Don’t sell your data without a broker.

I spent years in affiliate networks and payment processing. The job was the same shape. Merchants had offers. Affiliates had traffic. Processors had rates and rules. Nobody on either side saw the full set of options.

I had a wider view because I was already in the conversations. People paid a cut because the alternative was taking the first number that arrived, or locking a deal that closed off better ones later. Charm got the meeting; the cut stuck when the net result was better, or when a bad term was caught before it was signed.

The first offer is not price discovery.

Data owners are now getting cold emails from companies buying operational records for AI training. On October 9, micro1 announced a commitment to spend $1 billion over the following 12 months acquiring and licensing enterprise operational data. Some owners will treat the first serious number that lands in their inbox as the deal.

There is no transparent price sheet for a specific private archive. Public examples can provide context, but the rights, scope, quality and use case behind them may be very different. An experienced buyer may have seen many transactions while you have seen one offer. That information gap is a reason to test the market, not assume the buyer has already done it for you.

Find the buyer who wants your archive.

The work starts with mapping who is actually buying. A lab, an evaluation company, a vertical startup and an aggregator may want different records, rights and delivery arrangements. The buyer with the most aggressive outbound is not automatically the one with the highest willingness to pay for your specific archive.

In affiliate, the offer with the loudest creative was not necessarily the one with the best economics once you looked at the full terms and the residual. Sequencing who sees what, and when, serves the same purpose here: it gives you a basis for comparing offers rather than letting the first buyer define the opportunity.

Run a process. Not a file dump.

The process starts with a high-level teaser, followed by controlled disclosure where authorized, appropriate confidentiality terms and a clear timetable for serious parties. No one needs the full archive on the first call, and an NDA does not replace the review of what you are entitled to share.

Sending everything early, granting broad access or signalling that you must close this week can leak leverage. Respond promptly, but understand why each disclosure is needed. I watched the same thing with tracking links and merchant deals: moving fastest did not necessarily produce the better residual.

Find the problems before the deal finds them.

Diligence is where a proposed deal can get repriced or die. The questions are practical: who can actually license the data, not just who has the files? Are customer information, contractor IP or other third-party rights mixed in? Does a financing agreement or security interest need review? Will de-identification preserve the signal the buyer wants, and what restrictions apply to sensitive records?

A broker can coordinate the appropriate legal, privacy and technical specialists, not replace them. Finding a problem in week two is cheaper than finding it in month four.

Look beyond the headline price.

Price is one line. The agreement decides what you give up. That includes exclusivity, the records and uses covered, duration, resale or sublicensing, future deliveries, liability, change of control and audit rights. Retention and deletion need particular care in AI deals: deleting delivered files should not be confused with undoing training already performed.

The first draft contains the terms the buyer prefers. In payments and affiliate, the rate was never the whole deal. Holdbacks, rolling reserves, exclusivity on certain traffic and the right to swap or add offers later decided whether you had options left. An exclusive data license deserves the same scrutiny: it can be the right trade, but it may also prevent later licensing in a valuable category.

Operational records can sometimes be licensed more than once, or narrowly enough that other uses remain. Structure can matter more than the first number, though the options may be limited if there is only one credible buyer. The point is to understand which case you are in before you sign.

Not every owner needs an intermediary. If you already understand the buyers, have a well-defined scope and good counsel, you may be able to run the process directly. What matters is not confusing an offer with a view of the market.

The broker has to earn the cut.

A broker may charge a success fee, a retainer or a combination. The structure shapes incentives; even a commission-only arrangement can reward closing quickly rather than securing the best result for you. The test should be whether the expected outcome, after fees, time, legal costs and the rights granted, is better than taking the untested first offer.

If that isn’t true on your deal, don’t hire the broker.

The same test applied when friends in affiliate gave me a cut. They were paying for a wider view of who was buying and for someone who could catch a bad term before it was signed—not for personality.

Check the market before you close it off.

The first offer can be the right offer, but it is not automatically the only one worth seeing. Checking takes time and costs money; skipping that work can cost you a price—or rights—you never knew you could negotiate.