Google Studied 180 of Its Own Teams to Build the Perfect One. The Recipe Surprised It.

Google spent years and a frankly absurd amount of data trying to answer one question: what makes a team work? They went in expecting to find the perfect recipe of people. The right mix of star engineers, complementary skills, the correct personalities in the correct seats.

They didn't find it. The "who" barely mattered.

We make org chart software, so we'll say it plainly: an org chart is a picture of the who. The names, the titles, the boxes, the lines between them. It is a clean utility for mapping how an organization is arranged, and a good one is genuinely useful. It also doesn't pretend to do more than that. What makes the team inside the chart actually perform is a separate question, and the most data-obsessed company on earth ran the numbers on it. The gap between what a chart shows and what makes a team work is the most useful thing a manager can understand.

Key Takeaways

Google's Project Aristotle studied 180+ of its own teams and found that who is on a team matters far less than how the team works together.

The single biggest factor was psychological safety: whether people feel safe to ask questions, admit mistakes, and disagree without looking incompetent.

The five factors, in order, were psychological safety, dependability, structure and clarity, meaning, and impact.

The honest caveat: this was internal corporate research, not a peer-reviewed study, and a credible camp of psychologists argues personality matters more than Google found.

An org chart shows the structure. It can't show the trust. Knowing the difference is the whole job.

What was Google's Project Aristotle?

Project Aristotle was a research effort Google launched in 2012 to figure out what makes its teams effective. It was named for the line usually attributed to Aristotle, "the whole is greater than the sum of its parts," which turned out to be either prophetic or ironic depending on how you read the results.

The scale was real. Led by Google's People Analytics group, the team studied more than 180 teams, ran a couple hundred interviews, and looked at around 250 different team attributes through more than 35 statistical models. The goal was to find the variables that actually predicted whether a team did well.

The starting hypothesis was the intuitive one, the same one most of us carry around: build the best team by collecting the best people. Get the mix right. A brilliant individual contributor here, a seasoned manager there, the right blend of extroverts and introverts, the right seniority spread. Assemble talent, get performance.

The data refused to cooperate. As the manager who led the analytics work, Abeer Dubey, told The New York Times, they had heaps of data and nothing in it showed that a specific mix of personality types or skills or backgrounds made any difference. The "who" part of the equation, in his words, just didn't seem to matter.

What mattered was how the team operated. Less the roster, more the norms. Less who was in the boxes, more what happened in the space between them.

What are the five keys to an effective team?

Google's researchers landed on five dynamics that separated the strong teams from the rest. They're listed here in the order the research ranked them, which matters, because the first one outweighed the rest by a wide margin.

1. Psychological safety. Can you take a risk on this team without feeling insecure or embarrassed? This was the big one, the factor underneath all the others.

2. Dependability. Do people reliably do what they said they'd do, to the standard the work needs? Google's actual survey item was blunt: when teammates say they'll do something, they follow through.

3. Structure and clarity. Are roles, plans, and goals clear? Does the team have an effective way to make decisions? (This is the one an org chart gestures at and, on its own, mostly fails to deliver. More on that below.)

4. Meaning. Is the work personally important to the people doing it?

5. Impact. Do people believe their work matters and changes something?

Here's the part worth sitting with. Of those five, only one, structure and clarity, is even partly visible on an org chart. The other four live entirely in behavior. You cannot draw psychological safety. You cannot export dependability to PowerPoint. The chart captures the skeleton; the things that actually keep a team alive are soft tissue, and no diagram has ever rendered them.

What is psychological safety, really?

Psychological safety is the shared belief that a team is safe for interpersonal risk-taking. It's the freedom to ask the question you're worried sounds dumb, admit the mistake before it metastasizes, and disagree with the senior person in the room without quietly updating your résumé that night.

The concept predates Google. The researchers didn't invent it; they went looking for what explained their data and found Amy Edmondson's 1999 research on the idea, which filled the gap their statistical models couldn't. Edmondson, a Harvard Business School professor, had been studying it for years.

Her origin story for the concept is the most useful thing in it. Studying hospital teams, she expected better teams to make fewer errors. The data showed the opposite: the better-rated nursing teams reported more errors. For a moment that looks like a disaster. It wasn't. The strong teams weren't making more mistakes; they felt safe enough to report the ones they made, instead of burying them. The weak teams were just as error-prone and a lot quieter about it. Open teams surface problems early, while problems are still small and cheap to fix.

This is where most write-ups of psychological safety go soft and wrong, so let's be precise about what it is not.

Psychological safety is not niceness. It is not everyone agreeing, keeping the peace, and lowering the bar so no one feels bad. A team where nobody disagrees isn't safe; it's anesthetized. Edmondson's whole point is that safety exists in service of high standards, not instead of them. The goal isn't comfort. The goal is a team that can tell each other hard truths and get better because of it.

The mistake is treating "safe" and "demanding" as opposites you trade off. They're independent dials. Low on both, you get apathy. High demands but no safety, you get anxiety, where people hide problems to avoid blame. High safety but no demands, you get a pleasant team that ships nothing. You want both turned up: a team that holds a high bar and feels safe enough to be honest about whether they're clearing it.

The practical upshot is genuinely good news for managers. Edmondson's data shows psychological safety varies team to team within the same company. It isn't a culture handed down from the top that you're stuck with. It's built locally, by whoever runs the team, in how they react the first time someone admits they're behind or pushes back in a meeting. You don't need permission from the org to build it. You need to not punish the next person who tells you something you didn't want to hear.

What does this have to do with your org chart?

Everything, and almost nothing, which is exactly the point.

