Using Algorithms to Build a Better Workforce



Introduction

Abby Ludens, vice president of talent management at Mattress Firm in Houston, faced the daunting challenge of hiring and promoting enough gifted employees to keep pace with her company’s explosive corporate growth. Within about six years, the staff had increased to almost 7,000 workers, from less than 1,000.

In New York City, William Wolf, global head of talent development for Credit Suisse, sought new approaches to identifying high-performing employees at risk of leaving in an effort to prevent their departure from the global financial services company.

And at Cigna’s offices in Hartford, Conn., chief learning officer Karen Kocher wanted to figure out what set apart the global insurance organization’s model performers from its lower-rated employees.

To meet these challenges, Ludens, Wolf and Kocher all turned to people analytics.

Human resource executives are increasingly using this emerging science to guide them in making workforce decisions. By harnessing the power of algorithmic systems, they can better identify the recruits who are most likely to succeed, choose which high-performing workers to develop and promote, and retain talented leaders who often spur innovation and create value.

The growth of analytics-driven HR has been dramatic. In early 2015, only 24 percent of companies felt ready or somewhat ready for analytics. Just one year later, that number jumped to 32 percent, according to the Global Human Capital Trends 2016 report from consulting firm Deloitte.

"We’re moving toward a data-centric mindset" in HR, says Jason Taylor, chief scientist for New York City-based Infor Talent Science, which develops analytics software designed to predict which job applicants have the highest probability of succeeding. "We’ve seen lots of organizations making these shifts, measuring performance, identifying behavioral patterns and bringing in the right people based on those patterns," Taylor says. "If you can scale that, not for one person but for everyone, you have an entire organization poised for a high probability of success."

Data and Hard Judgment

Can algorithms produce better hiring results than people can? To answer that question, researchers from the University of Minnesota and Princeton, N.J.-based Educational Testing Service analyzed 17 studies of job applicant evaluations. Algorithms outperformed human decisions by at least 25 percent. The researchers maintained that their findings hold true "in any situation with a large number of candidates," regardless of the level of the position.

Hiring managers "are easily distracted by things that might be only marginally relevant and they use information inconsistently," researchers Nathan Kuncel, Deniz Ones and David Klieger wrote in the May 2014 Harvard Business Review. "They can be thrown off course by such inconsequential bits of data as applicants’ compliments or remarks on arbitrary topics."

For the best results, the authors recommend that organizations first use algorithms based on a large number of data points to narrow the field of applicants and then tap human judgment to pick from a few finalists. Better yet, have several managers weigh in on the final decision.

That’s similar to the approach Ludens took at Mattress Firm, which retained a vendor rather than building its own algorithm. Ludens tapped into software from Infor Talent Science to help her team evaluate applicants’ online assessments based on measurable data from 39 behavioral, cognitive and cultural traits. The job seekers’ characteristics were assessed against the profiles of Mattress Firm’s strongest performers—a benchmark managers used for predicting success in given roles.

"We promoted new hiring managers and hired new employees, and the system was easy and intuitive. We didn’t have to integrate any systems," Ludens says, adding that the applicants’ resumes and assessment scores were easy to access and view.

Ludens’ team recently examined two years of hiring and sales data. The salespeople that the system had recommended were 80 percent less likely to leave and sold 11 percent more product than others who were hired but not recommended, according to Ludens.

"When you’re hiring, it’s tempting to not want to listen to the data," she says. "It can be easy for a hiring manager to want to hire someone who walks like them or talks like them or thinks like them." But because Mattress Firm was hiring lots of workers over a relatively short period, Ludens says, "it was important to have the science behind it. We didn’t want to put our company brand at risk, and we wanted to hire people who were a good fit."

Of course, there’s a human element to hiring well, too. "We still talk to applicants. What we want to do is use the data and information to make better hiring decisions," Ludens says. "We won’t get it right all the time. That’s why it’s important to continue to use our hearts and our people to make some of those decisions."

Diversity and Quality of Hires

Google is a pioneer in using people analytics to create a better workforce. About eight years ago, Laszlo Bock, senior vice president of people operations at the technology giant, assembled a team of Ph.D.s, technologists and consultants to create algorithms that would identify which candidates were most likely to succeed. The team also produced an algorithm that reviewed rejected applications, which helped the company hire engineers that its screening process would have otherwise missed.

Google continues to refine its algorithms to uncover insights into employee retention, engagement and performance and to address diversity issues. Company leaders say they were able to refine internal processes to improve diversity in recruiting and promotions.

JetBlue Airways also created its own assessment program in an attempt to improve hiring decisions and, ultimately, the customer experience. An in-house team used advanced analytics to develop predictive models of performance for reservation agents and flight attendants. The New York City-based airline also works with an external business partner to assess prospective reservation agents using analytics, says Andrew Biga, director of talent acquisition and assessment. He oversees hiring for JetBlue, which has some 17,500 employees.

