Google's Project Oxygen: do managers matter?

A few years into Google's life, Larry Page and Sergey Brin wondered whether the company needed managers at all. In 2002 they tried a completely flat organisation and removed the engineering managers, to break down barriers to fast idea development and keep the collegial feel of graduate school. The experiment lasted a few months. They gave up when too many people came straight to Page with expense reports, conflicts with colleagues and other everyday problems.

The doubt stayed in the company's engineering culture, and in early 2009 Google's people analytics team, in the people function run by Laszlo Bock, set out to settle it with a project called Oxygen. Because Google's engineers would accept only strong evidence, the team framed the test the hard way: it tried to prove that managers do not matter. Its statisticians gathered more than 10,000 observations about managers, across more than a hundred variables, from performance reviews, feedback surveys and nominations for the company's best-manager award, which goes to about twenty managers a year. The attempt failed. Teams with higher-rated managers were happier and less likely to leave. The team then interviewed managers from the top and bottom of the ratings, without telling the interviewers which was which, to find what the best ones did differently.

Eight behaviours came out of it: being a good coach; giving the team room and not micromanaging; taking an interest in each person's success and wellbeing; being productive and focused on results; communicating and listening; helping with career development; having a clear vision for the team; and, last, having the technical skills to advise the team. Google turned the list into a twice-yearly anonymous survey of each manager's team. Bock reports that between 2010 and 2012 the average manager's score rose from 83 to 88 per cent favourable, and the lowest-scoring quarter rose from 70 to 77. In 2018 Google added two behaviours, collaborating across the company and being a strong decision maker.

The person who is best at the work is often the one who gets promoted to manage it, and Google's list shows why that so often goes wrong: technical skill came eighth. The behaviours at the top, coaching, giving people room and caring how they do, are ones you can see before a promotion, in who colleagues already go to for help.

Google's own evidence has limits. The ratings came from the managers' own teams, the study was never published, and teams with good managers might have been happier for other reasons. The stronger evidence comes from elsewhere. At one technology services company, where every worker's output was recorded, researchers ranked each boss by how much more the same workers produced under them than under other bosses. Moving a team from one of the weakest bosses on that ranking, the bottom tenth, to one of the strongest, the top tenth, raised the team's output by more than hiring a tenth person onto a team of nine would. At a large high-tech firm, managers whose teams rated their people skills highly lost fewer people, an effect that held up across several research designs, though it did not consistently improve other outcomes. Managers matter, mainly through who stays and how much a team gets done, and the evidence for that is stronger than the Google story alone.

Source: Google's founders and Project Oxygen: David Garvin, How Google Sold Its Engineers on Management, Harvard Business Review, December 2013. The survey figures: Laszlo Bock, Work Rules!, 2015. The output study: Lazear, Shaw and Stanton, The Value of Bosses, Journal of Labor Economics, 2015. The attrition study: Hoffman and Tadelis, People Management Skills, Employee Attrition, and Manager Rewards, Journal of Political Economy, 2021. The feedback study: Smither and colleagues, Personnel Psychology, 1995.

Work Rules!

Laszlo Bock, 2015

Google's head of people on what the company learned by measuring how it hires and manages, Project Oxygen included. Most of the methods, such as structured interviews and anonymous upward surveys, work at a company of any size.