How many people should be on your team? It is one of the most common questions in management, and the research has a surprisingly consistent answer. Across disciplines spanning social psychology, organizational behavior, military science, and software engineering, the evidence converges on a range: 4 to 6 people for most work, with diminishing returns after 10.
The reasons are structural, not cultural. As team size increases, coordination costs grow exponentially, individual effort declines, and communication becomes fragmented. Here is what the research says, and why it matters for how you structure your organization.
Dunbar's Number: The Cognitive Limit
British anthropologist Robin Dunbar proposed in 1992 that humans can maintain stable social relationships with approximately 150 people, based on the correlation between primate brain size and social group size. But Dunbar's model includes nested layers: 5 (intimate support group), 15 (close friends), 50 (band), 150 (clan) (Dunbar, 1992).
The innermost layer of 5 is the most relevant for work teams. This is the number of people with whom you can maintain the kind of deep, bidirectional knowledge required for true collaboration: understanding each other's strengths, communication styles, current workload, and emotional state. Beyond 5, the cognitive overhead of tracking those relationships begins to degrade performance.
Hackman and Vidmar: The 4.6 Sweet Spot
Harvard psychologist Richard Hackman and colleague Neil Vidmar ran experiments with teams of 2 to 7 people performing decision-making and judgment tasks. Participants rated their satisfaction with team size after the task. The ideal size, balancing performance and member satisfaction, was 4.6 (Leading Beat).
People preferred teams of 4 to 5. Anything below 4 felt too small to generate diverse perspectives. Anything above 5 introduced process loss: time spent coordinating, clarifying roles, and managing social dynamics rather than doing the actual work. Hackman later observed that teams performing highly interdependent work function most effectively when relatively small, and that teams larger than approximately 10 struggle to maintain engagement, clarity, and accountability.
The Ringelmann Effect: Why Bigger Teams Try Less Hard
In the early 1900s, French agricultural engineer Max Ringelmann measured the force exerted by individuals and groups pulling on a rope. He found that individual effort decreased as group size increased. One person pulled approximately 63 kg of force. Two people pulled 118 kg (a loss of 8 kg from the expected 126). Three pulled 160 kg (a loss of 29 kg from the expected 189) (BVOP Journal).
This effect, later termed social loafing, has been replicated across dozens of studies and task types. A meta-analysis of 78 studies found that social loafing is robust and generalizes across tasks and populations (Karau & Williams, 1993). The causes are diffusion of responsibility (my contribution matters less when there are more people) and coordination loss (harder to synchronize effort in larger groups).
Miller's 7 Plus or Minus 2
George Miller's 1956 paper "The Magical Number Seven, Plus or Minus Two" described the limits of human working memory. While Miller's research focused on information processing, not team size, the principle applies: humans can hold approximately 7 items in working memory simultaneously. In a team context, this translates to the number of people's contributions, roles, and perspectives you can actively track during a conversation or collaboration session.
When team size exceeds 7, members begin to lose track of who said what, who owns what, and what has already been discussed. This is why meetings with 8+ people tend to produce fewer decisions per minute than meetings with 4 to 5 people.
Amazon's Two-Pizza Rule
Jeff Bezos formalized the small-team principle at Amazon with the "two-pizza team" rule: no team should be so large that it cannot be fed by two pizzas, which works out to roughly 6 to 10 people (AWS Executive Insights). The rationale was reducing communication lines and increasing ownership. In a team of 6, there are 15 communication channels. In a team of 10, there are 45. In a team of 15, there are 105. Each additional person adds n-1 new communication paths, not one.
Gallup: Manager Engagement Peaks at 8 to 9 Reports
Gallup research on span of control found that manager engagement peaks at around 8 to 9 direct reports and declines as the span widens beyond that point (Gallup, 2024). The average number of direct reports increased from 10.9 in 2024 to 12.1 in 2025, a trend driven by cost-cutting layoffs that removed management layers without reducing workload. Managers with excessively wide spans report higher workload demands, more conflict, and less perceived control.
The median team size, however, holds steady at 5 to 6 employees per manager. A minority of very large teams pulls the average up. The implication: if your span exceeds 9, you are managing the structure more than the people.
Brooks's Law: Adding People Makes Late Projects Later
Fred Brooks observed in The Mythical Man-Month (1975) that adding people to a late software project makes it later. The reasons are onboarding cost (new people must be educated about preceding work, which diverts existing team members) and communication complexity (each additional person adds n-1 new communication channels) (Wikipedia). Brooks's Law is an extreme case of the team size problem: the point at which coordination costs exceed the marginal value of additional contributors.
What This Means for Your Organization
The research consensus:
- For deep collaboration: 4 to 5 people (Hackman, Dunbar's inner circle)
- For autonomous product teams: 6 to 8 people (Amazon two-pizza, Katzenbach and Smith)
- For manager span of control: 8 to 9 direct reports maximum for sustained engagement (Gallup)
- Hard ceiling: Beyond 10, process loss reliably exceeds the value of additional perspectives (Ringelmann, Brooks)
QuestWorks sessions are designed around this research. The flight simulator for team dynamics runs with groups of 2 to 5, deliberately sized at the small-group sweet spot where every voice matters, social loafing is structurally impossible, and coordination costs stay low. For organizations with larger teams, QuestWorks uses dynamic grouping to create quest groups within the optimal range from any team size. Sessions run 25 minutes on its own cinematic, voice-controlled platform, with Slack as the integration layer for install and onboarding.
QuestDash surfaces behavioral patterns at the individual and team level. Leaders get aggregate trends and strengths-based XP highlights through a weekly team health report. HeroGPT provides private AI coaching in Slack that never shares upstream. Everything is voluntary and never tied to performance reviews. $20/user/month, 14-day free trial.
The next time someone proposes adding a sixth person to a four-person team "for capacity," show them the Ringelmann data. The math does not care about intentions.
Try it free
