Breaking Away From Low-Quality Data

Scott McKinley’s Journey from professional cyclist and captain of the 1988 US Olympic road cycling team in Seoul, Korea, to CEO and founder data validation provider Truthset was more linear than one might think.

Cycling is “the ultimate test of truth,” McKinley says on this week’s episode of AdExchanger Talks. “You race 120, 130 miles and whoever gets to the other side first is the winner. It’s very clean, it’s very responsible, it’s very individual.”

But as doping became more widespread in cycling in the 1990s, McKinley took it as a sign to stop.

“Swindlers came, drugs came, and I didn’t want to take part in that,” he says.

After retirement, he worked as a website manager for Cox-owned TV stations in the late 90s before co-founding and leading several measurement and analytics software companies, and later joined Nielsen as EVP of product innovation.

But by that point, about 18 years into his career, something had become abundantly clear to him: digital advertising has a serious data quality problem.

“And I decided, basically, either I was going to leave this industry, because I was so tired of the snake oil salesmen, the obfuscation and the BS,” he says, “or I was going to create a company that would try to clean up the mess a little bit and give everybody a better opportunity of using the data to predict who someone might be on the other end of the device.”

In 2019, McKinley founded Truthset, a startup that verifies the accuracy and quality of datasets.

According to Truthset’s analysis of the public data market, the average accuracy of age data is 32%, meaning that most age data is wrong. Meanwhile, the average accuracy of gender data in publicly available segments is 61%, just a tiny bit better than flipping a coin.

So why do advertisers continue to buy this data?

The ad industry’s “addiction” to size at the expense of precision is one reason, McKinley says.

But there’s also a lot of “pretending” going on in the supply chain, he says, “what we like to call probabilistic modeling.”

“It’s basically a euphemism for pure guesswork, right, with an incentive to increase scope at all costs,” McKinley says. “When you have those two variables, how can you possibly trust what comes out of the other side?

Also in this episode: What happens when marketers use incorrect data, McKinley’s unvarnished take on it ID bridging (he’s, uh, not a fan) and why cycling isn’t the best way to introduce girls to high school.

For more articles on Scott McKinley, click here.