Truthset.

There is a number that everyone in advertising knows and almost nobody says out loud, which is that a large fraction of the consumer data being bought and sold - the stuff that decides which ad you see, and which you don't - is simply wrong. Not stale, not fuzzy. Wrong. Independent measurements put the error rate at up to 60%. And here is the fun part: after that data gets "onboarded" into the plumbing that actually runs a campaign, it loses roughly another 40% of its accuracy. So a segment that was 55% right when you bought it can arrive at the finish line about 33% right, which is to say slightly worse than a coin you paid a premium to flip.

This is the market Truthset decided to build a business inside of. The company, founded in Oakland in 2019, does something that sounds almost too plain to be a startup: it tells you how likely a piece of data is to be true. It does not sell the data. It is emphatic about this - Truthset is not a data broker, does not want to be one, and treats that distinction as the entire point. Its product is a verdict, not a list.

The verdict is called a TruthScore. For any given attribute about any given person - is this individual male, does this household have kids, does this person own a car - Truthset produces a number from 0 to 100% representing the likelihood that the attribute is actually accurate. Seventy percent chance this person is male. Forty percent chance that email belongs to a homeowner. It does this across billions of data points, covering about 95% of the U.S. Census population, and it does it as an outsider - the referee, not one of the teams.

The Nielsen Lesson

The referee framing is not an accident, and it is worth understanding where it comes from. Founder and CEO Scott McKinley spent years at Nielsen, the company that rates television audiences, and he took away a very specific business lesson: the entity that rates a market can be worth more than most of the companies competing inside it. "I was at Nielsen, so I learned how valuable it is if you're the company that rates things," he has said. "Nielsen had a market cap when it got taken private again at $16 billion for rating a $70 billion market. That's pretty incredible."

The bet embedded in Truthset is that consumer data deserves the same treatment. Marketers spend enormous sums on audience data, and the quality of that data is - to borrow McKinley's phrasing about the industry's polite euphemisms - "basically pure guessing, with an incentive to maximize scale at any cost." If you are a data provider, more names in your file means more to sell, and accuracy is somebody else's problem. Truthset's proposition is to make accuracy everybody's problem by making it a visible, comparable score.

Turning Accuracy Into a Grade

The clever move - the thing that turns a research methodology into a product people can buy - is that Truthset compresses all this probabilistic scoring into letter grades. AAA, AA, A, B. It is, quite deliberately, the visual language of credit ratings, and it does the same job: it lets a buyer who does not want to read a statistics paper glance at a segment and know, roughly, how much to trust it. A brand that needs precision buys AAA. A campaign chasing reach can slum it in the A tier and know exactly what it is trading away.

In April 2025, Truthset packaged this into a product called Data-Rated Audiences, which lets brands and agencies pick accuracy-rated segments from the providers in what the company calls its Data Collective - more than 20 leading data companies who have agreed to be scored by the same outside referee. That last part is the quietly impressive achievement. Getting one data provider to submit to independent grading is a sales call. Getting twenty competitors to agree on a common scorekeeper is closer to diplomacy.

Why Anyone Plays Along

The obvious question is why a data provider would volunteer for a report card. The answer is that in a market where everyone claims their data is accurate, a provider who can prove it has something the others don't. A Truthset Certification badge means a provider was actually vetted and scored - which is only worth anything precisely because a provider could fail to earn it. Trust marks that everyone gets are worthless; trust marks you can flunk are valuable. Providers like 180byTwo, Stirista and Audience Acuity have leaned into "Truthset Scored" as a differentiator, and the company has wired its scores into the broader ecosystem through partnerships with LiveRamp, Magnite and PMG's Alli Marketplace.

For the buyer, the value is more straightforward: less waste. If you know a segment is 33% accurate after onboarding, you can stop paying premium CPMs to reach the 67% who aren't who you think they are. Truthset's whole pitch reduces to a slogan on its own site - "nothing performs better than the truth" - which is the kind of line that is easy to write and hard to earn. The earning is in the billions of records and the patented scoring underneath it.

The Money and the Modest Size

Truthset is not a large company, and it does not pretend to be. Roughly 20 people. About $1 million in annual revenue by outside estimates. In October 2024 it raised a $5 million Series A - investors included super{set}, the venture studio run by ad-tech veterans Tom Chavez and Vivek Vaidya, along with Data Point Capital, Revel Partners and others - bringing cumulative reported funding into the neighborhood of $10 to $16 million depending on who is counting. These are not the numbers of a company trying to win by outspending everyone. They are the numbers of a company betting that being the neutral scorekeeper is a structurally good place to stand, and that the position compounds as more of the market agrees to be measured by it.

Whether that bet pays off the way Nielsen did is an open question - referees only matter if the teams keep showing up to be officiated. But the underlying observation is hard to argue with. An enormous amount of money moves on the assumption that data is accurate, the data frequently isn't, and until recently nobody was independently keeping score. Truthset's answer is not to fix the data or to sell better data. It is to stand off to the side, hold up a number, and let everyone see how much of it was ever true. That turns out to be a real business - and a slightly uncomfortable mirror for an industry that would rather not look.