What Google learned by firing all its managers
Google fired every manager in 2002 and regretted it within months. Then it spent 10,000 data points learning what a manager actually does — and the answer fits in two lines.
60% of managers fail in their first 24 months.[1]
82% of them never get training.[2]
There is a correlation here.
Where 600 new managers end up — figures modeled
Trained108 · 18%
Untrained492 · 82%
The 82 / 18 split and the failure baseline are published figures (CMI 2023; Arruda, Forbes 2023). Effectiveness rates are modeled from the measured effect of manager training (Lacerenza et al. 2017). The animated version of this figure requires JavaScript.
A cohort of 600
Everyone starts by being good at the work.
Each dot is one person — an engineer, a nurse, an analyst. None of them manage anyone yet.
The promotion
Then one Monday, they're in charge.
Most were promoted for being the best at the old job. Leading the people who now do that job is a different job.
The split
82% never get training.
Of our 600, that's 492 handed a team with no preparation at all — the research calls them accidental managers. If that's you, you're not the exception. You're the rule.
82% + "accidental managers": Chartered Management Institute, 2023.
The odds
60% of managers fail in their first 24 months.
Applied to the untrained majority, that leaves about 40% effective. Across 335 studies, management training shifts the odds enough to raise it to roughly 67%.
At today's mix, about 270 of 600 come out effective.
60%: Arruda, Forbes, 2023. Training effect: Lacerenza et al., 2017, J. Applied Psychology (335 studies). Effectiveness rates modeled — see the note beneath this figure.
If everyone were trained
About 130 more cross the line.
Same 600 people, same bar — the only change is training. Roughly 400 of 600 effective instead of 270, half again as many. And a manager is an amplifying role: each one who rises lifts a whole team with them.
That's the line Google decided to move.
Modeled at the trained rate (~67%) — see the note beneath this figure.
Google · 2002
Google bet that managers didn't matter.
So it ran the experiment: Google eliminated its engineering managers entirely — a completely flat organization, roughly 130 engineers reporting to a single executive. If managers were overhead, the work should have sped up. Spoiler alert — it didn't.
Garvin, HBR, 2013; Levy, In the Plex.
The flat org
The experiment lasted a few months.
It went poorly. The flat organization created more problems than it removed — and Google not only put the managers back, it studied what separates the good managers from the bad.
Project Oxygen · 2009
A great manager does two things.
Google went looking for what actually made its best managers good — 10,000 pieces of data: performance reviews, 360s, award nominations. The conclusion: get results today, and develop the team for results tomorrow.
Project Oxygen, 2009.
Google manager training
Google made training mandatory.
Built from what Project Oxygen found, for every manager. They made the investment because the data showed this job is learnable.
The numbers behind the dots
| Group | Managers | Effective (modeled) | Struggling (modeled) |
|---|---|---|---|
| Trained · 18% · ~67% | 108 | 73 | 35 |
| Untrained · 82% · ~40% | 492 | 197 | 295 |
| All managers | 600 | 270 | 330 |
How the rates are built: 60% of managers fail in their first 24 months (Arruda, Forbes, 2023) → the untrained baseline is taken as ~40% effective. Training lifts on-the-job leadership behavior by ~0.7 SD (Lacerenza et al., 2017; 335 studies). The trained rate is derived with a normal-threshold model, p = Φ(d − Φ⁻¹(1 − p₀)) ≈ 67%. The effect depends on training quality — lecture-only runs ~0.3 SD; practice + feedback reaches ~0.96 SD. Train-everyone counterfactual: ~403 of 600 effective at the same trained rate (modeled).
Managers are built, not born.
The most difficult transition I’ve had was going from running a team of 20 to a team of 160. I requested training on a particular topic for 2 years. Finally, a 1 hour class was created. And in that 1 hour, I learned everything I needed to know. I never struggled in that topic again.
But that training has not been offered again in the 10 years since.
So few companies train their managers. I’ll never understand that.
The training is expensive to develop, creating a barrier for small companies. I can understand that barrier, but I can’t understand why so many large companies would fail to train their managers.
I can only guess they point to the 70-20-10 rule — the Center for Creative Leadership finding that only 10% of learning comes from formal training.[6] This may be true, but the CCL never meant to say that training is dispensable. It was meant to define a system of training, guidance, and OJT.
If you are a manager, you deserve training.
I’ll tell a counter example in this essay.
Google has 30,000 managers today.
They all get mandatory training.
It wasn’t always like that.
Google fired all their managers in 2002. They regretted that decision quickly and brought the function back. They realized the decision was based on observations of bad managers who added little value instead of the good managers who added tremendous value.
They set off to determine what makes a manager great.
And then they built a training program to get consistent results out of their managers.
This essay describes what they learned.
The story: Google’s experiment firing all managers
“A great manager will make a tough job challenging and rewarding. And they’ll make a great job fantastic and transformational.
