Denver Business Journal | Denver tech firm tries to solve political-polling errors by ditching surveys
Denver Business Journal | Jan 11, 2021
By Ed Sealover – Senior Reporter, Denver Business Journal
Skylar White had majored in political science and spent his career in data sciences, but he was blindsided by President Donald Trump’s victory over Hillary Clinton in 2016 at a time when polls predicted a very different outcome.
Rather than just wring his hands, White dug in, especially after seeing how wrong polls were again for Mike Johnston’s gubernatorial campaign on which he was working in 2018. (Polls showed Johnston getting about 9% of the vote when he captured 23% in the Democratic primary.) And his work has led him to launch a new kind of polling firm that captures voter sentiment without ever picking up a phone and dialing anyone.
His firm, unumAI, predicts outcomes based on Google searches of candidates, which White argues is a much better indicator of voter sentiment at a time when voters are increasingly distrustful of pollsters and when more calls from survey firms are blocked as spam. And his methodology led to the correct outcome in 93% of swing U.S. Senate races this fall and 58% of swing U.S. House races — the latter an outcome that was twice as accurate as more traditional polling firms, according to his research.
The methodology isn’t perfect; just last week, he predicted four-point victories by Republicans in both Georgia U.S. Senate races, which were won by Democrats. But after a second presidential-election cycle in which many polls were off, he and his partners are banking they can grow business not just with political campaigns but with corporate leaders who need to know about public sentiment to judge their investments or product-release plans.
“The applications are really boundless, and now we are starting to test this out in many other ways,” he said in an interview.
White has spent time in and around politics: The University of Denver graduate had a father who served in former Gov. Dick Lamm’s administration and himself spent time working in former U.S. Sen. Mark Udall’s office. But he also worked for Deloitte in the national security and government efficiency space, understanding data gathering as he dealt with clients such as the Central Intelligence Agency and the U.S. Department of Defense.
Gathering data from millions of Google searches, as opposed to cold-calling some 500 potential voters, reflects a greater depth of sentiment and also offsets the increasing tendency of U.S. voters to lie to pollsters because they don’t trust them, White said. He then adds in the historical components of partisan voting patterns and fundraising and produces outcomes that he believes reflect the “silent majority” of voters that many polls don’t capture now.
White and two of his DU master’s program classmates launched unumAI in 2019 with the expectation that campaigns would want this new form of research, but as the coronavirus pandemic took hold, he found that many hunkered down and wanted to stick with traditional methods. He doesn’t believe that hesitancy will exist going forward at this point, however.
In addition to selling its results, unumAI can sell its costs — about $5,000 to $10,000 for the amount of research that goes into a poll that typically runs anywhere from $15,000 to $60,000. And he noted that the same type of research can analyze candidates’ and officeholders’ social media to understand what issues their constituents and voters are talking about, which can help them pinpoint messaging.
In addition, the two full-time employees of unumAI are beginning to pitch their service to private-equity firms, to understand, for example, the voter risk around industries that receive incentives or other government funding. Searches can predict outcomes such as the share of energy that consumers will seek to get from renewable or traditional sources and the share of vehicle sales that various types of engines will take too.
So, is this a time that political campaigns and corporations are ready to take a chance on new methods of understanding public sentiment after a half-century of traditional surveys? White believes that the next two years will give him a lot of insight into that question — and into the future of his business.
“I think that campaigns should see that key need to be on the cutting edge with technology or face peril,” he said. “We’re trying to find how we fit. We need more partners. We need more testing. We need more real-life projects. But we have something to offer.”