Data broker vs data marketplace vs data transaction layer: what is the difference?

The short answer: who controls the data and the deal

The three models differ on one question above all: who holds the data and who runs the transaction. A broker holds its own copy and sells it. A marketplace hosts listings and leaves most of the deal to the two sides. A transaction layer works for the company that created the data and manages the license from first review to final payment, without ever owning the records.

That last model is how SourceX works. It sources companies that hold valuable operational records, reviews their rights, helps them inventory what exists, agrees one all-in price with them, presents the opportunity to AI labs and data buyers, and handles contracting, delivery and payment. SourceX does not train AI models and does not resell information it has gathered about people.

The labels matter in practice. When an owner hears the phrase sell your data, the picture that comes to mind is usually a broker trading consumer profiles. A partner who can explain the difference in two sentences removes the biggest early objection.

Side-by-side comparison of the three models

Marketplaces differ widely, so treat the middle column as a general pattern rather than a description of any one platform.

How privacy law looks at the difference

Privacy law cares about what a business does with personal information, not what it calls itself. Under the California Consumer Privacy Act, consumers can opt out of the sale or sharing of their personal information, as the California Attorney General's CCPA overview explains, and the statute defines both terms and requires a written agreement limiting use when a business sells or shares personal information. A platform that calls itself a marketplace can still be selling personal information, while a company licensing redacted work records sits in a different position.

AI training adds its own scrutiny. FTC staff have warned that quietly changing terms of service or a privacy policy to permit sharing data for AI training could be unfair or deceptive. That is one reason SourceX focuses on business records the company created, treats mainly consumer personal data as a red flag, and agrees redaction rules before any work begins.

This is general information, not legal, tax or financial advice. Confirm with your own counsel, tax adviser or professional body before acting.

When each model fits

Use these rules of thumb when an owner asks which route makes sense:

• If the company's main asset is personal data about consumers, it is not a fit for SourceX, and it needs privacy counsel before any data deal.
• If the company already runs a data product with its own sales, packaging and support, a marketplace listing or direct licensing may suit it, because it can do the work an intermediary would otherwise do.
• If the company is a very large content platform, it may license directly. Reddit, for example, disclosed in its 2024 IPO registration statement that it had entered data licensing arrangements in January 2024 with an aggregate contract value of $203.0 million over terms of two to three years. That is a multi-year contract total, not annual revenue, and it reflects a company with in-house legal and data teams.
• If the company is an established operator with years of internal records across many systems and no data team, a managed transaction layer fits, because the hard parts are rights review, inventory, pricing and delivery rather than finding a listing page.

Most companies a professional advisor knows fall into the last group.

How a data transaction layer handles a deal

The data transaction layer explainer covers the concept. In practice, a SourceX deal runs in this order:

1. A partner introduces the company, or the company applies directly.
2. SourceX qualifies it on size, history, data breadth and rights.
3. The company completes a data inventory: which systems, how many years, what can be exported.
4. The company agrees price and terms; nothing is binding until it signs.
5. AI labs and data buyers review the opportunity; once a company is deal-ready, buyers typically respond within about two weeks.
6. The deal closes, the data is prepared under the agreed redaction rules and delivered, and the company is paid.

The full sequence, including where partners step back, is on how it works.

What to tell an owner who asks whether this is data brokering

Keep the answer factual and short:

Owners who want to know what a buyer can do with licensed records afterwards will find the page on who owns an AI model trained on licensed data useful. For how to open the conversation in the first place, see warm introductions vs cold outreach.

Who the managed model is built for

A transaction layer only works for companies with something substantial to license. The baseline is a US company that has had 50+ full-time employees at peak (contractors excluded), several years of documented operations, clear rights to its records and an executive who can authorize the process. Strong candidates tend to keep records across 10-15+ systems, from email and Slack or Teams to CRM, finance, support and engineering tools, and long or archived histories help.

It is not built for consumer data sets, records that belong to a company's clients, protected health information without authorization or de-identification, or data already licensed for AI training. Run a preliminary check with the company fit checker or read who qualifies.

Next step

If you know an owner who has been put off by the idea of selling data, register as a partner and make the introduction, or point the company to sourcex.si/apply with your referral link.