Team Formation Optimizer

Two arithmetic facts drive most of team-size reasoning. First, the number of distinct pairwise communication links in a group of n people is n(n-1)/2, the edge count of a complete graph: 6 links at 4 people, 15 at 6, 28 at 8, and 66 at 12. Coordination cost tracks that quadratic while output at best tracks n linearly, which is why adding people eventually subtracts throughput. Second, individual effort falls as groups grow, an effect this model approximates as a flat 5 percent decrement per member beyond four, so a team of eight delivers about 80 percent of each person's solo output, roughly 6.4 person-equivalents in total. Optimal size is set here by task complexity at 3 for low, 5 for medium, 7 for high, and 9 for very high, which brackets the 5 to 9 range most practitioners settle on for knowledge work. Several assumptions deserve scrutiny. The linear effort penalty is a stylised heuristic rather than a measured coefficient, and it breaks down badly at large n, implying zero output somewhere past 24 members. Real teams do not communicate as complete graphs; deliberate structure such as an interface owner, subteams, or written asynchronous updates cuts the effective link count well below n(n-1)/2, which is precisely how large organisations function at all. Seniority is treated as a single uniform percentage, ignoring that one architect and one experienced tester are not interchangeable. Skill coverage is a ratio heuristic that rewards matching skill count to headcount and cannot know whether the skills present are the ones the work actually needs. Use the output to compare candidate splits, not as an absolute forecast.

The empirical foundation is unusually old. Max Ringelmann, a French agricultural engineer, ran rope-pulling and cart-pushing trials in the 1880s to measure how much work men and horses could deliver in groups, and found that combined pull rose far more slowly than headcount, with groups of eight producing roughly half the sum of their individual efforts. His results were published in 1913 in the Annales de l'Institut National Agronomique and largely forgotten for decades before social psychologists rediscovered them. Ivan Steiner formalised the logic in Group Process and Productivity in 1972 with the framing that actual productivity equals potential productivity minus process losses, splitting those losses into coordination and motivation components. In 1979 Bibb Latane, Kipling Williams, and Stephen Harkins isolated the motivational half experimentally and named it social loafing. The software industry supplied the other half of the vocabulary. Fred Brooks, drawing on managing the IBM System/360 operating system project, published The Mythical Man-Month in 1975, where the n(n-1)/2 communication-path count underpins his observation that adding people to a late project makes it later. J. Richard Hackman spent decades at Harvard arguing that teams are routinely oversized and that stable membership and clear boundaries matter more than raw talent. Practice followed: Amazon popularised two-pizza teams under Jeff Bezos around the turn of the 2000s, and agile frameworks codified small units, with Scrum long recommending development teams of roughly three to nine people. Anita Woolley and colleagues added a measurement angle in 2010 with evidence in Science for a general collective intelligence factor in groups.