The Power (And Peril) of Predictive Hiring Analytics
According to a new study, a growing number of new hires — as well as top-performing employees — are choosing not to stick around for long.
The 2014 PwC Saratoga U.S. Human Capital Effectiveness Report, which is based on a survey of 375 employers, finds the percentage of employee headcount for new external hires increased by nearly 40 percent from 2010 to 2013. Unfortunately, more of those new hires are leaving before their one-year anniversary: The report finds that first-year-of-service turnover increased for the second consecutive year, from 22.6 percent in 2012 to 34.1 percent in 2013. It also finds the separation rate among high-performing employees has risen to its highest level in 10 years, from 5 percent in 2012 to 6 percent last year.
The report finds that some organizations are turning to predictive analytics to boost their “quality of hire,” and thus reduce the likelihood of first-year turnover by ensuring a better fit between job candidates and open positions, says Ranjan Dutta, a director of PwC Saratoga and an author of the report.
“If employees join an organization and leave within the first year, whether it’s involuntary or not, it’s indicative of poor selection and poor onboarding,” he says.
Smart companies are turning to predictive analytics to do a better job of selecting candidates, says Dutta. Predictive analytics may help HR identify candidates who are most likely to stay (and thrive) with the company by identifying the characteristics of top performers who’ve held pivotal roles and then looking for those characteristics in candidates for those positions, according to the report. These tools can examine data from inside and outside the company, as well as input by the candidates via pre-hire assessments.
The use of predictive analytics for hiring is a natural occurrence, given the sheer amounts of data that technology has allowed companies to amass, says Jason Corsello, vice president of corporate development and strategy at Santa Monica, Calif.-based talent-management vendor Cornerstone OnDemand.
“Companies are sitting on more data than ever: They know who their people are, what skills they have, their educational backgrounds, their job performance,” says Corsello, who says Cornerstone is developing new applications that make use of predictive analytics. “The other part of the equation is that advances in technology mean that the cost of gathering all this data has gone way down.”
Companies’ use of predictive analytics in areas besides hiring has generated some controversy, however. In his 2012 bestselling book, The Power of Habit: Why We Do What We Do in Life and Business, New York Times reporter Charles Duhigg drew attention to the use of PA by major retailers such as Target to predict what their customers will buy. In some cases, Duhigg wrote, retail chains can tell when a customer is pregnant even before she does by analyzing her buying patterns.
“The use of big data such as predictive analytics can be very powerful in terms of predicting what someone’s going to do,” says David Walton, a shareholder at Cozen O’Connor in Philadelphia and vice chair of the law firm’s labor and employment group.
It can also be used for questionable purposes, he adds, citing a Belgian firm that uses big data to create credit profiles of potential bank customers. The firm uses correlations which it says have revealed that people whose Facebook friends tend to write posts in all caps are potential credit risks, for example, says Walton.
“If they can do that with credit monitoring, then what can they do with employment practices, and what are the potential implications? It’s scary and amazing at the same time.”
The use of PA to make hiring decisions could lead to discrimination if not used properly, says Dutta.
“Organizations need to be very careful in how they’re using this software,” he says. “You don’t want to end up with a model that tells you not to hire people of certain ethnicities, for example.”
None of the three employment-law attorneys interviewed for this story knew of any current or recent employment-discrimination litigation concerning the use of predictive analytics for hiring. However, all agreed that the potential for adverse impact exists.
“Yes, I absolutely think it’s a legitimate concern,” says Erin Schilling, a shareholder at the Polsinelli law firm in Kansas City. “[The use of predictive analytics] could have a disparate impact on minorities, women or any different kind of class of worker.”
To stay out of trouble, she says, companies should first perform an “adverse impact analysis,” or a statistical test, on the predictive analytics they use to determine whether it screens out particular groups.
“In theory, predictive analytics should — if applied properly — broaden the possibilities for applicants to get hired and, theoretically, should result in possibly a more diverse workforce than what might have otherwise come through the door based on standard hiring processes,” says Peter Gillespie, an employment attorney with Fisher & Phillips in Chicago. “But, there’s always the risk of testing bias, and we’ve certainly seen the EEOC pursuing concerns about potential bias in the hiring process.”
HR leaders should ensure they understand how the factors that are being considered were selected and how the process was developed, he says. Should they use an outside vendor, they should check with other clients to see whether they experienced any negative outcomes with respect to adverse impact, says Gillespie.
Proponents of predictive analytics say the software can actually dispel myths that often stand in the way of people with certain backgrounds getting hired.
Michael Housman, Evolv’s chief analytics officer, says the predictive analytics his firm provides to its clients — a number of them large employers with many hourly workers — allow them to “hire and promote in a data-driven way.”
The company takes the information applicants provide via online pre-hire assessments and uses it to rate them as green, yellow and red. “Green” applicants are those who are most likely to stay with the company longer, while those rated yellow or red are more likely to leave, he says. These are important factors to consider for companies concerned about high turnover rates, he adds.
“We’ve been able to dispel a lot of myths around what predicts success in these roles,” says Housman.
For example, Evolv has discovered that applicants with a history of “job hopping” in their previous experience ended up having a job tenure just as long as employees without such a history. It also found that people who had previously been unemployed for long periods performed just as well in their jobs as those with more traditional work histories, he says.
Employees with criminal convictions on their records, says Housman, also ended up performing on the job just as well as colleagues with clean records.
“It’s not a matter of where you’ve been, but whether you’re a good fit for the job you’re applying for now, and the best way to assess that is to determine who you are as a person,” he says.
With respect to potential adverse impact, Housman points to Evolv’s recently issued workforce code of ethics.
“Any time we deploy an assessment, we conduct an adverse-impact study to ensure that it won’t screen out members of a protected class — if it does, then we don’t deploy that assessment,” he says.
“All the data we use is legally submitted — we’re not going around scraping people’s Twitter feeds without their permission,” says Housman.
Shaker Consulting Group, a Cleveland-based vendor, also creates predictive-analytic tools designed to help its clients make better hiring decisions. The firm performs a validation analysis with each client to ensure there are no disparate impacts, says Brian Stern, president of Shaker Consulting Group.
“You should be using big data to help you ask smarter questions to help inform your decision-making, as opposed to letting this data make decisions for you,” he says.
Similar to Evolv, Stern says predictive analytics has helped dispel myths that had previously been treated as rules of thumb by many recruiters.
“We found that financial institutions were weighting candidates that had a lot of cash-handling experience more heavily in the process, but in fact, once you have a little bit of cash-handling experience, having more of it doesn’t make a difference,” he says.
Dutta suggests that companies interested in using PA in the hiring process begin with a pilot project first and evaluate the results carefully. HR should also keep in mind that the software should never play a dominant role, he says.
“Use the software as a tool, not as a decision-maker,” says Dutta.