Each new hire adds conversations as well as hands
Headcount grows in a straight line. The number of conversations it takes to keep people aligned grows much faster, which is why small teams feel fast.
The more people we have on our team, the less room there is in the elevator and the more complicated everything gets.
A team of four has six possible pairs of people who need to stay in sync. A team of twelve has sixty-six. A team of forty has seven hundred and eighty. The arithmetic behind David Karp’s joke about the elevator is old: Fred Brooks used the same counting in The Mythical Man-Month in 1975, to explain why adding programmers to a late software project tends to make it later.
Tumblr, which Karp started in 2007, was known for running a very large service with a very small staff. That kind of team has advantages that are hard to see until they are gone. Everyone knows what everyone else is working on. Decisions happen in a hallway conversation. A new idea can be built and shipped in days, because there is no one to consult who is not already in the room.
Each hire changes that a little. A new engineer needs context, and someone has to provide it. Two people working on related parts of the code now have to agree on an interface. Once a team passes roughly the size of a dinner table, it needs meetings, written plans and managers, and those are real costs even when they are done well. The phrase “the more complicated everything gets” is the honest description of what those costs feel like from inside.
The practical lessons are about sequencing. Hire only when a specific piece of work is clearly suffering for lack of a person. Prefer people who can take on a whole area over people who need close direction. Keep the product small enough that a small team can hold it in their heads.
The case against the line is just as concrete. A service that grows faster than its staff will show the strain somewhere, often in outages, slow support and exhausted engineers. Being understaffed is also complicated; it just produces a different kind of mess. Some work, such as operations, sales and customer service, scales with the number of users whatever the founders prefer.
The goal is not the smallest possible team. It is the smallest team that can do the work well, and a willingness to pay the cost of complexity when growth demands it. A crowded elevator is better than a company that cannot keep its site up.