The article states that in true Google fashion, they are tacking this problem like any other. They are using math, in particular a very complex algorithm that will predict which employees are at risk for leaving.
I find this fascinating as a person that loves HR Metrics and as a person that loves predictive HR Metrics.
Google is not explaining the details but say, "the inputs include information from surveys and peer reviews, and the algorithm already has identified employees who felt underused, a key complaint among those who contemplate leaving."
Because many of their top executives have recently left, Google felt the need to understand why people were leaving and who is vulnerable, so they could intervene prior to departure.
One of the comments on the Wall Street Journal online community from Gregory Lynn states that, "Google needs to develop a human touch versus always looking to technology." Great point, Gregory!
While the algorithm may give you who is vulnerable, you have to actually do something about the issues that caused disengagement with the company. (human touch)
So while I love the idea of using predictive metrics in HR, you have to balance that with the fact that you are measuring people and they are complicated. And sometimes, you need a little insight with your data. I think when you marry good data with company insight you get some great intelligence to use for better decisions.