unumAI says, “We Know What Went Wrong”
While the nation waits to inaugurate its 46th president, one top-of-mind question being aggressively pondered is why traditional survey polls continue to fail at generating accurate predictions the public can trust. According to data science expert Skylar White and his Colorado-based start-up unumAI, the answer lies in new technology.
White is reimagining where and how to connect with voters by observing their online search behaviors in a way that reveals a far more accurate picture of how voters are really feeling about candidates or issues in their communities. And his strategy is demonstrating unprecedented successes. Nationwide, unumAI correctly predicted competitive 2020 Senate races 93% of the time, compared to the 58% accuracy seen in traditional public survey polling – an unmatched low margin of error at just 1.46%. In competitive suburban U.S. House districts where polling predicted another blue wave this year, polling misjudged significantly and unumAI predicted their actual outcomes about five times more accurately.
With the nation’s eyes on the upcoming Senate runoffs in Georgia, current poll data presents a murky picture, frustrating political analysts, constituents, and investors alike. Current polling data reveals significant volatility and lead switching among candidates in both races. The latest show Democrats winning both races. In contrast, unumAI’s data offers far more clarity and a more consistent picture that demonstrates Republican candidates have held leading positions throughout the race, with a four-point lead in both races today. White suggests, “Bad data can not only erode voter confidence, but it also impacts financial markets that are scared of one-party government and are trading off of these predictions.” White adds, “Regardless of party affiliation, having good data is critical to democracy – everyone loses when the numbers are wrong or misleading. People may believe that the whole system is rigged or develop conspiracy theories when their expectations aren’t met.”
Since the inception of public survey polling in the late 1930s, little has been done to keep pace with technology and the onslaught of rapidly changing messaging, fed to a growing, more diverse voter pool. White says, “The industry has failed to reconsider the inherent pitfall within traditional survey polling where people are inclined to compromise their honesty to fit into stereotypes.”
Instead, unumAI’s technology talks to Google to collect search data en masse and then applies proprietary analysis. White says, “People’s internet searches reflect what they do at the ballot box and what they say and feel at home when they’re not fearful of being judged by a stranger.”
By eliminating social desirability bias, White believes his process is more accurate, “For the first time, voter sentiment is collected organically, inclusively, and unbiasedly. We get a better picture because we sample everyone on the internet, not those who choose to participate in polls.”
Ultimately, unumAI is motivated by its desire to restore public trust in government. White says, “Our current process for candidate and issue polling is broken; Americans aren’t being adequately heard. The silent majority need no longer be silent. We sample millions of voters over time, rather than the traditional 500 people over one weekend.” The firm’s new strategy presents a better picture of how people are motivated politically, which White says yields compassion and forgiveness between each other over political differences. “Our strategy validates every perspective as reasonable and relevant, which makes room for healthy debate, and allows for us to strive towards an increasingly more perfect union.”