What we know about how teams
work — and why they don’t

The dominant belief about team performance is that the right people, working hard, with a clear goal, will deliver results. It’s intuitive. It’s also consistently wrong.

Decades of research in organisational psychology, behavioural science, and systems thinking have converged on a more uncomfortable finding: team performance is primarily a function of conditions, not capability. The quality of the people matters — but it explains far less of the variance in outcomes than most leaders assume.

In 2012, Google launched Project Aristotle: a five-year study of 180 internal teams designed to answer one question — what makes a team effective? The hypothesis was that the best teams would share certain qualities: similar backgrounds, complementary skills, high individual IQ, or strong social bonds. None of these predicted team performance.

The single strongest predictor — by a significant margin — was psychological safety: the shared belief that the team is a safe environment for interpersonal risk-taking. On teams with high psychological safety, members were more willing to raise problems, admit mistakes, ask questions, and offer ideas. They made better decisions, learned faster, and delivered more.

The implication is profound: you can assemble a team of exceptional individuals and achieve mediocre results, while a team of ordinary people — working in the right conditions — can consistently outperform them. Psychological safety isn’t a personality trait or a culture initiative. It’s a structural condition. It either exists in how the team operates, or it doesn’t.

Richard Hackman, Professor of Social and Organisational Psychology at Harvard, spent four decades studying what makes teams work. His conclusion was sobering: approximately 60% of the variance in team performance is explained by conditions established before the team begins its work — the clarity of its purpose, the soundness of its structure, and the quality of its design.

Hackman identified five conditions that enable team effectiveness:

• A real team — bounded, stable membership with clear interdependence. Many “teams” are actually working groups: people who share a manager but work independently. They need different management, not team interventions.
• A compelling direction — a purpose that is challenging, clear, and consequential. Vague or unmotivating goals produce compliant effort, not committed performance.
• An enabling structure — team size, composition, and task design that make excellent performance possible rather than preventing it. Most teams are too large, poorly composed, or working on tasks that don’t require genuine interdependence.
• A supportive context — the organisational conditions (information, resources, recognition, training) that allow the team to do its work without having to fight the system.
• Expert coaching — team-focused, not individual, coaching at the right moments in the team’s development cycle.

Most team performance interventions address one of these conditions — usually coaching — while leaving the others untouched. This explains why so many well-intentioned programmes fail to produce lasting change.

Trust is the most cited factor in team performance and the least well understood. Most leaders think of trust as a feeling — something that develops over time through shared experience. That model isn’t wrong, but it’s incomplete.

Amy Edmondson’s research distinguishes between interpersonal trust (confidence in others’ reliability and intentions) and psychological safety (confidence that the environment is safe for risk-taking). They are related but not the same — and they have different drivers.

Trust is also not a single thing. Research identifies at least three distinct types relevant to team performance:

• Competence trust — belief that teammates have the skills to do what they say they’ll do.
• Integrity trust — belief that teammates will behave consistently with their stated values and commitments.
• Benevolence trust — belief that teammates genuinely have your interests at heart, not just their own.

Teams can be high on one dimension and low on another — and the specific pattern matters enormously for how problems manifest and what interventions will work. A team with high competence trust but low benevolence trust looks very different from one with the opposite profile, even if both are described as “having trust issues.”

One of the most counterintuitive findings in team research is that teams can deteriorate without any identifiable incident or failure. No conflict. No departure. No strategic change. Just a gradual erosion of the conditions that enabled performance.

This happens because teams are living systems. The conditions that work at formation — when roles are clear, purpose is fresh, and relationships are new — often stop working as the team matures, grows, or encounters sustained pressure. What was alignment becomes assumption. What was accountability becomes avoidance. What was healthy challenge becomes silence.

The most dangerous phase is often when a team believes it is performing well. The absence of visible problems is not the same as the presence of high performance. Many teams are functioning effectively at 60–70% of their potential without anyone — including the leader — being able to name what is holding them back. This is the diagnostic gap. And it’s what most performance interventions fail to close — because they respond to the symptoms that are visible, not the conditions that created them.

