(MoneyWatch) Turnover is a huge problem in any company. So what if you could predict, with reasonable accuracy, which of your team members were most likely to quit?
A number of companies have tried. Eric Siegel, a former professor at Columbia University, and founder of Predictive Analytics World, writes in his new book Predictive Analytics, about HP's Flight Risk program. This program used data to predict which HP employees were most likely to leave. Siegel says (in an interview) that such programs are "becoming increasingly common, especially with large organizations." In a pilot group, Siegel writes, HP was able to reduce turnover from 20 percent to 15 percent when armed with this knowledge.
When I first heard about this idea, I thought it sounded a wee bit Big Brother-ish, or like something out of Minority Report. But then I realized two things.
First, looking at which employees might leave involves the exact same process companies use -- widely -- to figure out which customers are most likely to defect. Of course, a person's livelihood seems like a more weighty matter than whether you're going to switch to a competing pizza chain, but as long as the data is used responsibly, it doesn't mean someone will be pushed out before they can quit or be denied promotions. After all, most companies want to keep their employees. Knowing someone you value is likely to quit could result in you, as a manager, giving that person a more interesting assignment or a raise. That's not a bad outcome for that team member.
And second, managers already make predictions about which employees are likely to quit -- some of which may not be fair. Notes Siegel, "In general, the point of making data-driven decisions is to move away from the gut and more toward empirically validated decisions." A manager who learns an employee is pregnant with her second child might assume she won't come back from maternity leave since that happened with someone else three years ago. But data could remind that manager that any person's decision to stay or leave is based on multiple factors that humans just can't weigh well. Data might tell that manager that an employee who goes out on leave for a bit is actually less of a turnover risk than someone who was just given a lot more responsibility -- but no raise. "The core science is about how to integrate multiple factors," says Siegel.
How do you decide which of your employees is most likely to quit?