Flat Structures After 100 People — The Hierarchy Is Already There, It's Just Unwritten

Flat structures past a hundred people are rarely flat. What they usually have is a hierarchy that nobody designed and nobody can see: the few people everyone checks with before acting, whose approval nobody wrote down and whose job description says nothing about it. The honest question for a growing company is therefore not whether to have hierarchy but whether to write it down. A written hierarchy can be questioned, paid for and changed, while an unwritten one simply accumulates.

We meet the flat-structure question most often in companies that did something right early. At thirty people, having no managers meant speed, directness and a founder who knew everyone's work. At a hundred and twenty, the same design feels different from the inside, and the leadership is usually the last to feel it, because the founders are the one group for whom access has not changed.

"Everyone reports to the product"

The founders of a fintech company in Riyadh told us they had deliberately avoided a management layer as they grew past a hundred people. Their phrase was that everyone reported to the product. Teams formed around problems, dissolved when the problems were solved, and the two founders did performance conversations for the whole company twice a year.

The picture from the floor was different. Decisions happened in a handful of messaging groups that new joiners were not in. Two senior engineers had become the people everyone asked before merging anything or starting anything, so they spent their days unblocking others and their evenings doing their own work. They had no title for that role, no pay for it and no authority to say no. Within the same quarter both resigned, and the company discovered how much of its coordination had been running through two people.

One of the founders said it plainly afterwards: "We thought we'd removed hierarchy. We'd just stopped paying for it."

The Four Jobs Test

The debate about flat structures goes in circles because it treats hierarchy as one thing. It is really four jobs that a management layer happens to bundle together. A company can stay flat past a hundred people only if each of the four is being done by something, and it should add managers for whichever jobs are not.

The fintech above had solved one of the four. Teams formed around problems, which is a coordination mechanism of sorts. Deciding had fallen back onto the founders, people care happened twice a year from two people for a hundred and twenty, and the coordination that did happen ran through two unpaid, untitled engineers. That is a flat chart over a very steep informal hierarchy.

Here's the contrarian part

Most writing on this topic sets flat against hierarchical. The distinction that matters is written against unwritten. Every company past roughly a hundred people has hierarchy, meaning some people's views weigh more on some decisions than others. The only choice is whether that weight is visible, deliberate and matched with the time and authority to use it well. Jo Freeman made this point about activist groups in the early 1970s in The Tyranny of Structurelessness: a group that refuses formal structure does not eliminate power, it hands power to the informal network, where it cannot be challenged.

The second point surprises founders. Add the first managers for people care, not for control. Founders tend to introduce managers when approvals become a bottleneck, and so the first managers arrive as approvers, which is exactly the bureaucracy the company was trying to avoid. The job that usually breaks first is people care, because it scales worst: two founders can make decisions for a hundred and twenty people for a surprisingly long time, but they cannot know a hundred and twenty people's ambitions. Start the management layer there, keep spans wide, and hand over decision rights later and deliberately.

Why it works

Several strands of evidence point the same way. Anthropologist Robin Dunbar's work suggests that stable, trust-based relationships top out at around a hundred and fifty people, which is roughly where informal coordination stops working without help. W. L. Gore, famous for its lattice organization with no traditional bosses, has long kept individual sites small, around the size where people can still know each other, which is a structural answer to the same limit rather than an exception to it.

The attempts to prove managers unnecessary are instructive too. Google's Project Oxygen began, by the company's own account, as an effort to show that managers did not matter in an engineering culture, and ended by concluding that good managers made a measurable difference to team performance and retention. The work then turned to identifying what those managers did, and most of it was people care and coordination rather than control. Holacracy-style experiments, such as Zappos's in the mid-2010s, show the other side: replacing managers with explicit roles and rules can work, but only because it writes down far more structure than a conventional company ever does, which is the opposite of structurelessness.

In our own work the pattern is consistent. The companies that stay effectively flat past a hundred people have unusually explicit decision rights and unusually disciplined coordination rituals, the kind we discuss in our piece on the daily huddle. The companies that struggle have the same flat chart and none of that scaffolding, which is why the org chart and the organizational structure can tell completely different stories about the same company.

A practical checklist

• Map the informal hierarchy first. Ask ten people whom they check with before a significant decision. The names that keep appearing are your current management layer.
• Run the Four Jobs Test. For each of deciding, coordinating, people care and accountability, name the mechanism that does it today, or write "nobody."
• Recognise the people already doing the work. If someone is carrying a coordinating or people-care load, give them the title, the time and the pay, or take the load away.
• Start the first managers on people care, with wide spans, before giving them approval authority.
• Write decision rights by area, with thresholds, so that adding managers does not mean adding approval steps.
• Watch new joiners. How long it takes a new person to work out whom to ask is one of the best signals of hidden hierarchy.

Ask yourself

• Who are the three people everyone checks with before acting, and does their role description say so?
• When did each person in our company last have a meaningful conversation about their own growth?
• Which decisions still reach the founders only because nobody else has written authority to make them?
• If our two most-consulted people resigned this quarter, which coordination would stop?
• Are we avoiding managers because we believe in autonomy, or because we don't want to decide who they should be?

The takeaway

Past a hundred people, the choice is not between flat and hierarchical but between a hierarchy you write down and one you pretend not to have. Find out which of the four jobs of management are not being done, add managers for those, starting with people care rather than approval, and keep everything else as flat as the work allows. That is how companies like the one we described in the real challenges of scaling a startup keep their speed without leaving their best people to carry an invisible load.