Death by a Thousand Meetings
why modern organisational malaise is a problem of scale
Something important about our time is that the things that made us successful is now groaning under the scale.
The FDA had been exceptionally successful in it's initial days, and now groans under the weight of successive ages of regulations
Local government is usually a brilliant idea, to get decisions made closer to the site or the action, and at scale this becomes almost impossible to govern easily
As GM grew to become the largest automaker in the world by the mid-20th century, it became extremely bureaucratic and slow to adapt to changing consumer preferences. This contributed to its decline and eventual bankruptcy in 2009.
Once the largest retailer in the US, Sears struggled to adapt its massive bureaucracy to shifts in consumer shopping habits and competition from big box stores like Walmart and Target. It filed for bankruptcy in 2018.
IBM.
As the public school districts grew to administer large networks of schools, layers of bureaucracy developed that slowed innovation and made schools less responsive to community needs.
As things grow in size, the problems emerge, and often the canonical ones of modernity!
We see it everywhere. The received wisdom is that larger companies are bureaucratic and frustrating to deal with, while smaller companies are lightning quick and able to accomplish in hours what takes the larger organisations days.
And then somehow in the last six months we have seen the fastest response by incumbents I think I’ve ever seen in my working life. Since the word about generative AI came out and ChatGPT hit product-market fit, it’s been incorporated by pretty much most large companies in a matter of months and rolled out to their customers.
This happened at the same time that people were building enormous businesses with smaller headcounts than most early stage startups. Even OpenAI only had like 200 people when they built GPT-4. Midjourney was making hundreds of millions of dollars with barely a dozen people. It was, as Tolstoy would say, the best of times, the epoch of belief.
And yet, it’s Microsoft who have been leading the way but others including Google are fast following. The best open source model was from Meta, which got broken by a team in UAE for a few weeks, and then the crown is back with Meta. Each week a new release or a new partnership is announced, killing a few dozen startups along with it. This is the Empire Strikes Back of technological disruptions.
But still, the conventional wisdom is conventional for a reason. And it’s true because for a large number of organisations they do become less efficient as they get bigger. Things do slow down. As Alex’s brilliant slide deck shows, the organisational headwinds start to gale.
It talks about the idea of small problems growing into large problems because solving any small coordination problem is time and resource intensive, often on the part of the manager, which is an inherently scarce resource. Even if not scarce time-wise, it’s scarce in terms of your reputation points you’re willing to spend. So things fester and get out of control.
But there’s also a separate and equally large problem that doesn’t get enough airtime. Which is that the work you do in larger organisations is not the same work that you do in smaller organisations even if you’re doing the same things.
We used to advise startup CEOs about this all the time. Every six months you’re running a new business. Getting bigger means different focus. Recruiting becomes an enormous chunk of your day, as does firefighting problems and smoothing things over, in essence solving the very problems Alex identifies.
And suddenly you realise despite what you used to profess, you need an HR team. A legal team. Maybe a PR and comms team. Maybe you need people purely dedicated to internal roles, whose job it is to find out and smooth problems that exist which you can’t predict, and couldn’t have predicted.
You start having dozens of direct reports so you create layers in the middle. This means managing information that trickles through to you and doing “surprise” visits and skip-level 1-on-1s. But that’s also (obviously) inefficient.
The problem, simply, is that any configuration of people you have (a team) have a certain capability and it stretches a little but isn’t infinite, and coordination is a meta-layer that leeches off the existing capabilities no matter what you do.
This could be for very good reasons, but still remains true. Here are a couple examples from real life.
You want to launch a new product but need to ensure the marketing is in-line with other products, the design is the same, the GTM team understands what they need to do, the sales team have their pitch, not to mention the engineers have to build it in a consistent fashion that works with the existing background
You want to launch in a new country but need to ensure the legal and compliance is tip-top and you’re not running afoul of any regulations in the home country, and that you are not accidentally creating perverse incentives for your customers (pricing it 2x cheaper for them there than here)
The multiple managerial ambitions collide, which means that any project has to simultaneously satisfy too many considerations to succeed
Larger companies deal with larger projects with means there’s just much more information that needs to be gathered before launching anything! Imagine Barclays launching a new product vs a startup neo-bank, or a niche ecommerce retailer selling shoes for pets, compared to what Amazon would need to do
This is why skunkworks exist!
