Transcript
One data program's page says, in big letters: earn a hundred thousand dollars to four million dollars and more. Scroll down the same page, and its own calculator says your data could be worth up to a hundred thousand. Both were checked on the same day. So which number is your data worth?
Let's start with three ways to put a number on data, because you'll hear all three used.
Data is valued three ways. The cost approach asks what it cost to create or replace. The market approach asks what comparable deals paid. The income approach asks what license income the data can earn over the term. Each gives a different answer, and none of them is a price list.
Start with cost. What did your tickets, documents or code cost to produce, and to clean up for a buyer? That sets a floor for negotiation, but not a price. And you won't find it in your accounts, because an international accounting standard keeps internally generated archives off the balance sheet.
Next, the market approach. Few deal values are disclosed. One health data company's filing states two hundred million dollars over three years from one buyer, with fifty million upfront. A court recorded a ten million dollar winning bid for a bankrupt airline's de-identified data. That sale wasn't approved as of November ninth.
Here's the catch. Comparables travel badly. Each deal prices one data type, one rights package and one term. Clinical records and book titles set no rate for an archive of support tickets or email. So a big headline deal tells you little about yours.
The income approach looks at what the data can earn over the term, and filings show how uneven that is. A big social platform says substantially all its data license value comes from two partners whose agreements are approaching the end of their terms. An academic publisher's AI revenue for one quarter fell from twenty-eight point nine million dollars to thirteen point seven.
A stock media company shows it clearly. Its data, distribution and services revenue was down seventy-six percent in the third quarter of 2026 against a year earlier, and its data offering was down ninety-one percent. Data income can drop fast after the first term, so a past deal says little about a future one.
Now, the part where most owners get misled: claims versus deals.
A buyer program's payout range is that buyer's own claim about its own offers. A disclosed deal value is a figure a party filed, signed, or a court recorded. They look alike on a page. They aren't the same kind of evidence.
What about the calculators? One estimator returned four different ranges for the same inputs on four page loads. A broker's calculator labels its output a scenario, not a valuation. A calculator shows one buyer's illustrative range for its own program, not what your data is worth.
Payout figures come in four levels. First, an advertised range, which is a sales message. Second, a claimed paid-out total, the payer's own count. Third, a contract value or a bid on the record. Fourth, an amount the recipient says was received. Only that last level shows money actually paid.
So when you see a total, ask. Paid to whom? Over what period? Gross or net? Cash or equity? One big deal, or many small ones? Most published totals leave these open, which is why you can't compare one with another. And as of November ninth, no seller had publicly confirmed an amount received under the company data programs.
So what actually moves the value of your data?
Six drivers move the value: rights, consent, personal data, exclusivity, uniqueness, and volume with history and freshness. The first three can take it all the way to zero.
Rights come first. Data is worth only what you can license. One program's master terms have the seller warrant it hasn't granted anyone else the same rights on similar terms. And customer contracts can bar a license altogether, which removes that data from the sale instead of lowering its price.
Uniqueness matters too. One intermediary says about thirty-seven percent of the raw code offered through its service is removed in cleaning, as third-party or generated code and duplication. Duplicates and borrowed material don't get paid for.
Exclusivity comes in three forms. Exclusive by default. A time-limited exclusive window, such as ninety days. Or non-exclusive with a no-prior-grant warranty. No party has published an exclusivity premium, so the price effect gets set in negotiation.
Now, say an offer lands in your inbox. How do you read it?
First, an estimate isn't an offer. Buyers label their calculators indicative or illustrative. Only a written offer tied to an order form states a price the buyer proposes to pay. And a written offer usually follows a sample of your data.
Every offer comes down to four things: the price, the payment trigger, the review window, and the grant of rights. Plus the conditions that are still open. No program publishes its pricing formula, so you read the terms, not the math.
Offers price your data in four ways. A fixed sum. A fee per delivered record. An upfront fee plus a revenue share. Or a revenue share alone, with nothing upfront.
Then ask when the money arrives. Buyers publish six triggers. At signing. On delivery. On acceptance. After invoicing. After approval plus net thirty to sixty days. Or on each resale. Same price, very different wait.
The review window matters as much as the price. One program's master terms give thirty business days after delivery to reject. After that, the data is deemed accepted. A refund clawback survives acceptance if your data manifest is materially inaccurate.
The same price can buy very different rights. One grant is nonexclusive, irrevocable, perpetual and sublicensable. Another is exclusive by default, with a kill clause, and the buyer says trained models cannot be unlearned. A third is a ninety-day sole selling window.
And the headline price isn't what you keep. Some programs say they cover extraction, anonymization and packaging themselves. A data marketplace charges listing fees of one and a half to three percent. One broker is paid by the buyer when a license closes. Ask who pays for what.
Before you answer, put questions to the buyer, and get the answers in writing. Which unit and quantity is the price based on? Who pays for extraction and de-identification? What triggers payment, and on which date? How long is the review, and what does silence mean?
Then four more. What survives acceptance, such as a clawback? What are the scope, term and exclusivity of the grant? What happens to the data and the trained models when the grant ends? And which conditions are still open?
You have three answers. Accept. Counter on terms. Or decline. Declining is right when the terms cost you more than the price pays. You can also ask other buyers for written offers, or license a narrower set of the data.
One more warning. An unsolicited offer that names a large sum before anyone has seen your data is a reason to check who is offering before you reply.
So how do you reach a number you can defend? Inventory your data. Check rights and consent. Set the de-identification scope. Describe it in a manifest and send it to more than one buyer. Then compare written offers on the same terms. That beats any calculator, and it gives you something to negotiate with.
So, in one line. Value your data three ways, treat every advertised range as a claim, protect your rights, and compare written offers on price, payment trigger, review window and grant. Not selling is a valid result too.
Next up is lesson A6, The four terms that decide what you give away.
Read and watch the full lesson free. Enter your email to unlock every lesson of the academy.