Don’t make a fetish of having a flat organization
At a minimum, you need a healthy bureaucracy to keep track of money, therefore designing a great bureaucracy is essential to your company's growth
(The following is an excerpt from a book I’m writing for entrepreneurs and managers at small to medium sized startups.)
Over the last 15 or 20 years, one vice that’s become popular among Silicon Valley types is a fetish for “flat organizations.” Supposedly these will save us from the horrors of bureaucracy.
For the sake of humor, let’s take a quick side trip down history’s lane, in particular regarding the opposite paths blazed by two words: divan and bureau. For centuries, the divan were the ministers who served the shah in the Persian empire. When declaring a judgement, or making a new ruling, they normally sat upon a traditional structure of wood and cloth. Western merchants and diplomats went to Persia and studied the empire’s system of government, eventually concluding, “I really like that couch.” They brought the designs home. Nowadays you can go to any furniture store and buy one. Meanwhile, in the West, a bureau was a kind of cabinet where you could store things. Towards the end of the 1800s, as government got bigger, Max Weber saw that leaders had more and more secretaries who were writing things down, such as legal agreements — then storing the documents away, and later finding those documents and citing them for various important reasons. Weber decided this was a new kind of government, and he invented the word “bureaucracy” to describe it. So “divan” was a system of government that became a piece of furniture whereas “bureau” was a piece of furniture that became a system of government.
Because bureaucracy seemed impersonal and objective, Weber also predicted that it would be the most rational form of government ever invented. There is a sense in which he was correct — a bureaucracy tends to do what you tell it to do, not what you want it to do. Anyone who has ever tried to write software is aware how maddeningly literal computers can be: they forever do what you tell them to do, not what you want them to do. If you tell software to do two things which contradict, you’ll end up with a bug that causes all kinds of unpredictable errors, and since most software is somewhat buggy, it’s apparently easy to tell a computer to do contradictory things. Bureaucracy, as it turns out, is as rational as any computer. Given contradictory instructions, it will produce bizarre errors. And when a bureaucracy is many decades old and has had multiple leaders with diverse objectives, it ends up being as difficult to predict as old software that’s become fragile because too many people have given it too many contradictory instructions. But just like good software that is carefully written and well maintained, a well-designed bureaucracy can be a powerful tool which can take your organization to a level of success that nothing else will achieve. Any large organization that is successful will be process oriented, and those processes will be overseen by a well-run bureaucracy.
By contrast, the divan ruled over Persia during a simpler time, when the whole vast empire had a smaller population than New York City does today. As such, legal conflicts were fewer, and the divan could take their time and consider each conflict carefully. This didn’t lead to better government, but it was a more personal system. They didn’t necessarily reason from abstract principles of justice; rather, they could consider what would be fair for each person who came before them, given the context of who that person was.
Successful startups must initially be run by divan, and then later, as they grow, they must be run by bureaucracy. But a startup is always run by something. No startup is ever truly flat, and insisting they can be flat tends to be a smoke screen that certain irresponsible leaders use to disguise how much they are trying to abdicate their obligations.
How much flatness is good? Answer this pragmatically, not ideologically
Among the entrepreneurs I’ve known who have been militant on the issue of insisting that their organization remain flat, most have had personal reasons for their rigidity. One of them grew up in a household where, sadly, a parent had wrestled with alcoholism and was prone to episodes of rage. Another such entrepreneur grew up in a happy household, but the parents were what I think could be fairly described as counter-cultural hippies, and they homeschooled all their children. Having grown up in a household that treated authority as a political question, it was natural that he should take that attitude into his business life.
This is a tricky issue, so I want to give recognition to all sides. When we, as citizens, think about our nation, obviously questions such as hierarchy and authority are legitimate political questions, and over the course of decades we can decide if we want to live in a more hierarchical society or a less hierarchical society. We can also recognize that society is composed of many different hierarchies, and we might tolerate an increase of hierarchy in one area, such as the specialization of labor, while demanding a decrease in hierarchy in some other area, such as corporate power.