An org chart is the single clearest artifact of the "who." It shows reporting lines, titles, headcount, who sits under whom. It's genuinely useful for that. When someone is new, when a team has grown past the size you can hold in your head, when you need to explain the structure to a board or a new hire, a clear chart does real work. Worth drawing, worth keeping current.

But notice what Google's five factors do to the chart's reputation. Four of the five things that make a team work do not appear on it at all. The chart shows you the structure and clarity bucket, partially, and goes completely silent on safety, dependability, meaning, and impact. You can have a flawless org chart, every box aligned, every line crisp, sitting on top of a team where nobody trusts anyone and half the commitments quietly slip. The diagram will look great. The team will not be fine.

This is the trap the chart sets, and it's worth naming because it's so easy to fall into. Redrawing the boxes feels like fixing the team. It's visible, it's satisfying, it produces an artifact you can point to in a meeting. But moving names around a diagram doesn't create trust, and it doesn't make people follow through. If the real problem is that your team can't be honest with each other, no reorg solves it. You'll just have the same problem in a tidier shape.

So here's the honest version of what an org chart is for. It's a map, not the territory. It tells you the formal structure, which is real and which matters. It tells you nothing about whether the people in those boxes feel safe enough to do their best work. The chart is a symptom of a healthy team, not a cause of one. Draw it clearly, keep it current, and then go do the actual work, which happens in conversations no diagram can hold.

Is Project Aristotle actually settled science?

No, and you should be a little suspicious of anyone who presents it as gospel, including the dozens of leadership blogs that have spent a decade citing it as proof of whatever they were already selling.

Two honest caveats, because the brand that taught you to distrust airport WiFi pricing is not going to hand you a tidy research finding without the asterisks.

First, it was never peer-reviewed. Project Aristotle was internal Google research, popularized for the rest of us by journalist Charles Duhigg in a 2016 New York Times Magazine piece. That doesn't make it wrong. Google had more data on real teams than most academics will ever touch. But internal corporate research that became a viral story is a different animal from a replicated, peer-reviewed study, and it's worth holding it as "Google's compelling findings" rather than "settled science."

Second, there's a credible camp that thinks the "who" matters more than Google found. Business psychologist Tomas Chamorro-Premuzic, who has run a major personality-assessment firm, and his collaborator Dave Winsborough argue that personality and team composition predict performance more than the Aristotle story suggests. Their reading: the who and the how aren't really separable, because who is on a team shapes how that team behaves in the first place. A team's norms don't come from nowhere; they come partly from the people in it.

The fair synthesis is less dramatic than the viral headline, and more honest. Google didn't prove that people don't matter. Of course they matter; you can't staff a team with the wrong people and norm your way to greatness. What Google found is that how people treat each other outweighs their résumés on paper, and that the best collection of individuals will still underperform if they can't be honest with one another. That's a sturdier claim than "talent is irrelevant," and it's the one the evidence actually supports.

It also happens to be more useful, because you can act on it. You can't always pick your team. You can almost always change how it operates.

What can a manager actually do on Monday?

Read the research and the practical advice converges on a few unglamorous things. None of them require a reorg.

Treat the chart as a starting point, not the finish line. Get the formal structure clear and correct, because ambiguity about who does what is its own quiet team-killer. The fastest way to get there is usually to build the chart from a spreadsheet you already have rather than redrawing boxes by hand. Then stop pretending the structure is the substance. A clean chart buys you clarity, not trust.

React well the next time someone admits a mistake. This is the single highest-leverage thing in the whole literature, and it costs nothing. The first time a report tells you something went wrong and you respond with curiosity instead of blame, you've done more for your team's performance than any restructuring. The first time you respond with blame, you've taught everyone watching to go quiet.

Frame the work as something you're all still figuring out. Edmondson's advice for building safety is almost suspiciously simple: treat the work as a learning problem rather than a pure execution problem, admit your own fallibility out loud, and ask a lot of questions. Managers who act like they have all the answers get teams that stop offering theirs.

Watch dependability, not just talent. The flashy hire who doesn't follow through hurts a team more than a steady one who does. Google's second factor was reliability for a reason. Notice who closes the loop and who quietly drops things, and treat that as the signal it is.

The org chart will tell you none of this. It'll tell you who reports to whom, which is worth knowing and worth getting right. But it can't tell you whether the people in those boxes trust each other. That part lives on no diagram. You build it one honest conversation at a time.

Frequently asked questions

What is Project Aristotle in simple terms?

It was Google's internal research project, started in 2012, that studied over 180 of its own teams to find what makes a team effective. The headline finding: how a team works together matters far more than who is on it, with psychological safety as the single most important factor.

What are the five keys to a successful team according to Google?

In order of importance: psychological safety, dependability, structure and clarity, meaning, and impact. Psychological safety, the freedom to take interpersonal risks without fear, outweighed the other four by a significant margin.

Is psychological safety the same as being nice?

No. Psychological safety is the freedom to speak up, disagree, and admit mistakes. It exists to support high standards, not replace them. A team where everyone agrees to keep the peace is not psychologically safe; it's just quiet, and it tends to hide problems until they're expensive.

Does an org chart show team health?

No. An org chart shows formal structure: reporting lines, titles, and headcount. Four of Google's five effectiveness factors, including trust and follow-through, don't appear on it at all. A chart is useful for clarity, but it can't tell you whether a team actually works well together.

Was Project Aristotle a peer-reviewed study?

No. It was internal corporate research, popularized by journalist Charles Duhigg in 2016. The findings are compelling and data-rich, but they were never published as peer-reviewed academic work, and some psychologists argue it understated how much team composition and personality matter.