The market, in public numbers

What has actually been paid for data

Every figure on this page is a public deal or a company's own published statement, with who reported it. We keep it here because almost everything written about “selling your data” is guesswork, and because an owner deserves to see the market before deciding anything.

Last verified 9 September 2026. Nothing on this page is an offer, a valuation, or a prediction about your company.

Seven deals on the record

Google — Spirit Airlines' internal operating data

$10,000,000

Reported by Business Insider, August 2026; sold through a bankruptcy-court auction.

OpenAI — News Corp archives

$250,000,000+ over five years

Widely reported licensing agreement.

OpenAI — Reddit

~$70,000,000 a year

Reported terms.

Google — Reddit

~$60,000,000 a year

Reported terms.

Five AI buyers — Shutterstock library

$104,000,000 in one year

Disclosed in the company's own 2023 results.

Amazon — The New York Times

$20–25,000,000 a year

Reported terms.

OpenAI — Dotdash Meredith

~$16,000,000 a year

Reported terms.

These are large publishers and one airline — shown for scale, not as a comparison to a local company. What they prove is simple: data that records how something real was done is now a purchased asset, and the buyers are named companies with budgets.

What the buyers are short of right now

Human, first-hand records of skilled work. Leading AI labs are each reported to spend on the order of a billion dollars a year on human-produced data, in a category growing more than 50% a year (Dealroom, July 2026).
Simulated jobs with a grader — the “does the model actually do the task” test. Interviews across eighteen people at the firms building them put contracts at six to seven figures per quarter, individual graded tasks at $200–$2,000, and replicas of real software tools at roughly $20,000 each (Epoch AI; SemiAnalysis).
Expert judgment, by the hour. One of the large data marketplaces publishes an average contracted rate of $113 an hour across its experts, and says it pays out millions a day.
Physical work on video. Programs pay most for hands-on, mobile or spoken work, because that is the part of the economy no dataset has ever captured.

Why small businesses have been left out

Every published program we have read has an effective floor around twenty to thirty employees. A six-person shop is under the minimum everywhere — not because its records are worthless, but because buyers price on volume multiplied by uniqueness, and one small company is thin on the first half.

What we are doing about it

Pooling has already worked in every other data category:

Thirty US health systems pooled de-identified records covering 130 million patients; investors and seventeen of those systems put $320,000,000 into the venture at a valuation above a billion (Reuters, January 2025).
Pooled hourly-workforce data from more than 100,000 small businesses is now cited by the St. Louis Fed, the Wall Street Journal, CNBC and Barron's.
A location-data company sourcing from hundreds of app publishers crossed $100,000,000 in annual revenue (TechCrunch, August 2024).
More than a thousand financial institutions have their data monetized through a single intermediary.

Nobody has done it yet for main-street operating records — the dispatch board, the estimate history, the intake sheet, the job photos. So we are asking the buyers directly, in writing, whether they will buy one dataset representing roughly forty same-vertical small businesses, and at what price. If none of them will name a price, we will say so on this page rather than sell a category that does not exist.

What we will never do

Hold, copy, store or resell your company's data. Our job ends at the introduction.
Move customer records, patient information, student information, or anything privileged.
Name your company publicly before you have signed something.
Promise you a number. The buyer decides what it pays, and most companies are declined.

See where your company lands Ask for a free review