Span of Control

How many people can one manager actually manage? The question sounds administrative. It is architectural. The answer determines the shape of the entire organisation — how many layers sit between the CEO and the front line, how fast information travels, how much distortion accumulates in transit, and whether decisions are made by people with context or people with titles.

Classical management theory — Lyndall Urwick in 1956, drawing on military command structures — settled on five to seven direct reports as the effective maximum. The reasoning was cognitive: a manager tracking five people can hold each person's context, priorities, blockers, and development needs in working memory. At eight, the tracking degrades. At twelve, the manager is scheduling one-on-ones they can't prepare for and reading status updates they can't act on. The manager becomes a router — passing information between people who could communicate directly if the hierarchy didn't exist.

Jeff Bezos rejected this framing entirely. His "two-pizza team" rule — no team should be larger than two pizzas can feed, roughly six to ten people — was not about span of control in the classical sense. It was about eliminating coordination costs at the team level so the manager's job shifted from information routing to decision-making. Amazon's insight: the problem isn't that managers can't handle more reports. The problem is that large teams generate exponential communication overhead. A team of six has fifteen communication paths. A team of twelve has sixty-six. A team of twenty has one hundred and ninety. The manager becomes the bottleneck not because they're incompetent but because the mathematics of group communication are unforgiving.

Google ran the opposite experiment. Under Larry Page's return as CEO in 2011, some engineering VPs managed twenty or more direct reports. The logic: wider spans force managers to delegate deeply, prevent micromanagement, and push decision-making authority down to the people with the most context. Google's Project Oxygen research found that the best managers coached, communicated vision, and removed blockers — functions that don't require narrow spans. The implicit assumption: if you hire people who don't need close supervision, you can widen the span dramatically.

Both approaches work. Neither is universally correct. The variable that determines which is right is the nature of the work. Routine, well-defined tasks tolerate wide spans. The manager of a twenty-person customer support team with documented processes and clear escalation paths doesn't need narrow oversight — the process itself provides the structure. Novel, ambiguous, high-judgment work requires narrow spans. The VP of product managing three product leads who are each making bet-the-company decisions about market direction needs to be deeply embedded in each person's context. The span must match the cognitive load of supervision, not an arbitrary number from a management textbook.

Every additional management layer between the top and the bottom adds latency and distortion. A directive from the CEO passes through four layers of management and arrives at the front line as a different directive — not because anyone intentionally changed it, but because each layer filters, interprets, and rephrases based on their own context and incentives. Bezos's observation cuts to the core: "Communication is a sign of dysfunction. It means people aren't working together in a close, organic way." The ideal organisation doesn't need to communicate across layers because the layers don't exist — or because the spans are designed so that each layer adds decision-making value rather than translation overhead.