Why New Hiring Algorithms Are More Efficient - Even If They Filter Out Qualified Candidates
Ironically, this ease in applying is the reason why the process is so frustrating for both sides of the job application process. Qualified jobseekers never hear back and recruiters don't know how to sift through the hundreds of thousands of resumes they regularly receive.
"The Internet has democratized the entire application process," says job site Bright.com CEO Steve Goodman. "Anybody can go online and spray and pray their resume all over the place."
That's why it's actually OK if increasingly complicated algorithms accidentally filter out some qualified candidates in order to identify the really good ones, Goodman tells Business Insider.
Launched in early 2011, Bright.com works almost like a dating site, using data and algorithms to match candidates up with potential jobs and hiring managers with star performers. The company currently has 30 million job descriptions and about 10 million job seekers using the site.
Similarly to when dating sites "match" up singles based on common interests, Bright.com offers a "Bright score" to both jobseekers and employers. This score incorporates hundreds of variables, including education, prior employment, and skills listed to infer other skills that may not be listed. For example, if you're in public relations, the company assumes that you're also a good public speaker and speech writer even if you don't actually list those skills on your resume.
"We take your resume and build a bigger resume around it," says Goodman.
The algorithm also picks up on specific employer's hiring practices to come up with patterns. For example, the algorithm knows if Citi likes to hire people who attended Columbia University or worked at JPMorgan, but typically doesn't hire people from Morgan Stanley.
"Over time, the algorithm learns," says Bright.com's Chief Scientist David Hardtke. In much the same way that Google learns what their users' interests are over time by documenting search history, Bright.com analyzes the hiring patterns of employers to predict the type of candidate that employer would most want to hire.
With all of the gathered information, the algorithm then assesses each candidate's fit for specific available jobs and provides a numerical assessment on a scale from zero to 100, with a score of 70 meaning the candidate is minimally qualified. The database finds the highest score matches and provides the information to both jobseekers and employers.
Perhaps problematically, this filtering system also eliminates some qualified candidates. For example, a candidate who never worked at JPMorgan or went to a non-preferred school might be given a lower score for a job with Citi even if they would be a good fit.
"Perfect is the enemy of good," says Hardtke. "It turns out that, yes, we miss 35% of the good matches, but compared to what's out there now, it's way, way better. What do companies do now? Deloitte has turned off all recruiting except for new Ivy league graduates and internal referrals. The people they look at is limited to these two groups. That's eliminating way more than 35%."
Hardtke says this filtering system is also better than companies using keywords to eliminate candidates, which weeds out an "enormous" group of qualified candidates.
"People do fall through the cracks, there's no question about it," says Goodman. "But people don't fall through the cracks with every job. They fall through the cracks with one job here, one job there."
Goodman says that as long as the algorithm is able to make the "whole hiring process more efficient," it's OK that a few qualified candidates were eliminated from jobs they would be a good fit for.