But a bad manager will make a great job so-so and a tough job soul crushing.”
Google was founded in 1998.
In 2002, with only 1 humble pop culture reference to date, Larry Page and Sergey Brin decided to remove their engineering managers entirely — a completely flat organization, with roughly 130 engineers reporting to a single executive.
That experiment went poorly — priorities became unclear, projects stalled, expense reports piled up on a single desk. They realized managers make a difference.
And they learned not all managers are created equal.
They brought managers back into the business.
And they started asking what separates the good managers from the bad.
In 2006, with more than 10,000 employees, they hired Laszlo Bock to run Human Resources. Laszlo started running annual 360 degree assessments. In 2007, they hired Prasad Setty to stand up a “people analytics” group. And in 2009, that HR and People Analytics group stood up “Project Oxygen” with the goal of finding out what makes a great manager. They looked at 10,000 pieces of data from things like performance reviews, 360 reviews, nominations for top-manager awards.
And they concluded a great manager only needs to do two things:
- Get results today
- Develop the team for results tomorrow.
They built a training program to consistently build great managers.
Remember that one-hour class I waited two years for? That’s what it was. Someone finally taught the job instead of letting me guess at it — one hour of training closed a gap that two years of doing hadn’t. Google decided to do that on purpose, for every manager, at scale.
Now, when you become a new manager at Google, they let you struggle for 30-60 days to develop curiosity. And then they pick you up and bring you to mandatory training. They also provide each manager a coach to ensure their continued development.
How’s that for learning from their decisions?
The framework: what a manager does
Google concluded a manager has two main functions:
- Deliver results
- Develop your team
Google also defined 10 behaviors split into these two categories:
Getting results
- Have a clear vision and strategy for the team
- Be productive and results oriented
- Collaborate across the organization
- Be a strong decision maker. Know when to take a decision and when to escalate one.
- Have key technical skills to help advise the team
- Understand the team’s work enough that you can roll up your sleeves and dive in when needed
- Learn new skills to meet the business’s needs.
Developing your team
- Create an inclusive environment and show concern for the success and wellbeing for the people on the team
- Empower the team, don’t micromanage
- Support career development and discuss performance
- Be a good coach
- Be a good communicator
- Communicate about the team
- Communicate about the company
As we’ve discussed before, a Team is a bounded, stable group of people interdependent on achieving a common goal.
Therefore, the team leader’s role is to get better outcomes from the team than the team could have delivered on their own. That’s the value leadership brings: they build a team that works well together and creates the environment for success through coaching and removing barriers.
And as the leader supports the team’s achievement of the shared goal, they balance 1) near term results and 2) growth for the future.
Now I’ll break down the definition of a team to illustrate.
Stable
“Stable” means the group’s members are unchanging and “bounded” means they are known. Obviously people move in and out your business, and in and out of your team within your business — nothing is ever set in stone. But this happens relatively slowly and the team’s membership is always clearly defined.
And this means the team can practice working together — they have time to get to know each other, to learn each other’s strengths and weaknesses, to learn the norms of the team, and improve their collaboration.
This is valuable for success today and success tomorrow.
Interdependent
“Interdependent” means the team must work together to create something no single person could have created. Remember to build psychological safety to get them to work together effectively.
There are only two reasons why you need a team.
Either you need to add people to deliver at Scale, or you need to add expertise to deliver your Scope.
Scale
Say your business is making pizza. You can personally make 1 pizza, or maybe even 10 pizzas. But you can’t make 100 and you can’t do it 7 days a week. Here you’ll add a team to deliver your solution at Scale.
Scope
Imagine you want to make a special telescope to look at how the universe started (like the James Webb Space Telescope).
But to do that, you need more than the telescope itself. You also need a rocket to send the telescope up into space.
You need a radio system to send the telescope’s pictures and data back to Earth. You know how to make the telescope, but you don’t know how to make rockets or radios.
So, you get help from a team of experts who know how to make rockets and radios. Together, you all work to make the whole project happen.
You are adding a diverse team of specialists to deliver the full Scope of your solution.
Common Goal
And that team needs to be aligned on a “Common Goal”.
We build alignment on Finite Goals (Tasks and Projects) and Infinite Goals (Shared Purpose).
This means we need success today, and we need success tomorrow.
I really like this fable from Peter Drucker:[7]
Three stonecutters were asked what they were doing.
The first replied, “I am making a living.”
The second kept on hammering while he said, “I am doing the best job of stonecutting in the entire county.”
The third one looked up with a visionary gleam in his eyes and said, “I am building a cathedral.”
Three people doing the same job.
Only one can see his daily work in the larger context.
Now imagine if the whole team felt that context.
That is the job of the manager.
Communicate the near term task.
Communicate how it fits into the larger context.
Create a shared sense of purpose.