But in the short term, when dealing with your own organization, I urge you to think of this as a purely pragmatic question. “What works best for us right now?” is the only productive lens through which a business leader can approach this topic. No other issue is as subtle or as context-dependent. You might have some great mid-level managers who can take on the burden of managing a large number of direct reports, and therefore you can keep your structure relatively flat. Or you might have multiple products, each needing their own development team, sales team, and marketing team, and therefore its own division with its own vice-president, so you need a relatively high level of hierarchy. The right test is always, “Can we increase sales if each leader delegates more authority to someone below? If so, how much can we delegate before we have to create a whole new layer of management?”
A blind commitment to flatness damages your organization
There are multiple ways that a rigid commitment to flatness damages your organization:
1. Each top leader (because even “flat organizations” do typically still have a CEO, a CTO, a CFO, and perhaps even a CMO) will have so many direct reports that as a practical matter they are no longer really leaders. They become symbolic figureheads who no longer know what people are doing. They might boast that they’ve avoided the vice of micromanaging, which is true, but they’ve gone to the other extreme and largely shirked their responsibilities.
2. It’s unfair to the most experienced people on the teams. A good friend of mine was, for all practical purposes, the head of devops at a very large analytics firm, and he had to architect an ambitious system for consuming and processing a terabyte of data a day. But he can’t put “Head of devops” on his resume because such titles didn’t exist at his company. The organization's strict commitment to flatness robbed him of the recognition he deserved.
3. Ironically, one still ends up with the worst kind of micromanagement. Regarding the friend I just mentioned, in theory everyone on the devops team was equal, no matter how much experience they had — so when there was a dispute, my friend couldn’t simply resolve the issue. He wasn’t the team leader because there were no team leaders. Everything had to be escalated to the CTO. This is the worst kind of micromanagement. The CTO will tend to be poorly informed on most issues, as there are too many issues for any one person to keep track of. If there had been a “frontend team leader” and “database team leader” then most disputes could be resolved at a lower level, but creating such mid-level managers would create hierarchy. Without them, every heated argument percolates up to the top.
Do you want every decision to go to the very top?
Some of the advocates of flat organizations have said to me, “So long as everyone acts like a real adult and talks things over with real maturity, then the teams can resolve their disputes on their own, without having to escalate issues.” I suspect that such advocates are subconsciously thinking of family dynamics: if children could act like real adults, then they wouldn’t have to escalate their disputes to the parents. But that assumes the fight is over something trivial, like toys or clothes or video games. Among professionals, the issues will often be over strategic initiatives, where two options are both valid, but each commits the company to a different path. A question like, “Should the database team use a distributed, high write-volume NoSQL database, or a traditional database with a RabbitMQ queue in front of it it?” is not like two children arguing over a toy. Strong arguments can be made for either option. But it is a major architectural decision and it will shape how software is developed at the company for many years to come. “Act like an adult” also means “defend your professional opinion based on all you’ve learned over your whole career,” which doesn’t facilitate compromise on important issues. Just the opposite, you’ve been hired as a highly skilled professional so you can contribute a highly educated opinion and defend it. But somewhere in the system there has to be a leader who can make a choice and then accept all the consequences and responsibilities from having made that choice. So the only real question is, will your company empower mid-level managers to make those decisions, or will every decision get escalated all the way to the top?
A point of contrast might be Eric Garside at Freshly. Recall that he had 15 direct reports when the team had 60 people; that’s 15 people he was trusting to make most day-to-day decisions so he didn’t have to think about them. Later, when he needed even more distance from the inner workings of the tech team, he elevated his 2 most trusted lieutenants and left nearly all the work to them. By all accounts, Garside managed this transition in a pragmatic fashion, without any rigidness about keeping the organization flat nor any particular desire to make it more hierarchical. He simply did what he deemed best for the company.
You cannot build a vibrant, successful organization if you try to follow simplistic slogans such as, “All bureaucracy is bad.” Nor is it valid to say, “All bureaucracy is good.” The only appropriate approach is to figure out when your current managers are at their maximum and then to introduce new layers of management to facilitate future growth.
Actually, depending on how fast you are growing, you might need to hire managers in excess of what you currently need. Peter Drucker, when he was working as a consultant, would ask CEOs, “How would your organization be able to react if a big sale came in that doubled your sales?” If the CEO responded, “Well, of course we’d be happy for the new business but, wow, that would be a burden to try to meet that demand,” then Drucker would answer, “You’re too lean. You need to start hiring and training more managers right now.”
