Frida Polli (MBA 2012) came to HBS as a neuroscientist in search of a business problem to solve, and she found it during the recruitment season. All of these companies and job-seekers searching for hints of a good fit in handshakes and CVs—it seemed so, well, unscientific to her. Besides, the world was already awash with sophisticated algorithms employed to match people with the perfect Amazon purchase, Netflix binge, or Friday night date. Why hadn’t anyone applied these models to match people to their perfect career?
• Polli matches characteristics to career paths on the Skydeck podcast
Polli launched Pymetrics in 2013 to fill this gap. A PhD in neuropsychology who spent a decade as a researcher at Harvard and MIT, Polli started working on the company while at HBS, building a series of games—used for decades in neuroscience research experiments—with former research colleague Julie Yoo to assess cognitive and emotional traits. The games didn’t ask personal questions, they measured responses, providing objectivity in a way that the traditional Myers-Briggs personality test could not. “The analogy I always like to give is, ‘Would you get a better read out of someone’s weight if you asked them how much they weigh, or if you just put them on the scale?’ ” says Polli. The online game interactions are precisely timed, so they also offer greater depth of data. The difference between a normal questionnaire and the games, Polli notes, is like the difference between the scattershot of a dot matrix printer and the precision of a laser printer.
HBS students were the first data source; stats from their game-play helped Polli and Yoo to build their algorithms and career models. “We were able to really tell the difference between consulting and finance and marketing and entrepreneurship,” says Polli. Career paths in accounting or consulting, for example, usually correlated to attention to detail and a lack of impulsive behavior, she says; high-performing salespeople tended to exhibit the opposite.
Illustrations by Marcos Chin
Hiring companies employ Pymetrics to source additional talent and filter massive applicant pools as well as to boost retention and serve as a development tool. “We follow the talent life cycle—everything from bringing someone in and even off-boarding,” says Polli. “If somebody is looking to let go of an employee, we will give them career matching and help them connect to other opportunities. It is important to have a solution that will basically work from start to finish.”
The company has grown rapidly—expanding from 12 employees last year to a current tally of 25—and counts about two dozen clients. Among them is consumer goods giant Unilever, which has seen a 75 percent drop in hiring time since it started working with Pymetrics. “It used to take them four months to make a decision, now it takes four weeks,” says Polli.
Still, Polli regularly runs up against something she calls “algorithm aversion.” In other words, “I am a human, and I do a better job than an algorithm.” She gets it—HR professionals are worried about their jobs. But that’s not her aim. “We are going to free up your time to do much better, more high-value activities than scanning a résumé for six seconds—which is what the average recruiter does,” she says.
But just as important as efficiency to companies like Unilever is the diversity of applicants that Pymetrics delivers—a trait unique to the company, says Polli, because its technology is bias-free. While countless studies have shown that résumé reviewers often exhibit bias—consciously or unconsciously—against women and minorities, Pymetrics’s matches arrive without any demographic data. “And there are no known gender or ethnic biases in the traits that we measure, so we end up with neutral algorithms,” says Polli. “Which means [the algorithms] will recommend an equal number of men and women and a group that is proportional across ethnicity.”
This is something, she notes, that an algorithm can just do better. “We like to say that we are more humane than humans,” says Polli.
From Baker Library:
Class of MBA 2012, Section D