The Mythical Man-Month
1. More people create more communication
Software work is not a pile of identical tasks that can always be divided among more people. New contributors need context, existing contributors stop to teach them, and every added dependency creates another path along which plans can diverge. The effect is strongest when a project is already late because the remaining work is often the least separable and the deadline leaves no time for onboarding to repay its cost.
The management lesson is not to avoid growth. It is to distinguish work that can be partitioned from work that depends on a shared mental model. Add capacity early, make interfaces crisp, and protect experienced contributors from becoming an unplanned help desk. When the critical path is tightly coupled, reducing scope can outperform increasing headcount.
Imagine a six-person team trying to stabilize a release in its final three weeks. The failures cross the same authentication flow, database migration, and deployment process, so each fix depends on knowledge held by several people. Adding four engineers may look like a 67 percent increase in capacity. In practice, the original team must explain the architecture, reproduce fragile environments, review unfamiliar changes, and reconcile more competing diagnoses. The calendar gains people while the critical path gains interruptions.
Communication cost does not rise only through scheduled meetings. It appears in duplicated investigation, incompatible assumptions, review queues, merge conflicts, changed interfaces, and decisions that have to be retold. If every person must remain aligned with every other person, the number of possible working relationships grows much faster than the headcount. A team can reduce that burden with stable boundaries, written decisions, clear ownership, and work that can be accepted through an interface rather than a chain of conversations.
There are still good reasons to add people. A separate documentation effort, a well-bounded migration, test automation, customer support, or an independent service can often proceed without sitting on the same narrow path. The useful question is not ‘Will another person help?’ but ‘What result can this person own with limited new coordination, and when will their contribution repay the onboarding cost?’ The earlier that question is asked, the more likely staffing can change the outcome.
This principle also exposes a scheduling mistake: treating all unfinished work as interchangeable. A project may report twenty open tasks, but if nineteen depend on one risky integration, nineteen extra people cannot make that integration happen in parallel. Managers need a dependency view, not merely a task count. Identify the truly sequential work, the decisions that only one group can make, and the places where knowledge is concentrated. Those are the constraints that determine whether headcount changes the delivery date.
The idea has limits. Small teams can become overloaded, and chronic understaffing should not be defended with a slogan about communication. The point is timing and structure. Before adding people to late work, write down the critical path, the learning each newcomer needs, who will teach it, and which deliverable can be separated cleanly. If that exercise produces no independent ownership, choose a smaller release, remove a dependency, or move the date. Capacity matters, but coordination determines how much of that capacity reaches the product.