Foot Locker Inc. has never had difficulty finding job candidates. The New York-based company’s multiple brands attract more than 1.5 million applicants across its more than 3,400 locations each year, as plenty of potential retail workers envy the chance to don those distinguished referee-style uniforms.
But like most retail outfits, Foot Locker in the years leading up to 2010 had experienced high turnover at its stores, resulting in muted sales growth and higher training costs.
“We were looking for an increase in performance,” said Alexis Trigo, Foot Locker’s director of store capability. “We wanted an increase in sales per hour. We wanted a positive impact in our customer experience. And we knew that a side effect of that would be that the teams should be staying longer.”
In fall of 2010, Trigo and Foot Locker’s talent management team set out to improve the company’s front-line worker hiring by tapping a discipline gaining traction across the rest of the human resources industry: predictive analytics.
The goal, company executives said, was to use predictive data to narrow down massive candidate pools to pinpoint individuals with strong behavioral profiles ideal for Foot Locker’s retail sales.
Through a partnership with New York-based HR technology firm Infor, Foot Locker implemented Talent Science, a cloud-based predictive analytics software system that helped the shoe retailer streamline the hiring process of its front-line workers.
With Talent Science, applicants take an online assessment at sneakerjobs.com or by visiting the brand’s various jobs extension websites, such as footlocker.jobs. Answers from the assessment then produce a “fit score,” which indicates how closely candidates’ behavior aligns with a job profile.
By using predictive data to help identify ideal in-store worker behaviors, Foot Locker said it was able to reduce its turnover while finding better candidates more likely to stay with the company long term.
“We didn’t want just a warm body who was in our store,” said Robert Perkins, Foot Locker’s vice president of talent management. “We wanted someone who had higher sales capabilities.”
Still, the biggest challenge in shifting to the new data-driven retail-hiring model was convincing the company’s senior leaders that a problem existed in the first place.
With more than a million applicants yearly, Foot Locker’s retail talent pipeline was flowing steadily, meaning its talent management team needed to show the business case for using Talent Science across its brands, which include Lady Foot Locker, Kids Foot Locker, Footaction, Six:02 and Champs Sports.
Predictive Science
While technology has made it more accessible, companies in both the private and public sectors have historically tried to use data to predict human performance, according to David Lewin, managing director and head of the labor and employment practice at Berkeley Research Group, a data analytics advisory.
What first started in the 1940s with psychologists’ interest in the concept of quantification ultimately led to use in the U.S. Army as the institution aimed to use behavioral data to identify top-performing soldiers for its upper ranks. Both World Wars featured nascent use of personality and behavioral assessments to predict worker and soldier behavior.
The advent of mainframe computers in the 1960s and ’70s digitized the experience, Lewin said, making the practice much more accessible. This has only grown more widespread as computers have evolved — from mainframes to desktops to laptops, and now smartphones, tablets and wearable devices.
Similar technological evolution has also played a heavy hand in how people apply for jobs at Foot Locker.
Since 2008, Foot Locker’s job-application process included an online application for full-time workers, Trigo said. But prior to Talent Science’s implementation, the company’s part-time workers — like those working in its stores on the front lines — had to apply in person, with paper résumés and applications.
“The efficiencies we were going after was those part-timers,” Trigo said.
If selected for an interview, applicants were asked behavior-based questions. Trigo said the old questions were fairly generic. Interviews consisted of four questions in four categories — customer service/sales, basic store operations, leadership and interpersonal people development — for a total of 16 questions.
With Talent Science, as few as three questions are asked from a library of several hundred. This shorter list of questions aims to better predict if the candidate holds the ideal behaviors needed for the job,Trigo said.
Applicants are first put through an online assessment. Store managers then see where the candidate sits along a spectrum of four categories based on their answers: recommended, recommended with qualifications, recommended with reservations and not recommended.
Once selected for an interview, Talent Science provides questions tailored to each individual candidate based on their answers to the initial assessment. Three to five questions are generated; store managers then have three additional questions they can ask based on the needs of the store and the candidate’s responses.
But before Foot Locker could fully implement the streamlined process, the talent management team needed to sell the company’s senior leadership on the idea that spending on analytics software was a worthy investment.
