Google has come up with an algorithm to work out which of its employees feel under-used and are most likely to quit.

The American internet search giant has begun crunching data from employee appraisals and promotion and pay histories in a mathematical formula which can identify those of its 20,000 employees who feel most unhappy at work and are likely to leave, according to the Wall Street Journal.

The move is one of a series Google has made to prevent its most promising engineers, designers and sales executives from leaving now that the firm's start-up atmosphere has faded because of the size of the company, the newspaper claimed.

Google's HR chief in the US, Laszlo Bock, said the formula could supplement traditional engagement techniques such as training.

The algorithm would "get inside people's heads even before they know they might leave," said Bock.

Advertising sales boss Tim Armstrong and display advertising chief David Rosenblatt have both quit the company in recent weeks, while senior managers including desginers and engineering directors have flocked to start-ups such as Facebook and Twitter.

Bosses at Google are worried the organisation is losing talent because employees feel they cannot make the same impact they once did.

However, Bock stressed not all the top talent had left the company. "We haven't seen the most critical people leave," he said.