The research on team performance interventions is humbling. A 2008 meta-analysis by Shuffler, DiazGranados, and Salas reviewed decades of studies on team training and found that most interventions produce short-term behavioural changes that fade within weeks without reinforcement. The ones that produce lasting change share three characteristics:

• They address specific, diagnosed conditions — not generic team development. Teams that received targeted, diagnosis-informed interventions showed significantly greater and more durable performance gains than those receiving standardised programmes.
• They involve the whole team system — not just the leader or a subset of members. Change in one part of the system without corresponding change in related parts is quickly absorbed and reversed.
• They are followed by structured reinforcement — regular attention to the conditions being developed, not a one-time event followed by return to normal operations.

The implication is direct: generic team-building, off-the-shelf programmes, and leadership coaching that doesn’t address the team system are unlikely to produce the results leaders hope for. Not because the interventions are poor — but because they are not matched to the actual cause of the problem.

Matching the intervention to the cause requires knowing the cause. That’s what the diagnostic is for.

In 1968, Austrian biologist Ludwig von Bertalanffy published General System Theory, proposing that the principles governing living organisms — interdependence, feedback, emergent behaviour — applied equally to all complex systems, from ecosystems to economies to organisations. The insight was radical: the whole is not merely the sum of its parts. It is shaped by the relationships between the parts.

A team is a system in this precise sense. Its performance cannot be understood by analysing individual members in isolation. It emerges from the interactions between them — the patterns of communication, accountability, trust, and shared meaning that form through repeated contact over time. Change one part of the system, and every connected part responds.

In 1990, Peter Senge introduced systems thinking to a leadership audience through The Fifth Discipline. His argument: organisations fail not because of bad intentions or incompetent people, but because their leaders are trained to think in straight lines (cause → effect) while the systems they lead operate in circles (cause → effect → feedback → cause).

Senge identified several “learning disabilities” that afflict teams and organisations precisely because of this mismatch:

• “I am my position.” People focus on their own role and lose sight of how their actions affect the broader system. Problems that cross boundaries go unowned.
• “The enemy is out there.” When something goes wrong, teams look for external causes. The internal dynamics — the feedback loops their own behaviour creates — remain invisible.
• “The illusion of taking charge.” Leaders respond to symptoms with visible, decisive action. The underlying system dynamics continue unchanged. The problem returns, often worse.
• “The boiling frog.” Gradual deterioration in team performance isn’t noticed because each change is small. By the time the problem is visible, the system is already in significant distress.

These aren’t failures of intelligence or effort. They are the predictable consequences of trying to manage a system with a non-systemic mental model.

Donella Meadows, a systems scientist at MIT and Dartmouth, spent her career identifying where interventions in complex systems actually produce change — and where they don’t. Her 1999 paper Leverage Points: Places to Intervene in a System identified twelve points of intervention, ranked from least to most effective.

The counterintuitive finding: the most obvious places to intervene are usually the least effective. Adjusting numbers (budgets, headcounts, timelines) produces the smallest change. Shifting information flows, feedback loops, and ultimately the goals and mental models of the system produces the largest.

Applied to teams: training individuals in new skills is a low-leverage intervention. Changing the feedback mechanisms — how performance is visible, how accountability is held, how information flows between members — is higher leverage. Shifting the shared mental model of what the team is for and how it should operate is the highest leverage of all. This is why the diagnostic focuses on the system — not on any individual’s performance or development needs. The leverage lies in the conditions, not the capabilities.

One of the most important — and most overlooked — features of team systems is that they are self-reinforcing. The conditions a team operates in are not simply imposed from outside. They are continuously recreated by the team’s own behaviour.

A team where it is unsafe to raise problems will not raise problems. Because problems go unraised, they compound. The team’s performance suffers, which increases pressure, which makes it even less safe to raise problems. This is a reinforcing feedback loop — a vicious cycle that amplifies itself over time without any external change.

The same logic runs in the other direction. A team with high psychological safety raises problems early. Problems addressed early don’t compound. Performance remains strong, which reduces pressure, which reinforces the sense that it is safe to speak up. A virtuous cycle.

What distinguishes high-performing teams from underperforming ones is not that the former have fewer problems. It is that they have feedback systems that surface problems early, at low cost, before they become crises. The diagnostic identifies whether those feedback systems exist — and where they are breaking down.

Emergence is the systems theory concept that explains why a group of individually capable people can produce collectively poor results — and why a seemingly ordinary group can produce extraordinary ones.

Emergent properties are characteristics of a system that cannot be predicted from the properties of its components. Water is wet; hydrogen and oxygen are not. A team that trusts each other will make decisions its members could not make individually. A team in conflict will fail to use the capabilities each member brings.