The team that advocated the new predictive analytics hiring system consisted of four people, including Trigo and Perkins. Members from HR, finance and sales also gathered to compile analytical findings to support rationale for the program, Trigo said.
When selecting the software to use in the program, the team looked for simplicity, as well as something that would have a validated process to prioritize candidate flow. They also wanted metrics to prove return on investment.
To bring metrics to its business case, the team looked at sales per hour, in-store experiences of customers and retention rates, among others. Another critical component included internal testimonials of store managers.
Perkins said applicant volume and the amount of time it took store managers to comb through them was an additional selling point. With as many as 300 résumés received for each open position, store managers had no systematic way to interview talent. As a result, decisions were often made based on likability or convenience.
“That’s not the best way to make a hiring decision,” Perkins said.
The team also presented a budget and the initial funds for a pilot program. Leadership remained unconvinced.
“Because from their standpoint, again, we didn’t have an issue with candidate flow. We weren’t a company that had a difficult time trying to attract people to come and work at Foot Locker,” Perkins said.
So the team presented predictive analytics’ use as “an opportunity for us to increase our effectiveness, not only in our hiring processes but also in the sales performance that would be generated from the hiring of these candidates,” Perkins said.
Senior leadership was still hesitant. “My leadership didn’t necessarily trust the science behind it,” Perkins said. “It was almost like a drug that’s in the early research stages that’s very promising and people hope and think it’s going to work, but it hasn’t been tested on enough people yet to know for sure.”
The Shoe Fits
Jill Strange, Infor’s director of human capital management and behavioral science, said the science behind its predictive analytics software helps users take some of the decision-making out of the hiring process.
When initiating a new client, Infor determines what’s driving success in its business. It then compares that with what it hopes to attain moving forward.
To do this, Infor has current employees —in this case, Foot Locker’s in-store associates — take Infor’s 210-item assessment that measures 39 behavioral, cognitive and cultural characteristics. When Foot Locker began the program, it used 38 characteristics. Assessments take between 20 and 30 minutes.
“We’re measuring what we call their ‘behavioral DNA,’ ” said Strange, who has a doctorate in industrial and organizational psychology. “Everyone has their own unique behavioral DNA. It’s sort of what makes you who you are.”
Those assessments are then put through an algorithm that compares responses with employee performance to identify which characteristics make them successful. This leads to nearly 40 behaviors that carry varying significance in profiles, including ambition, attention to detail and self-reliance.
‘It [predictive analytics] was almost like a drug that’s in the early research stages that’s very promising and people hope and think it’s going to work, but it hasn’t been tested on enough people yet to know for sure.’
— Robert Perkins,vice president of talent management, Foot Locker Inc.
Ambition, for instance, “measures the degree to which a person is motivated to advance their career,” Strange said. High scorers on this attribute give their maximum effort to advance within a company. People who score low aren’t driven by the need to advance their careers, meaning they could work well in low-visibility roles.
To write these items, industrial and organizational psychologists adhere to tightly constrained definitions of each attribute. Each attribute is then piloted with hundreds of thousands of individuals. Strange said Infor tests 14 million people per year.
Because different factors make different organizations successful, each company has its own unique profiles. For example, Foot Locker profiles are different from its other brands, like Champs Sports, because they are different stores with different cultures.
“The reason why we have different profiles for different brands is that those characteristics that are driving success may differ based upon the organization,” Strange said.
Gaining Support
The talent management team eventually gained full leadership support through an 18-month pilot program starting in July 2011.
About 250 Foot Locker stores and 50 Champs Sports locations participated in the first phase of the pilot, with the second phase adding another 300 stores. Perkins said the pilot stores chosen were the most representative of Foot Locker’s general customer populations.
At the six-month mark of the pilot, the team said it saw double-digit reductions in staff turnover and double digit increases in sales-per-hour. Managers were also reviewing far fewer applicants, reducing résumés seen from as many as 300 to as few as three for each job opening.
Foot Locker’s senior leadership was also sold. The company got the green light to implement the system across all of its stores. “We quickly shifted from ‘Do we trust the science?’ to ‘How quickly can we accelerate rolling this out to as many of our stores as possible?’ ” Perkins said.
The full implementation took eight weeks, Trigo said, with district managers attending three-hour workshops on the new system. Others attended 90-minute WebEx sessions to learn how to navigate the new Talent Science process. Foot Locker conducted the training without third-party assistance.
