covers the rest of the company's systems, the rights check asks the questions above in order, and the guide to proving quality without sharing the data shows how to answer a buyer's questions before anything leaves.

Two worked examples

These examples are illustrative.

A home goods brand selling online. A brand with about 40 staff has sold through its own Shopify store for seven years, and through Amazon for the last three. Its records hold every order, return, discount code and support ticket, its product catalogue with every price change, its purchase orders to two factories, and its stock across a warehouse and a third-party fulfilment centre. This is the record Mercor and Polyshares name, and it clears most size lines. From its Shopify store it exports orders, returns and products, codes every customer, generalises addresses to region, drops payment details and joins orders to their tickets and returns; it switches off Shopify Network Intelligence if it wants to offer exclusive rights. From Amazon it takes only what is its own (its listings, prices, stock and fulfilment), leaving Amazon's customers out until it has read Amazon's terms for merchants. Offered with its email, finance and support records, it is valued as one company record.

An industrial parts distributor. A distributor with about 90 staff has run on NetSuite for twelve years. Its records show every quote, sales order, purchase order, receipt, stock transfer and credit note, with the approvals behind each and the email with customers and suppliers around them. Its customers are businesses, so there is little consumer data, but its prices to each customer and its costs from each supplier are confidential under its contracts. It reads those contracts, codes the names of every business and person, keeps the price structure in relative form where a contract asks for confidentiality, and offers the history of how it quoted, bought and managed stock. Centific names its industry, and Handshake AI and Nyne name its ERP.

Questions people ask

Do AI companies buy e-commerce and order data? Several name it. Mercor lists order records, inventory logs and listings for retail and e-commerce, and Polyshares takes Shopify, Amazon Seller Central, WooCommerce and BigCommerce exports. They buy the order history as a record of how the business runs, with customers taken out.

How much is my order data worth? Mercor's listing for payments and commerce businesses pays $80K to $900K a company for transaction, order, inventory or shipment tables. Across all company record sales the median is $72.5K per company. A valuation of the whole record is the way to a figure for one business.

Can I sell my customer list for AI training? Not as a list of people. AI buyers want the operating record with customers taken out; a list of customers' names and purchases is a different product, usually sold to a competitor, and selling it is a sale under California's law with notice and opt-out duties.

Does Shopify own my store's data? No. Shopify's terms say it takes no ownership of the merchant's Materials but takes a licence to them, and its Network Intelligence uses customer data alongside other merchants' unless switched off.

Can a business that sells on Amazon or eBay license its sales data? Its own part, yes: listings, prices, stock and operations. eBay's agreement bars using information about users without their consent, so buyers' details stay out. Amazon's terms for merchants set their own rules on buyer data, and a merchant should read them in its account first.

Is ERP data worth more than email? Buyers say so. Troveo reports ERP systems in higher demand than email and Slack, and Grepped's instant quote scores ERP above email and chat.

What this means for you

Order and inventory records sell when they show how the business runs, and the store or ERP platform rarely stops a business licensing them. What decides a sale is how much history the records hold, how well they join up, and whether the customers can be taken out:

1. Name what you hold. Orders, returns, catalogue and pricing history, stock, purchasing and fulfilment, and the systems that hold them.
2. Separate your own records from what eBay, Amazon or Etsy holds. Your own store's history is yours; their buyers are not.
3. Read your platform's terms for its licence and aggregate, and switch off any pooling you do not want before you offer exclusivity.
4. Take the people out: customers, staff and the contacts at your suppliers and business customers, with card details and health-related purchases left out entirely.
5. Join the records into a history and keep the decisions in it.
6. Value your data with your store and ERP among your systems, and check the buyer board for the buyers that name your kind of business.

AIData.Broker introduces the holder to the buyers that take what it holds, free for the holder because the buyers pay us, or, where a sale we run earns more, acts as its agent for a fee fixed in the mandate and due only if the sale closes; either way the holder itself licenses its data to the buyer. We run sales for holders based in the United States, the United Kingdom and Canada. For the wider picture, see how to sell company data to AI companies and what a data licensing agreement contains.

This guide is research from public sources, not legal or tax advice.