This has a direct implication for leaders: you cannot build a high-performing team by focusing exclusively on individual development. The performance lives in the space between people — in the quality of their interaction, the clarity of their shared understanding, and the conditions that make those interactions productive. That space is not visible to the people inside it. It requires an external view to see clearly.

Senge uses a supply chain simulation — known as the beer game — to make systems dysfunction visceral. In the game, players take on the roles of retailer, wholesaler, and manufacturer, each making independent ordering decisions based on the information available to them locally. There is no communication between players. No one is trying to fail.

What happens is always the same. A small uptick in consumer demand triggers a cascade of overordering up the chain. By the time the manufacturer responds, the retailer has already been flooded with inventory. Costs spike. Orders crash. The system oscillates wildly around a point of equilibrium it never quite reaches — not because any individual made a bad decision, but because the structure of the system made coordination impossible.

The beer game has been run with thousands of executives, MBA students, and senior leadership teams. The result is always the same. And when debriefed, players consistently blame the other players — the wholesaler over-ordered, the manufacturer was too slow. The systemic cause — poor information flow, no shared visibility, decisions made in isolation — goes unidentified until it is pointed out.

This is exactly what the diagnostic is designed to surface: not which person is the problem, but where the system is generating dysfunction that rational, capable people cannot solve from inside it.

In 1896, Italian economist Vilfredo Pareto was studying land ownership in Italy. He observed that 80% of the land was owned by 20% of the population. Curious, he looked at other countries. The same ratio appeared. He looked at other domains. It appeared again. Pareto had discovered something fundamental about how outcomes distribute in complex systems: they are almost never equal. A small number of inputs account for a disproportionately large share of results.

This became the Pareto Principle — more commonly known as the 80/20 rule. It has since been validated in domains as varied as software engineering (20% of bugs cause 80% of crashes), healthcare (20% of patients account for 80% of costs), and sales (20% of customers produce 80% of revenue).

In team performance, the pattern is consistent. Roughly 20% of the conditions shaping a team’s performance account for approximately 80% of the friction. Not every dimension of how a team operates is equally weighted. Two or three factors — the specific ones the diagnostic identifies — are doing most of the damage. Everything else is secondary.

This is why broad-based team interventions so rarely work. When you address everything, you address nothing with sufficient depth or precision. Resources are spread across ten issues when concentrating them on two would move the system. The effort is visible. The results are not.

The Pareto Principle also explains why the same generic intervention produces dramatically different results in different teams. A communication workshop will transform a team whose root cause is communication breakdown and do nothing for a team whose root cause is strategic misalignment. The intervention isn’t wrong — it’s just applied in the wrong place.

Finding the 20% is the diagnostic’s primary purpose. Not a comprehensive picture for its own sake — but a precisely targeted view that tells you where a small, focused intervention will produce the greatest impact on team performance.

Donella Meadows called these points “leverage points” — places in a system where a small shift produces large change. Pareto’s contribution was demonstrating that they are not randomly distributed. They are consistently concentrated. You can find them. And when you do, the intervention becomes straightforward — not because the problem is simple, but because you finally know where it is.

In 1913, French agricultural engineer Maximilien Ringelmann asked individuals and groups to pull on a rope, and measured the force exerted. The finding was uncomfortable: the more people in the group, the less effort each person applied. A group of eight pulled with less than half the combined force of eight individuals working alone.

This wasn’t laziness. It was diffusion — the natural human tendency to reduce individual effort when contribution feels invisible or unaccountable. Ringelmann called it “social loafing.” It has since been replicated across dozens of contexts, cultures, and task types.

The implication for team leaders: size isn’t strength. And accountability — genuine, visible, mutual accountability — isn’t a cultural value. It’s a structural condition. It either exists in the way a team is designed, or it doesn’t.

The diagnostic measures this directly. Not by asking whether people feel accountable — but by identifying whether the structural conditions that make accountability possible are present or absent in the way the team actually operates.

In 2001, Jim Collins published Good to Great — a study of eleven companies that made the leap from good to exceptional performance and sustained it for at least fifteen years. One of his most cited findings: great leaders didn’t start with a vision or a strategy. They started by getting the right people on the bus, the wrong people off it, and the right people in the right seats.