Get near term results and develop the team.
The implications on management
Before I walk through them, I’ll tell you why I think most management training doesn’t stick: it lacks a holistic view. We teach in vignettes — a feedback class here, a delegation module there — as if each tool stood alone.
But feedback is one tool in an interdependent toolbox. We are trying to achieve a Common Goal. We need each other to achieve it. Therefore we need feedback. That chain — goal, interdependence, then the tool — is what the vignettes skip, and it’s where this framework differs. Every task below hangs on it.
Delegation
A core task of management is delegation.
When we do so, we consider near term success and growth for the future when we ask:
What team skills should I be leveraging?
Know your team including strengths and weaknesses
What team skills should I be developing?
Understand who wants to learn what.
Understand what the future may look like, and develop relevant team skills.
Feedback
Another core task of management is feedback.
When we fix people’s substandard work (rewriting the slide deck at 11 p.m. instead of sending it back), we do so to get results today. But we sacrifice tomorrow’s results because we teach them to continue producing substandard work.
Silence guarantees nothing will change.
When we give them feedback, we give them awareness of the gap to the standard, and we help them reach it. This gets results today.
Coaching
When we coach people, we bring employee learning to the next level.
As Sir John Whitmore says:[8]
“Coaching unlocks people’s potential to maximize their own performance. It is helping them to learn rather than teaching them.”
Hiring
A core task of management is hiring.
Hiring for a short term lens may sacrifice long term success. Especially as the hiring process drags on and a tendency to settle sets in.
When you hire by committee, you spread the pressure to fill the role across more people helping maintain a short and long term focus.
And so on.
Every management task can be executed to get results today and develop the team for results tomorrow.
One book, one podcast, one video, one article
Here’s the best stuff I’ve found while researching this.
“People with a growth mindset knows that it takes time for potential to flower.”
The book. Reading Carol Dweck’s Mindset is a prerequisite to attending Google’s manager training. She helps us see how Michael Jordan’s “Growth Mindset” enabled him to take feedback that he wasn’t good enough at shooting, ball handling, and defense to inform a development agenda making him the best in the world at all three areas. Pardon the Amazon link, the publisher doesn’t have a page anymore. Please buy in your local bookstore. They won’t survive without you.
“Give managers the feedback they need to get better.”
The podcast. Sarah Calderon ran Google’s manager training program for years. She described Project Oxygen and their training program on the Google Partners podcast in 2017 — an episode that has since all but disappeared from the internet (Google deleted the original audio). A listing lingers on Castbox (33 minutes).
“The high performance but low trust person is toxic.”
The video. Simon Sinek describes a version of the 2×2 Psychological Safety matrix but as “Performance” and “Trust”. The Navy SEALs would rather have someone of Medium Performance and High Trust than someone of High Performance and Low Trust. I had posted one of my Psych Safety videos on Facebook and a friend sent this as a reply. It’s so good. Watch on YouTube here (3 minutes).
“Engineers hate being micromanaged on the technical side but love being closely managed on the career side.”
The article. This article describes Google’s manager training program, and includes several examples of the manager feedback that was used to create it. “How Google Sold Its Engineers on Management” by David Garvin, Harvard Business Review.
- [1]Arruda, William. “Why Most New Managers Fail And How To Prevent It.” Forbes, February 15, 2023. ↩
- [2]Chartered Management Institute. Taking Responsibility — Why UK PLC Needs Better Managers. Research Report. CMI, 2023. ↩
- [3]Lacerenza, Christina N., Denise L. Reyes, Shannon L. Marlow, Dana L. Joseph, and Eduardo Salas. “Leadership Training Design, Delivery, and Implementation: A Meta-Analysis.” Journal of Applied Psychology 102, no. 12 (2017): 1686–1718. Cited in Exhibit A (training effect, 335 studies).
- [4]Garvin, David A. “How Google Sold Its Engineers on Management.” Harvard Business Review, December 2013. Cited in Exhibit A (the 2002 flat-org experiment).
- [5]Levy, Steven. In the Plex: How Google Thinks, Works, and Shapes Our Lives. Simon & Schuster, 2011. Cited in Exhibit A (roughly 130 engineers reporting to a single executive).
- [6]McCall, Morgan W., Michael M. Lombardo, and Ann M. Morrison. The Lessons of Experience. Center for Creative Leadership, 1988. The 70-20-10 ratio as published in Lombardo & Eichinger, The Career Architect Development Planner, 1996. ↩
- [7]Drucker, Peter F. The Essential Drucker: The Best of Sixty Years of Peter Drucker’s Essential Writings on Management. Collins Business (HarperCollins), 2007. ↩
- [8]Whitmore, John. Coaching for Performance: GROWing People, Performance and Purpose. 3rd ed. People Skills for Professionals. Nicholas Brealey, 2002. ↩