EN
Back to the archive

The archive · Work & Ways of Doing · Operational decision · 2008–2018

Project Oxygen turns 'great manager' into a data list — 8 behaviors, then 10, Google-wide

When no one could say what made a manager great, Google mined its own data, found eight behaviors, trained on them — and re-ran the experiment as it grew.

Google

The ideaMine employee feedback to find what top managers do, make that list the training program, and re-test it as Google grows — management by data, not opinion.substantial

What it had to solve

Google grew so fast that people were promoted into management for technical skill rather than leadership, and nobody had a shared definition of what a great manager actually does. In 2008, People Operations launched Project Oxygen to replace intuition with evidence.

How it works

By 2008 Google was a company where the route into management often ran through technical skill, and no one had a shared answer to a basic question: what does a great manager actually do? People Operations answered with Project Oxygen — a research effort to determine, from Google's own data, what makes a manager great.

The team studied employee feedback and manager surveys, and found eight behaviors common among the highest-performing managers: being a good coach, empowering rather than micromanaging, showing interest in employees' success and well-being, staying productive and results-oriented, communicating well, supporting career development, holding a clear vision, and having key technical skills. Google publicized the list, built it into manager development programs and trained managers on it.

The results were concrete: Google reported improvements in management and in team outcomes like turnover, satisfaction and performance. In 2018 the company repeated the exercise — employees wanted stronger cross-team collaboration and decision-making — and updated the list: behaviors 3 and 6 were reworded, and two new behaviors, 'collaborates across Google' and 'is a strong decision maker,' were added. The ten-behavior list proved even more predictive of turnover, satisfaction and performance than the original eight.

What made the idea travel was publication. Google shared the full method on re:Work, its public people-ops platform, so any company could borrow the approach — turning management from folklore into an instrument you can measure, train and re-run.

Why it lands

  • Data beats intuition: the behaviors came from Google's own employee feedback, not from what executives assumed managers should do.
  • It made management trainable: the list became surveys, feedback tools and development programs, not a poster on the wall.
  • It kept iterating: the 2018 refresh showed the model had to grow with the company, adding collaboration and decision-making.
  • It measured outcomes: improvements showed up in turnover, satisfaction and performance, not just in manager ratings.
  • Open publication multiplied the impact: re:Work let outside companies adopt the same evidence-based method.

What it did

Google says publicizing and training the eight behaviors improved management and team outcomes — turnover, satisfaction and performance — over time, and that the 2018 ten-behavior list was even more predictive of those outcomes. The findings were published openly on re:Work so other companies could copy the method.

Write-upGoogle's Project Oxygen post in full

What you can take

If you can't describe what good looks like, measure it: turn a vague quality into a testable list, train against it, and re-run the experiment whenever the organization changes.

Since then

Project Oxygen's behaviors remain the backbone of Google's manager development, and Google re:Work became a free public library of people-ops research — Project Oxygen, Project Aristotle and hiring guides — that HR teams around the world consulted. The broader idea, that management quality can be measured and deliberately trained rather than left to personality, entered the mainstream conversation about people analytics.

Sources

spotted an error? The archive wants to know.

Your turn

You just read one. Describe the brief you are staring at, and see who has been given the same problem.

Free account · 3 free questions · no card

Related cases