“The right people on the bus” has become one of the most repeated ideas in leadership. It’s also one of the most frequently misapplied.

Collins was not arguing that assembling talented individuals creates high performance. He was arguing something more specific: that people with the right values, work ethic, and character create the conditions in which strategy and structure can actually work. “Right people,” in Collins’s framework, is not primarily about competence. It is about character and fit.

The distinction matters because it changes what leaders should be doing when performance suffers. If the problem is a values misalignment or an accountability deficit, leadership development and culture work may be appropriate. If the problem is something else — unclear purpose, poor process design, trust deficits, misaligned priorities — the same intervention will achieve nothing.

What distinguishes high-performing teams is not simply who is on the team. It is whether the conditions exist for those people to work at their best. That is a diagnostic question — not an intuitive one.

The meeting that ends without a decision. The feedback that doesn’t land. The question that doesn’t get asked because the relationship doesn’t quite support it. The assumption that goes unchecked because it would be uncomfortable to challenge.

These aren’t communication failures. They’re trust failures, expressed through conversation. And trust — the specific kind that enables performance — isn’t built through team days or values workshops. It’s built through consistently safe, clear, accountable interaction over time.

Amy Edmondson’s research distinguishes between interpersonal trust (confidence in others’ reliability and intentions) and psychological safety (confidence that the environment is safe for risk-taking). They are related but not the same — and they have different drivers.

Research identifies at least three distinct types of trust relevant to team performance. Competence trust — belief that teammates have the skills to do what they say they’ll do. Integrity trust — belief that teammates will behave consistently with their stated values and commitments. Benevolence trust — belief that teammates genuinely have your interests at heart, not just their own.

Teams can be high on one dimension and low on another — and the specific pattern matters enormously for how problems manifest and what interventions will work. A team with high competence trust but low benevolence trust looks very different from one with the opposite profile, even if both are described as “having trust issues.”

The diagnostic measures the conditions that either support or undermine this kind of trust. Not by asking “do you trust your team?” — but by identifying the structural conditions that make trust available or unavailable.

In 1982, General Motors closed its Fremont, California assembly plant. It had the worst productivity and quality of any GM facility in North America. Absenteeism ran at 20%. Wildcat strikes were routine. Workers had filed more than 5,000 outstanding grievances. Leadership and the UAW local were in open war. The plant was, by every measure, broken.

Two years later, in 1984, GM and Toyota jointly reopened the same plant under a new name — NUMMI: New United Motor Manufacturing, Inc. They rehired 85% of the same workforce, including the most militant union members. The building was unchanged. The union contract was unchanged. The equipment was largely the same.

Within two years, NUMMI was producing the highest-quality vehicles of any GM plant in North America. Absenteeism had fallen to 2%. Grievances had dropped from thousands to a handful. The workforce — the same people who had been the worst in the system — were now among the best.

Toyota did not change the people. They changed the system the people worked in. Specifically, they changed three things.

First, they gave workers the authority to stop the production line. Every worker had a cord they could pull — the “andon cord” — to halt assembly if they spotted a defect. At GM, stopping the line meant discipline. At NUMMI, it meant you had done your job. Problems were surfaced immediately, at the point of origin, before they compounded downstream.

Second, they replaced individual inspection with team accountability. Workers were organised into small teams, responsible collectively for their section of the line. Quality was no longer a function of inspectors catching problems after the fact. It was a function of teams owning their process in real time.

Third, they invested in standardised work — not as a constraint on workers, but as a baseline from which improvement could be measured and shared. Variation was the enemy of quality. Standardisation made variation visible.

The workers who had been labelled unmotivated, adversarial, and incapable were not the problem. They were responding rationally to the system they were in: one that punished transparency, removed accountability, and rewarded compliance over performance. Change the system, and the behaviour changed with it — immediately, and dramatically.

This is the most important thing the NUMMI story proves: the diagnostic assumption that performance is a people problem is almost always wrong. People perform within the conditions available to them. When the conditions change, the performance changes. Not gradually. Not after cultural transformation programmes. Immediately.

The diagnostic exists to identify which conditions are limiting performance — because those are where the leverage is. The people, in the overwhelming majority of cases, are not.

In 2010, a research team led by Anita Woolley at Carnegie Mellon and Thomas Malone at MIT set out to answer a simple question: do teams have an intelligence? Not the intelligence of their individual members — but a collective capacity that predicts how well the group performs across a wide range of tasks.

