Xerox runs 175 call centers around the world. In all, the centers employ more than 50,000 customer service agents who deal with questions about everything from cellphone bills to health insurance.
Teri Morse, who is in charge of recruiting all those people, says the company had a problem: It was hiring people who just weren't a good fit.
"People were in the training classes sharing with us that they weren't right for the position," she says. "You have to deal with a frustrated customer, hang up the phone and move on to the next, and not have to excuse yourself to go to the ladies' room and cry."
So a couple of years ago, Xerox hired a company to help the company do a better job of finding the right people
This company, called Evolv, began collecting lots of data about the people applying for jobs at Xerox call centers.
The applicants had to answer extensive surveys with questions like: "Which word better characterizes you: 'consistent' or 'witty'?"
Another question: "Can you name three pieces of computer hardware?"
Applicants were tested on pattern recognition and multitasking. They had to respond to a challenging customer service call.
Some of these people got hired. Not all of them were a good fit for the job. But the data gave the company a sense of which characteristics predicted that hires would be a good fit — and which didn't. Prior experience in a call center, for instance, didn't really matter.
A retail background was a predictor of success — except for people who worked as cashiers or in restaurants. Those people tended to do worse at the call centers.
With these new techniques, Xerox says it has been able to improve its hiring and significantly reduce turnover at its call centers.
Other companies that parse employee data are finding surprising results. Michael Rosenbaum of Pegged Software, a company that works with hospitals, says one piece of conventional wisdom is flat-out wrong: "We find zero statistically significant correlation between a college degree or a master's degree and success as a software developer."
Of course, using data to drive hiring decisions has its problems. Employers guided by data could wind up skipping over promising candidates. But Barbara Marder of the consulting firm Mercer points out that the way companies hire now has its own flaws. We like to hire people who are like us. People who went to schools we know. People who were referred to us by our friends.
"A lot of these new techniques do have the potential to eliminate biases," Marder says.
AUDIE CORNISH, HOST:
From NPR News, this is ALL THINGS CONSIDERED. I'm Audie Cornish.
JULIA KELLER, BYLINE: And I'm Melissa Block. When companies are looking to hire a new employee, they typically rely on a very low-tech tool: the resume. Resumes include the seemingly important information, where the person went to school, their previous employers and the skills they have, but some employers are trying out a new approach to hiring that makes resumes seem antiquated. It uses massive amounts of data and complex algorithms to help predict which job applicants are likely to work out and which are not. Here's Lisa Chow with our Planet Money team.
LISA CHOW, BYLINE: The company Xerox runs 175 call centers around the world, filled with customer service agents who deal with questions about everything from cell phone bills to health insurance. Fifty thousand people work in these call centers. Terry Moore(ph) is in charge of recruiting them. She says a lot of people they thought would work out didn't.
TERRY MOORE: We were spending a lot to recruit and even more to train. And people were in the training classes sharing with us that they weren't right for the position. You have to be able to deal with a frustrated customer, hang up the phone and get on to the next and not have to excuse yourself to the ladies' room and cry.
CHOW: So a couple of years ago, Xerox hired a company to help them do a better job at finding the right people, people who could hack it in the call centers. This company, called Evolve, began an experiment. When people applied for jobs at Xerox call centers, this company gathered a bunch of data about these candidates that went far beyond resumes. All applicants had to answer extensive surveys with questions like which word better characterizes you, consistent or witty. Also, can you name three pieces of computer hardware?
Applicants were also tested on pattern recognition and multitasking, and they had to respond to a challenging customer service call.
UNIDENTIFIED MAN: And I know for a fact that you sneak these charges in because people don't call about $1.10 because it's just $1.10, and you sneak this into people's bills, and everybody pays it. But this is criminal. It's awful. I just want this $1.10 removed from my bill, and I never want to be charged for data usage again.
CHOW: Some of these people got hired, some worked out great, some did not, but now Xerox had a way to tell what exactly should they be looking for when they were screening candidates. And there were some surprises. Moore says prior experience in a call center, for instance, didn't really matter.
MOORE: We've proven that just because you worked in a call center, and possibly even your reference came out well, it doesn't necessarily mean that you are going to be good in the future.
CHOW: On the other hand, if you had a retail background, that turned out to help unless you were a cashier or worked at a restaurant. Those people tended to do worse. With these new techniques, Xerox has been able to significantly reduce turnover. Michael Rosenbaum(ph) works at Pegged Software, which is developing big data techniques to find the right workers in hospitals. He says they've used it on themselves and overturned a basic piece of conventional wisdom, one of the things that people put at the top of their resume: where they went to school.
MICHAEL ROSENBAUM: The question is does a college degree or even a graduate degree tell you whether or not someone's going to be good in a particular job. And we find zero statistically significant correlation between a college degree or a master's degree and success as a software developer.
CHOW: Rosenbaum thinks companies are going to be doing a lot more data analysis like this in the future, which is way better than trying to read the tea leaves in a resume. Barbara Marter(ph) agrees. She works at Mercer, a consulting firm that specializes in recruiting.
BARBARA MARTER: And I think a lot of it will be looking at, you know, people who are performing really well in the job and finding out what is it about them that's making them so successful and equally the low performers or the people who either leave on their own or are asked to leave, what is it about them that makes them, you know, not really well-suited for the job.
CHOW: It is of course possible that data mining will miss some good people, but she says think about the flawed way in which we hire people now. There are all kinds of biases now. People tend to hire people who are like them or went to schools they know or who were referred to them by a friend.
MARTER: A lot of these new techniques do have the potential to eliminate biases.
CHOW: Marter thinks in the near future employers are going to be using a lot more data and relying a lot less on old paper resumes. Lisa Chow, NPR News. Transcript provided by NPR, Copyright NPR.