Quartz had been built flat. It had few levels of management, and decisions were made by talking them through. For four years that worked.
Then, within a single year, its staff rose above 150, and things began to jam. “Too often,” the cofounder Kevin J. Delaney wrote in 2016, “our staff was stalled because it wasn’t clear who was responsible for moving a decision forward.” Staff said the same in anonymous surveys. Some of these worries had been heard before. But, Delaney wrote, “they really came to a head as the company reached 150 people.”
That number was not a coincidence, he thought. It had a name.
A graph and a library
In 1991 the British evolutionary psychologist Robin Dunbar was looking at a graph. It plotted the size of primates’ brains against the size of the groups they lived in. The bigger the brain, the bigger the group. Dunbar wondered what group size the same line predicted for humans.
“The number my calculations gave was 150,” he later wrote. “Since this seemed low, I hurried off to the library to look for data on natural human group sizes.” Among hunter-gatherers, he found, the community level of organization came out at almost exactly 150.
His idea is that the brain can handle only so much social information, and that this caps how many relationships a person can keep up. The number, he stresses, is about meaningful relationships, not everyone you could name. He describes a person’s network as layers, each roughly three times the size of the one inside it: about five intimate friends at the core, around 150 meaningful relationships further out, and as many as 1,500 people whose names you know.
Later studies found the number turning up again. In 2003 Dunbar and a colleague studied who people sent Christmas cards to. The full networks averaged 153.5 people.
Split before it grows
Dunbar points to groups that seem to have found the limit on their own. The Hutterites, he writes, split their communities once they pass about 150, because beyond that peer pressure alone can’t run them; W. L. Gore, as he describes it, built a new factory next door rather than let one grow past 150. “Keeping things below 150 means you can manage the system by peer pressure,” he told Delaney, “whereas above 150 you need some kind of top down, discipline-based management system.”
A number under fire
The number is famous. Whether it is real is another matter.
In 2021 three researchers at Stockholm University, Patrik Lindenfors, Andreas Wartel and Johan Lind, redid Dunbar’s calculation with more data and newer statistical methods. Their estimates came out all over the place, with 95% confidence intervals as wide as 4 to 520. “Specifying any one number is futile,” they wrote.
Dunbar replied that they had used the wrong kind of regression. With the method he used, he wrote, their own data predicted 158. The Stockholm group answered that his preferred method, run on their data, gave not 150 but about 290, and over 400 if only monkeys and apes were counted. They also pointed out that a figure worked out as an average group size is often quoted as a maximum.
So 150 is best treated as a rough guide, not a law. The underlying observation is harder to argue with: relationships take time and attention, and a group that relies on everyone knowing everyone will, at some size, run out of both.
What changes past the line
Imagine a design agency that grows from 60 people to 170 in two years. Nobody can point to the day it broke. But the office manager no longer knows every name. Two teams build almost the same template without hearing about each other. A designer finishes a pitch and doesn’t know whether the account director or the studio head signs it off, so it sits for a week. A new hire asks who approves a client discount and gets three different answers.
None of this means the agency hired worse people. The informal system that worked at 60 (everyone overhears everything, and the founders settle disputes over lunch) has quietly stopped reaching everyone. Each of those problems is a question that used to answer itself: who’s working on what, who decides, who needs to know.
The fix is rarely to stop growing. It is to answer those questions on purpose, starting with the one that stalled Quartz. Its first change was to make fewer people clearly responsible for each part of the business, so that a stuck decision had an obvious owner. Spotify, in 2012, tried something similar from the other end, keeping each of its “tribes” small enough for people to know one another, under a limit it put at “100 people or so.”
Whether the line falls at 100, 150 or 290 matters less than noticing when you have crossed it. A good sign is when a question that used to have one obvious answer now has three.