Google is developing a staff-churn search algorithm in a bid to bolster talent retention among want-away personnel who are presumably no longer sold on the benefits of stock options, free gourmet meals, subsidised scooters and regular "brain-expansion" Tech-Talk lectures.

The typically quirky and decidedly different angle on HR automation uses Google's fabled data-collection and number-crunching processes to figure out which of its employees are unhappy and most likely to quit.

Laszlo Bock, Google's HR chief in the US, believes this new, more quantitative approach to personnel decisions should help the company perfect its talent retention policies and "get inside people's heads... even before they know they might leave".

It has already helped identify which of Google's 20,000 employees feel underused - a key complaint among those who contemplate leaving.

The algorithm works by applying behavioural analysis to personnel data mined from employee records, promotion and pay histories, surveys and peer reviews - but Google has so far been reluctant to share any further details. It will be used to supplement more traditional talent measures derived during appraisals, training and leadership meetings.

Researchers are working on other ways of triangulating data to validate trends and reduce errors in a wide range of business forecasts (see Data Feed, below).

The reason efforts have been focused inwards, on HR, is that despite receiving more than 700,000 job applications a year, the company is concerned about a potential brain drain to the likes of Facebook and Twitter.

Data Feed

Combining data from Google Trends with economic models cuts error rates in forecasts of car and vehicle parts sales by 15%.

About 90 million Americans a year search online for information about specific diseases, enabling Google researchers to predict flu epidemics in the US one or two weeks ahead of the official Center of Disease Control and Prevention.