The answer was yes. They called it the “c factor” — collective intelligence — and they measured it using a diverse battery of tasks: visual puzzles, negotiation exercises, moral reasoning challenges, architectural design problems. Teams that performed well on one type of task tended to perform well on all of them, just as individuals with high general intelligence tend to outperform across cognitive domains.

The c factor was real, measurable, and consistent. The finding that followed was the one that changed the conversation.

The researchers tested whether collective intelligence could be predicted by the average individual IQ of team members. It could not. They tested whether it was predicted by the highest individual IQ on the team — the “smartest person in the room” effect. It was not. They tested motivation, cohesion, and team satisfaction. None predicted the c factor.

Three variables predicted it. The first was average social perceptiveness — how accurately team members could read each other’s emotional states, measured by the “Reading the Mind in the Eyes” test. Teams whose members were better at reading each other performed better collectively.

The second was equality of communication — how evenly turn-taking was distributed. Teams where one or two people dominated the conversation performed worse. Teams where everyone contributed performed better, even when the conversation took longer.

The third was the proportion of women on the team — which the researchers noted was largely explained by the first two factors, since women on average score higher on social perceptiveness tests and tend to distribute communication more evenly.

If you want to build a high-performing team, hiring the smartest people is the wrong strategy. The research is unambiguous on this point. What predicts collective performance is not raw capability — it is the quality of the interactions between people, and the conditions that shape those interactions.

A team where one person dominates, where emotional signals go unread, where some voices are structurally absent from the conversation — that team is leaving most of its collective intelligence on the table, regardless of the credentials in the room.

The diagnostic measures the conditions that either enable or suppress collective intelligence. Psychological safety, communication quality, inclusion, and the quality of listening are not soft metrics. They are the variables that a peer-reviewed study published in Science identified as the primary predictors of how well a team actually thinks together.

When Amy Edmondson began her research into team performance in hospital settings in the late 1990s, she expected to find a straightforward relationship: better teams would make fewer mistakes. She had developed a measure of psychological safety — the shared belief that a team is safe for interpersonal risk-taking — and predicted it would be highest in teams that performed best.

The data came back inverted. The teams with the highest psychological safety scores were reporting more errors, not fewer. The teams with the lowest safety scores were reporting the fewest. If the theory was right, the data looked wrong.

Edmondson’s insight was to ask a different question: were the high-safety teams actually making more errors — or were they more willing to report them?

The answer, confirmed through subsequent research and independent observation of the teams, was the latter. Errors were occurring at roughly the same rate across all teams. But in high-safety teams, mistakes were surfaced immediately — discussed openly, learned from, and prevented from compounding. In low-safety teams, mistakes were hidden. Not because the teams were unethical, but because the cost of speaking up — judgement, blame, embarrassment — was perceived as higher than the cost of silence.

The low-error-reporting teams were not performing better. They were managing information better. And the information they were suppressing was exactly the information needed to prevent the next error.

This finding has a direct application to how leaders interpret team behaviour. A team where no one raises problems is not a team without problems. It is a team where the cost of raising problems is perceived as too high. The silence is data — but it reads as competence to anyone not looking closely enough.

In diagnostic terms: a team that scores low on psychological safety is not a comfortable team. It is a team that is accumulating errors, misalignments, and frustrations below the surface — accumulations that will eventually manifest as something larger and harder to fix.

The best teams are not the quietest ones. They are the ones where the difficult things get said early, at low cost, before they become crises. That capacity — the willingness to surface what is not working — is precisely what psychological safety enables. And it is precisely what its absence destroys.

The diagnostic is built on more than 25 years of applied research into what actually differentiates high-performing teams from those that underperform despite capable people. It draws on validated frameworks across organisational psychology, systems thinking, and leadership development.

The dimensions were developed through direct application across more than 80,000 individual assessments. They are not theoretical constructs — they are the recurring drivers that explain most of the variance in team performance across industries, team sizes, and organisational contexts.

The following references inform the diagnostic framework, its dimensions, and the conceptual models used in our reporting. Sources span peer-reviewed journal articles, meta-analyses, and established practitioner texts across organisational psychology, systems thinking, and leadership development.

APA 7th edition. Full internal reference list with constructs and business relevance available on request.