As a result of the company’s efforts, the quality of hires rose. "We get more customer compliments—it went up dramatically—and fewer complaints for crew members hired using the new method," Biga says. "The big savings is they’re less likely to leave the company and attrition has gone down. We’re getting people with the right motivational fit, so they’re happy in the job. The more engaged you are in the work, the better the performance." He says JetBlue plans to roll out data-driven hiring programs for other positions.

Another advantage of using data analysis to support hiring is that companies reduce the potential for bias in decision-making. "From a compliance perspective, we’ve gotten much better at doing things in a fair and consistent way. It’s a main reason to do something like this," Biga says.

Attrition and Incentives

At Credit Suisse, Wolf built a department of analysts in HR who specialize in using workforce data to aid in recruiting and to reduce voluntary attrition costs, without the involvement of the company’s IT team. Driving down the attrition rate among some 46,600 Credit Suisse employees by just 1 percent a year could save the company up to $100 million annually, he says.

Using insights gleaned from the data it collected, the analytics team developed policies and practices designed to retain workers. For example, Wolf’s team learned that filling open positions with internal candidates had been an effective tool for retaining employees in the past. So it developed a "Grow Your Own" campaign, which encourages those responsible for hiring to "look inside and give the right of first refusal" to current employees. HR professionals also contact employees with the appropriate skills and potential to ask if they’d be interested in the open positions.

"Five years ago, we only looked inside about 40 percent of the time, and now it’s up to 80 percent," Wolf says. "If I have a vice president position open and I’m growing my own, I’m pulling people up and giving them new responsibilities. We know that internal mobility is really attractive to people."

Each year, staff at Credit Suisse processes data involving 80 variables and 30,000 employees to help predict who might jump ship. The information collected includes workers’ performance ratings, their bosses’ ratings, their compensation changes from year to year, and the length of time in the job without a promotion.

The model "shows you who is at risk [of leaving the company], and it tends to be correct," Wolf says.

The organization generates reports for managers showing the relative flight risk of their employees. The managers then can give workers a raise, a promotion, access to career development training or something else to ward off a departure.

"The model is a better predictor of [a person] leaving than managers are of their own people," he says. "Analytics help you better understand what’s happening."

Gamification and Insights

At Cigna, Kocher used a gamified approach to analytics to help assess managers’ performance and upward-mobility potential. Palo Alto, Calif.-based company Knack supplied the games, which are played on a mobile phone and help to identify the skills, abilities and competencies at which the players excel, as well as the qualities they might consider developing. This is done by looking at their decision-making, actions and reactions in the context of the games.

Among the behavioral patterns evaluated: problem-solving, quick thinking, logical reasoning, executive presence, learning agility, risk-taking and innovation.

"We looked for people demonstrating the highest degrees of performance levels and pulled a subset. We had them play the Knack games to see about their ways of doing things," Kocher says. "From that, you form a profile."

The results can be eye-opening. At Cigna, for example, "What’s come back has been more insightful than we expected," Kocher says. "The Knack data gives you information at a granular level, and it becomes so much clearer what needs to be developed within the group and how it should be handled."

Knack’s analytics software brings up a dashboard that shows where employees in various groups excel or need improvement. The information enables Kocher’s team to blend "the qualitative with the quantitative" in looking for competencies that align with what the company is searching for in employees, such as perseverance, customer focus, willingness to take risks and interpersonal savvy. Now the group is using the Knack games for much larger groups to validate earlier findings and for quantitative insights.

"We’re looking at the characteristics of high-performing, effective managers. We have 4,000 managers around the world," Kocher says. "We’re hoping to better understand what it takes to be a higher-performing manager." Her team uses the information to assist people in developing their management skills based on the profile of top performers. "We’re not only having individuals be more successful with their teams; as an organization, we’re driving strategy forward."

AXA, a New York City-based financial services business, funnels considerable resources toward management training and development. But correlations linking the data from Knack, one of two talent analytics companies AXA hired, to employees’ behavioral patterns revealed surprising findings.

"We found we were lower than expected in certain skills as a company—in particular, learning agility, an attribute connected to learning new skills," says Rino Piazzolla, a senior executive director and chief human resources officer. "That was a surprise for us because we always spend a lot of money on management training. This is helping us rethink career development."

Costly technology and infrastructure changes aren’t necessary for companies to benefit from the broad insights and business value that analytics can provide, according to Piazzolla. "This isn’t about IT and infrastructure," he says. "What you need are different types of people in HR. The biggest challenge in the HR function is to have people with a much stronger analytical background who understand numbers and draw correlations and insight vs. the traditional HR professional who is all about feelings and perception and things you don’t measure." "Put aside your belief system for a moment. Get engaged," Piazzolla advises. "Decide collectively on this new way of doing things. We want to create HR services that are more meaningful to our people in the future. We want to be a successful company tomorrow. This is all about tomorrow." 

Carole Fleck is the content director for Diabetes Forecast magazine.
Illustration by Randy Pollak for HR Magazine.