article » People Analytics: ‘Moneyball’ for Human Resources

People Analytics: ‘Moneyball’ for Human Resources

August 1, 2014
6 min read

Lord knows it’s expensive and time consuming, and even when they get it, it’s not clear how much students actually learn. But the one thing you could always say about a college degree was that it led to better jobs because it served as a reliable and invaluable signal to employers about a job applicant’s intelligence and persistence.

Or maybe not. Michael Rosenbaum knows from fancy degrees — he’s got a BA and law degree from Harvard, along with a master’s from the London School of Economics. But at the two software companies he has founded in Baltimore, 40 percent of the programmers have no college degrees, and half of the others got theirs from community colleges. The reason is simple: Statistically speaking, a degree from a fancy college has zero correlation with success in writing software.

Such are the insights from “people analytics,” a hot new area in human resource management that aims to bring “big data” to the task of corporate hiring and promotion. What “moneyball” did for baseball, people analytics promises to do for human resources, replacing intuition, old-boy networks and outmoded rules of thumb with computerized tests, database searches and quantifiable performance metrics.

Given the time and money companies sink into hiring and recruitment, the results are decidedly mediocre. According to a survey conducted by Arlington-based Corporate Executive Board, nearly a quarter of all new hires leave within a year, while Gallup reports that half of those who do stay reported being “not engaged. The resulting drag on profits and productivity represent a multibillion dollar opportunity for firms such as Rosenbaum’s Pegged Software, which helps hospitals and nursing homes reduce turnover of entry-level workers by putting the right people in the right jobs.

This is actually an old idea, going back to the 1950s when large corporations first made widespread use of psychological and intelligence tests to screen applicants for jobs. But over the years, the process began to look more expensive and less effective. By the 1990s, companies had reverted to scanning résumés for relevant degrees and years of experience, followed by subjective impressions gleaned from unstructured interviews. The result, according to one study, is that more often than not, managers hire people just like themselves.

What’s surprising isn’t that companies are beginning to use more objective techniques to make personnel decisions, says Peter Cappelli, a professor at the University of Pennsylvania’s Wharton School of Management. “What’s surprising is that, for the last 20 years, they’ve been going with their gut.”

In the early versions of “people analytics,” companies asked managers to identify their best-performing employees and then tried to find patterns in each group based on biographical information, work history and their answers to computerized personality and intelligence tests. Job applicants were then “scored” based on how closely their background and test answers matched those high-performing employees.

Rosenbaum says the problem with the early efforts is that the results tended to correlate less with the actual strengths of the job candidates than with their cleverness in taking the tests. And by relying on the managers’ rating of current employees, the process reintroduced a high degree of bias and subjectivity into the scoring system.

The more recent and effective iterations of these systems start by collecting whatever data they can about the performance characteristics of current employees — how long they have been on the job, how many calls they process or sales they close. In the computerized tests, the purpose of many questions isn’t to see whether someone gives the “right” answer, but to see how long it takes them to answer it, or how consistent it is with other answers they have given, or what words they use to answer more open-ended questions.

Pegged Software bases its scoring of job applicants on several thousand data points. It has found that having more years of experience in a job doesn’t necessarily correlate with a greater probability of success. Moreover, factors that do correlate can vary markedly from one job category to the next — or even in the same job category at different companies. One example: Leadership in community organizations turns out to be a positive marker for acute-care nurses but not for those working in nursing homes.

Out in San Francisco, a company called Evolv has helped Xerox, AT&T and other big companies screen applicants for front-line sales and customer-service jobs. Evolv has found that, contrary to HR folk wisdom, there is no statistical reason not to hire a convicted felon, or a job hopper, or someone who doesn’t score well on general intelligence. Nor does it matter whether they are unemployed or how long they have been out of work. None of it correlates with turnover and job performance.

On the other hand, what may be useful predictors are how far somebody has to commute, whether they use a factory-installed browser or download a better one like Firefox or Chrome, or how many social-networking sites they use. In sales, Evolv has found that creativity rather than persuasiveness is a better predictor of success, while at call centers, rapport with customers is more important than either.

Different industries require different approaches to people analytics. A company called Knack in Silicon Valley uses video games such as Dungeon Scrawl and Wasabi Waiter to help Shell assess employee traits such as persistence and creativity, and the ability to learn quickly from mistakes.

And San Francisco-based Gild helps software companies recruit programmers through an algorithm that analyzes their language on professional-networking sites. Another evaluates the open software a coder has written for its simplicity, elegance and the number of times it has been adopted by other programmers.

In coming up with these systems, companies employ neuroscientists, psychologists, statisticians, econometricians, anthropologists and engineers. Yet even with all that expertise, says Michael Housman, the 32-year-old chief analytics officer at Evolv, they can’t always explain why some factors, or combination of factors, turn out to be more predictive of success than others. The focus is on correlation, not causation.

Another feature of People Analytics 2.0 is that it is dynamic. The artificial intelligence built into these software programs means that it is continuously learning from experience and refining its algorithms as more performance data is collected about employees who leave or stay or are promoted. Over time, some factors become more important, others less.

The results can be impressive. According to Rosenbaum, Pegged Software’s clients typically cut their first-year employee turnover by 50 percent or more — no small matter at hospitals where first-year turnover is 35 percent and the cost of replacing each employee can run $15,000 to $50,000.

At Evolv, chief executive Max Simkoff says his clients generally reduce turnover by 15 to 20 percent while improving productivity metrics — calls handled, sales made — by 3 to 5 percent.

Unlike some critics, I don’t get particularly worked up that computers have replaced humans in making personnel decisions. Most companies who use the software still do face-to-face interviews, albeit fewer interviews with fewer finalists. At the same time, the computer-driven process clearly reduces the irrational biases and prejudices that prevent lots of applicants from even being considered.

For me, a bigger concern is that these systems are likely to produce companies with rather homogeneous workforces. That may be okay at call centers, where people work independently of one another and there are clear, simple metrics of success. But in work settings requiring teamwork and collaboration on tasks involving creativity and judgment, I suspect there will be a danger in winding up with employees who all have the same personality traits or the same strengths and weaknesses.

In those settings, success often comes from having a kind of unplanned diversity in talent, outlook, experience and personality that fosters creativity and militates against the groupthink that has doomed even the mightiest of companies.

For access to the full-length article, click here.