|
|||
|
|
|
||||||
|
Predicting Retention
Google has developed an algorithm designed to identify which of its 20,000 employees are most likely to quit, according to a recent column in the Wall Street Journal. The WSJ notes that Google has suffered an exodus of top talent who have left the firm in recent weeks to join fast-growing companies such as Facebook, including display advertising chief David Rosenblatt, engineering director Steve Horowitz and search-quality chief Santosh Jayaram. The algorithm, which uses data from employee pay and promotion histories as well as performance reviews, helps the company "get inside people's heads even before they know they might leave," Google's Vice President of HR, Laszlo Bock, told the newspaper. Although using a formula for such purposes may sound revolutionary, organizations such as the U.S. Army have been employing statistical methodologies to determine the type of personalities that would fare best in a given occupation as far back as World War I, says Douglas Klein, president of Sirota Survey Intelligence in Purchase, N.Y. HR departments at large firms have long relied on statistical models that incorporate factors such as demographics and employee-attitudinal surveys to determine which employees are most likely to leave their organizations, he says. What's different about Google's algorithm, says Klein, is that it incorporates a wider range of variables than those that have traditionally been used. "Google is casting a much wider net, looking at things like promotion and pay histories, performance appraisals and other kinds of HR and non-HR elements that haven't traditionally been used in these sorts of things -- it's pretty unique," says Klein. "Will it be more predictive than other methods?" he asked. "I don't know. Sometimes more data is better, sometimes it's not." Google declined to comment for this story. Dr. Jac Fitz-enz, CEO of Human Capital Source, a consulting firm in San Jose, Calif., and founder of the Saratoga Institute, an organization that pioneered the use of HR metrics, says Google's algorithm is a sign of things to come. "If this algorithm can really do what Google says it can, then this is very exciting," he says. "This is what we've been talking about in terms of predictive management. If you want to find out the best way of finding and bringing into your organization the very best people for certain positions, an algorithm can show you the best way." Opinions were generally positive on the idea, but offered some cautions. Paul Sparta, CEO of Plateau Systems, a talent-management systems vendor based in Arlington, Va., says that, while he's impressed by Google's algorithm, it's crucial that companies planning to do something similar ensure the data they rely on is clean and accurate. "You've got to have very consistent and normalized data that's relevant to whatever your goal is, otherwise you'll end up with garbage if the data you're using is polluted or anecdotal," he says. And David Karel, who says the algorithm is a "very interesting idea," says the existence of such a formula may also signal a failure on Google's part to keep its employees engaged. "Once you're trying to figure out who's going to quit, you might have already lost the game," says Karel, vice president of product marketing at SuccessFactors, a talent-management systems vendor in San Mateo, Calif. It would be better for HR, he says, to focus on preventive medicine: Does the company have processes in place for measuring performance fairly and consistently? Is a pay-for-performance culture in place? Are employees being developed in a focused way based on performance gaps and strategic needs? Although Google hasn't indicated any plans to market its algorithm to outside firms, Sparta says, companies that wish to do something similar would probably be better off developing their own algorithms instead of using one developed at another company. "The algorithm that Google uses may not work for another company," he says. "The reason a company might need its own algorithm is that you might ask different questions on your performance reviews than Company B, for example. You can't just use a generic algorithm and expect it to work." Fitz-enz says his organization is in the process of developing an algorithm for the U.S. National Reconnaissance Office, the agency in charge of the government's spy satellites, that's designed to help the agency find better ways of preparing its program managers for future challenges. It's also partnering with KnowledgeAdvisors Inc., a Chicago-based software firm, to develop an algorithm-based product designed to help firms optimize their practices in areas such as learning, performance management and compensation planning, he says. However, HR leaders don't need to purchase algorithms from vendors or pay for them to be developed by expensive consulting firms, says Fitz-enz. "If you hire a couple of statisticians, they can help you develop an algorithm for your own organization." Even so, too many HR practitioners shy away from algorithms, he says. "Most HR departments are not well-versed in analytics, so they don't appreciate what they can do. Analytics has not been a strong part of HR in most companies and HR people, in general, are not analytically inclined."
June 17, 2009 Copyright 2009© LRP Publications |
||||||
Print
Email
Write to the Editor
Reprints
|
||||
|
||||
|
||||
|
© Copyright 2010 LRP Publications. All rights reserved.
|
||||