Because it’s only by completely stepping away from the existing organisational structure that you’re often able to create anything truly new without getting bogged down.
Because each team only has so much information-processing capacity, even when working flat out, any added information sources that you add on top will make them less efficient.
And not to mention, as Alex had it, as number of teams increases, the coordination cost amongst them increases too!
That’s a double whammy. This is why larger companies have problems!
See how the bright blue that peeks up at the bottom but got squashed down? That’s the ambition of every startup founder being crushed by the reality of running a large company. When they succeed!
Because when larger companies need to have larger number of teams work on a problem, they need to expend a lot more resources coordinating with each other.
And when they have to deal with the sheer complexity that comes from coordinating or dealing with a more complex market (more information sources to deal with and analyse), then their analytical capacity gets overwhelmed.
With this reality in mind, there’s the obvious question. Is there a solution?
Well, one that was actually done by the larger incumbents is to have smaller teams given relative carte blanche to go off and do work. By reducing N (number of teams) and by reducing M (information sources they need to deal with), things get better and faster.
In fact they get much faster than the startups in many cases because the employees are highly capable also, and they have an incredible amount of resources at their disposal! That increases how much information you can process as a team dramatically - it’s the organisational equivalent of what the washing machine did for the 1950s housewife.
Also that’s also the inherent ability for more teams to solve more information efficiently - eg with better systems or better processes. This is mildly incorporated already in the stylised example above, but it’s an area of massive boost! If you’ve seen startups across countries or continents, you know how big a difference this makes. For a few decades the consulting companies made bank by just explaining to companies in Country X how Country Y did things.
This is also the specialisation point, as in why hierarchies actually form, and tells you a lot about why companies have Centers of Excellence which are often neither central nor excellent.
Bezos’ two pizza rule is a smaller version of this, where you’re forcing a solution space that is functionally in a smaller space.
I wrote a while back about the odd levels of efficiency seen by the British East India Company and expanded on this with my friend
. In hindsight I’ve modulated the view somewhat, but a lot of it remains. They succeeded as a highly decentralised organisation, not just in the “who could make decisions” sense, but “who needed to be consulted” sense too.So it remains because it feels like larger organisations have high enough entropy that forcing them to do similar things on different aspects can only have the same outcome. And when they don’t, like putting small teams on high priority things like launching LLMs, they do excellent work. Focus and freedom are the key ingredients, and turns out you can’t beat the complexity tax purely by throwing resources at it.
The cost of pursuing efficiency, funnily enough, is tolerating higher levels of inefficiency. There’s a lesson in that!
isn't this the same argument from "The Mythical Man-Month" by Fred Brooks (1975)? Brooks argues that adding more programmers to a project leads to increased communication overhead and makes it harder to coordinate efforts and thus resulting in a counterproductive effect.
It's helpful to look at how bureaucracy evolved--what problems was it adopted to solve?
I see three. Size. Change. Diversity.
First two points are not controversial. First, size. "Bureaucracy" historically comes from oddly large orgs--military, church, shipbuilding--and was borrowed by other organizations. To get big, you need formal procedures that replace old customary or face-to-face procedures. Certain big organizations began to capture economies of scale and scope--in industry, government and social life--which led to the predominance of large orgs we see today.
Second point is change. The bureaucracy can work to bracket and in some ways mitigate changes in the modern economy. Its size in other words lets it provide insurance against unpredictable shocks. It can push its people and capital to smooth out short-term unpredictability and strategically navigate medium-term unpredictability. From the standpoint of the individual consumer the bureaucracy can provide a regular product or service even in the face of change. From the standpoint of the worker, the bureaucracy can offer a somewhat regular lifepath--with set daily activities, seasonal rituals, career progression, etc.
The final point, diversity, may be more controversial. I saw bureaucracies especially in social life arising to deal with exceptionally diverse groups of people. Cities exploded post 16 C--huge anonymous places filled with migrants. The experience of this diversity was uncomfortable. Bureaucracies helped to smooth this away. Instead of dealing with customary work relations you had written, explicit work relations backed up by formal supervision. Instead of dealing with willy-nilly social life, you had clubs with written rules and fines for breaking those rules. These let people (well, men) of different political, religious and geographic backgrounds come together temporarily.