In 2002 Larry Page and Sergey Brin tried something many engineers quietly wish for: they removed the entire engineering management layer and went completely flat. The aim was to clear the path between an idea and a product, and to recreate the collegial feel of graduate school.
The experiment lasted a few months. It ended because too many people went directly to Page with questions about expense reports, interpersonal conflicts and other everyday matters1.
That is the best detail in the whole story, and the one usually cut when it gets retold. Management was not restored because somebody wrote a persuasive essay about leadership. It was restored because without it the work jammed in the most ordinary places.
The problem: Google's engineers did not believe it
Even after that experiment, most engineers still regarded management as more destructive than useful, a distraction from real work.
And Google is a place where nobody accepts a top-down directive without question. To change anything there you need tight logic and plenty of data.
So people operations did the thing engineers respect: a study. They called it Project Oxygen.
Step one: prove that managers make a difference
The analytics team started with data it already had — Googlegeist scores, semiannual reviews — and compared managers on both satisfaction and performance.
The 2008 finding was the hinge of the whole project. High-scoring managers saw less turnover on their teams, and retention was tied more strongly to manager quality than to seniority, performance, tenure or promotions. Employees with high-scoring bosses also reported greater satisfaction across innovation, work-life balance and career development1.
Step two: find out what the good ones actually do
Knowing managers matter teaches nobody anything. So the team ran double-blind qualitative interviews with both high- and low-scoring managers, asking concrete questions such as how often they hold career development discussions with their reports.
Participants came from Google's three major functions, from all levels and geographies. The team also read thousands of qualitative comments from Googlegeist surveys, performance reviews and submissions for the company's Great Manager Award, for which Google picks about 20 managers a year on employee nominations. Coding it all took several months.
The output was eight behaviours shared by high-scoring managers. In the order printed in David Garvin's Harvard Business Review article1:
- Is a good coach.
- Empowers the team and does not micromanage.
- Expresses interest in and concern for team members' success and personal well-being.
- Is productive and results-oriented.
- Is a good communicator — listens and shares information.
- Helps with career development.
- Has a clear vision and strategy for the team.
- Has key technical skills that help advise the team.
Note where technical skill sits. At a company built on technical credibility, which had once concluded it needed no managers at all, technical skill is item eight.
Why the engineers listened
Michelle Donovan, one of the project's two leads, said the team hoped the list would resonate because it came from Google's own data. The attributes, in her words, were about us, by us, and for us.
The rollout was designed with the same care. Google built two upward feedback surveys, one for administrative and business functions and one for engineers. Employees rated managers on a five-point scale against specific activities, such as giving actionable feedback regularly and communicating team goals clearly.
The first surveys went out in June 2010, deliberately out of sync with performance reviews in April and September. Google had considered tying the scores to performance reviews and decided against it, because employees would read it as standards imposed from above and push back1.
That was the most important design decision in the programme, and it is the one companies most often get backwards.
The results, and their limits
Between 2010 and 2012 median favourability on the two surveys rose from 83% to 88%. The lowest-scoring managers improved the most, particularly on coaching and career development, and the gains were consistent across functions, levels and regions.
The article gives one individual example. Sebastien Marotte, a vice president who joined from Oracle in 2011 and ran a global team of 150, hit his sales numbers and then received shockingly low survey scores. After working through every comment with a people ops colleague and building a plan, he took his favourability from 46% to 86% within two survey cycles1.
Keep a cool head reading those numbers. This is Google's own data about Google's own programme, with no independent control group. Most of the findings are correlational rather than demonstrated cause. And the setting is unusual: at the time Google had more than 37,000 employees but only 5,000 managers, 1,000 directors and 100 vice presidents. Engineering managers with 30 direct reports were common, by design, so that there was only so much meddling anyone could do.
The list has changed, and how Google runs it now
Project Oxygen did not stop with the 2013 article. On re:Work, Google's own publication, the team says it analysed over 10,000 data points about its managers, drawn from performance reviews and feedback surveys, plus hundreds of pages of interview notes2. Its stated conclusion: teams with highly effective managers achieved better results, were happier, and had lower turnover2.
The eight behaviours later grew to ten and, together with the Project Aristotle research, became a company-wide framework called Google Manager Responsibilities2.
The operating cycle around that framework is the part worth copying. At the start of the year, managers set their own expectations against the framework. Each quarter they get feedback on their management from their own manager. Annually, their team rates how well they meet the responsibilities, and the manager receives a report with both qualitative and quantitative feedback2.
One small point in the re:Work guides matters for mid-sized companies: they advise managers to work on one or two behaviours at a time, and to ask their own team which ones would matter most3. Handing everyone a list of ten and asking them to improve on all of it is the surest way to change nothing.
What a Vietnamese company can take from it
The value of Project Oxygen is not the eight behaviours. The list is fairly obvious, and Google says as much.
The value is in three design choices.
First, they proved the problem before selling the solution. Companies here usually run that backwards: buy the management training programme, then work out how to justify it.
Second, they separated the manager survey from the performance cycle. Once survey scores feed pay decisions, they stop being a development tool and become politics.
Third, they described behaviour rather than qualities. Nobody can improve on "has leadership presence". Anyone can improve on "holds a career conversation with each team member at least twice a year" — it can be measured, taught and started next week.