The intuitiveness of Talent Science made training more of a formality. “It made sense,” Trigo said. “It’s something that didn’t take a lot to train on in the way the information was presented.”
Additionally, Foot Locker’s “Recruitment & Selection: Process Cycle Guide” was updated from its 2008 version. Changes included moving to online applications from paper as well as how to handle the new interview questions used in Talent Science.
The guide is a PDF file within Foot Locker’s internal learning portal. By having the guide online, Trigo said it’s easier to make changes. “More importantly, store managers can just review and print what they need, so they’re not taking up any more storage,” he said.
Three months after the initial training, management followed up with store associates through what the company calls “touchpoint assessments.” These questionnaires helped Trigo and his team determine if it needed a new strategy, which programs were working, and if and where there were knowledge gaps in the process.
Still, despite the system’s perceived efficiency, Foot Locker faced some initial resistance from store managers.
“We had pockets of resistance in the sense of ‘I know best. I’m not going to trust the system. I’m not going to trust the new process,’ ” Trigo said.
Time and patience proved to be the ultimate persuader. Three to six months after using the new system, “they did a radical 180,” Trigo said. “Those individuals that were kind of the resisters became our advocates.”
Talent Science went live nationwide between May and August 2013. It’s now used in about 2,400 of Foot Locker stores, with a goal to be in the company’s more than 3,400 locations, Trigo said.
Pushing Boundaries
Now a little more than two years into using predictive analytics, Foot Locker store managers say they’re seeing great benefits.
Henry Roberts, a store manager at Lehigh Valley Mall in Whitehall, Pennsylvania, said Talent Science has helped him stay more organized because of the reduction in paperwork. He said the flow of the hiring process is also more organized and intuitive.
“If you’re not able to do everything in one day, you can click on the person’s name, and it’ll allow you to see where you are within that flow and then pick up where you left off,” Roberts said.
Shannon Morris, district manager of Southeastern New York, has been with Foot Locker for 14 years. She said she has interviewed at least 300 people since the new system has been implemented and agreed that the new hiring program helps managers prepare better for interviews by providing them with more focused questions for each candidate.
“Not every manager has been around for 14 years,” Morris said. “So if we have a manager that’s only been around for one year, it was a lot more helpful for them to have these more educated questions to ask the associates that are in front of them.”
Although Morris said she puts more time into each interview than before, she hires fewer people thanks to a reduction in turnover. Morris said she’s now able to invest that time into training, which helps keep associates for longer and “focus on their futures with us,” she said.
Morris cited one assistant manager in particular that she hired using Talent Science as a shining example. The assistant manager, who Morris declined to name, had no prior experience with FootLocker, but was able to run a store after onlyseven months on the job.
Morris said the assistant manager worked at a previous retail job seven years and never ran a store. She said the predictive behaviors assessed through the new hiring system allowed the staff to recognize his management qualities and develop them accordingly. “He came to us with a goal in mind, and we hire with a goal in mind,” Morris said.
When asked about the quality of candidates Talent Science suggests for interviews, Morris said, “I thought they’ve all been a great match.”
Morris and Roberts agree that customer feedback has also improved, with higher scores seen after Talent Science’s implementation. “The scores have definitely changed for the better,” Roberts said.
Talent Science has also proved worthy from a return-on-investment perspective — Perkins declined to provide exact figures other than saying there have been double-digit increases in sales per hour and reductions in turnover. Numbers of associates in stores has remained the same, but associates are staying with the company longer, Perkins said, which has saved Foot Locker money.
“The feedback has been so positive and favorable, we are now expanding this internationally,” Perkins said.
Moving forward, Foot Locker said it plans to include additional points on Talent Science’s scoring scale from a five-point scale to seven points, Trigo said. It also plans to add Infor’s new behavior — creativity — to its job profiles. As the company rolls out the system internationally, profiles are also likely to change because of cultural differences.
In any event, Roberts said the greatest benefit since using predictive analytics has been the overall quality of hires at the Whitehall, Pennsylvania, store compared with the time it takes to source and hire them. Foot Locker is spending less time on the minutiae of finding and hiring the right candidates, and more time selling sneakers.
“They definitely respond a lot better to the workload that we ask of them,” Roberts said.