Put another way: if running a “flat organization” means that all of your top leaders have to work 60 hours a week to keep up with their many direct reports, then your organization is brittle and cannot react quickly to changing circumstances, because your leadership is already spread thin and exhausted. Likewise, if your top leadership only works 40 hours a week but is unable to talk with most of their direct reports, then your organization is, very simply, poorly organized and poorly led.
But won’t bureaucracy ruin our company?
Just to be very clear, there is something joyful about a small company. It is cozy. It is personal. It can be supportive in a way that is difficult for a larger organization. If you are the owner, you can decide that you want to keep your company small forever. That can be fun. In some ways it’s less stressful. If you don’t have investors, then this is entirely your choice. The rest of this chapter exists only for those entrepreneurs who aim to build the biggest organization they can. At some point they will need to build a process-driven organization, and they will need a bureaucracy to help enforce those processes.
Don’t ever think that a badly functioning bureaucracy is the natural state of bureaucracies. That is a fatal assumption.
As customers and as citizens, all of us have a few bad experiences with bureaucracies run amok. But ask yourself, is the system really broken from the perspective of those who are in charge of the system? If you call your cable company and tell them you want to cancel your account, but a month later they still bill you for the service, has the system malfunctioned for those who are running it? After all, they are happy to take your money for another month. In government, we sometimes witness a rogue agency that gains too much autonomy and starts granting itself a larger budget, feeding its own growth like a cancer. A government bureaucracy will only be accountable to the public will if the public elects someone strong enough, in office long enough, that the elected leader can gain control over the bureaucracy they nominally lead. Otherwise, no one is in charge. When an anti-government populist fury removes too many politicians, the government bureaucracies gain the independence to do whatever they want, free of oversight. We’ve recently seen that with some police forces, which sometimes have more power than the civilians who are elected to oversee them. Paradoxically, citizens would get more responsive government if they elected politicians for longer terms, as that would give the politicians the time to gain control over the systems they govern. (If this subject interests you, I strongly recommend the book Democracy for Realists by Christopher H. Achen and Larry M. Bartels.)
Much like software, a bureaucracy will usually do what you tell it to do, not what you want it to do. Just like software, ensuring that what you said and what you want are concordant takes constant effort. But it can be done. With enough effort, you can write software that does exactly what you want, and likewise, you can build a bureaucracy that does exactly what you want. And that is what you should aim for: a well-run, highly functional bureaucracy that empowers your company for further growth.
Take, for instance, money. A small team of highly ethical individuals can generally trust one another to safeguard money in an appropriate way. Many small churches and family businesses are run this way, with loose controls over who can deposit or spend. However, as a company grows, the chance you'll hire an unethical individual also grows. At a certain scale it is almost certain that a worker will steal some money, if they have the chance to do so. So a company must impose rules about handling money, and there needs to be a process for paying bills, with someone authorized to write checks and use the company credit cards. This is the beginning of your financial bureaucracy, and the tight limit on who holds the authority to spend is the beginning of hierarchy in this area.
Is this a bad thing? Again, you'll confuse yourself if you think about this in terms of good or bad. Just focus on your goals. What strengths do you want to build into your organization?
Nowadays, there is a widespread idea that bureaucracy makes an organization rigid. But consider the ways that a well-designed bureaucracy for money control can make your organization more flexible. If you decide to keep your money control bureaucracy small then you'll also need to keep your purchasing process simple, perhaps by centralizing all purchasing decisions in the hands of the CEO or CFO. That one individual then becomes a choke point that limits progress – nothing can get done until they find time to pay some bills and buy some necessary materials or services. And there is an obvious limit on how much the company can grow when the authority to pay a bill is so limited. But even when the money handling bureaucracy grows to three or four people, that’s still not enough to oversee complex spending patterns, so your company will still be forced to follow simple and inflexible rules. By contrast, what if you wanted to decentralize the power to make purchases? What if you wanted to empower each team with the ability to make whatever purchases they felt they needed, up to some reasonable limit? That would make your organization much more flexible, but it also makes it more difficult to track the money, and therefore it requires a larger bureaucracy for tracking each dollar. My point is, the larger bureaucracy can actually help enable some kinds of decentralized decision making, and thus improve the agility of the organization. A growing bureaucracy does not automatically equal growing inflexibility, rather, the whole point is to try to maintain flexibility at scale.