The School of Knowledge is the weekly newsletter for SME owners and investors who want frameworks they can actually use — frameworks, checklists, and operating manuals every weekend, built to read on Sunday and use on Monday.
This is the second instalment of my How to Design the Systems That Run Your Business series. A series dedicated to helping business owners and operators run their businesses more efficiently, including how I run mine. To read the first article on designing and monitoring feedback loops please click here.
Every system has a boundary, and the first act of the system builder is to draw it. This sounds like an analytical exercise—study the problem, identify the relevant variables, include them—but it’s not. Boundary decisions are strategic choices, often the most consequential ones an organisation makes, because they determine what you can control, what you can accumulate, and what you must take as given.
Draw the boundary too narrowly, and you will be blindsided by forces you assumed were someone else’s problem. Draw it too broadly, and you will drown in coordination costs. The history of corporate strategy is littered with examples of both errors, and the theoretical and practical tools for thinking about boundaries—from Coase and Williamson to Ashby—remain among the most powerful and underused frameworks in organisational design.
Why firms exist: Coase, Williamson, and the logic of boundaries
Negotiating terms, writing contracts, monitoring performance, and finding the right counterparty all have a price attached to them—directly or indirectly. Ronald Coase’s 1937 paper “The Nature of the Firm” asked a deceptively simple question to his economic counterparts: if markets are efficient, why do firms exist at all? His answer was transaction costs. Using the market isn’t free. When costs are high enough, it makes sense to bring them in-house, where management can direct their resources through authority rather than by price. Hence the need for companies.
This presents a boundary question: how big should a company be? When the cost of organising one more transaction internally equals or exceeds going to the market, then going to the market is the most efficient thing to do. The company has expanded past its boundary and it is no longer cost-efficient to keep it internal.
The implications are significant:
Firms are not just production functions—they are governance structures that exist because markets are imperfect
The boundary of the firm is an economic decision, not an organisational one
Vertical integration, outsourcing, and make-vs-buy decisions all flow from this logic
The business operator should constantly be asking: where is my boundary? and what makes sense to keep in-house versus outsource? This applies not only to finances, but to all resources: time, energy, and capabilities. At the micro level, the operator should be checking that what sits within their boundary is making them more resourceful—not less.
There are no hard and fast rules when it comes to boundaries. They contract and expand as markets change, meaning you must adapt to them as well.
Oliver Williamson, joint 2009 Nobel Prize winner in Economics, operationalised Coase’s insight through the concept of asset specificity—the degree to which an investment loses value when redeployed to a different use or relationship. Williamson identified four types: site specificity (an oil refinery cannot be moved), physical asset specificity (specialised tooling designed for a single product), human asset specificity (knowledge and skills tied to a particular relationship), and dedicated asset specificity (general-purpose investments made for a specific buyer that would create excess capacity if the contract ended).
When asset specificity is low, market transactions work well. Buyers and sellers are interchangeable, contracts are straightforward, and competition disciplines behaviour. When asset specificity is high, bilateral dependencies emerge. A factory line built to a single customer’s requirements, a skilled worker trained in proprietary systems, or a contractor who has built up years of accumulated client knowledge cannot easily walk away. This creates hold-up risk: the other party can renegotiate better terms after you’ve committed, knowing you can’t easily leave. You have that sunk cost feeling.
Williamson states that when asset specificity is high, transactions should be internalised, where managerial authority replaces contractual negotiation. The company’s boundary, then, is a governance decision: it sits where the risk of opportunism in the market exceeds the cost of bureaucratic coordination within the hierarchy.
Boundary decisions done well
Apple’s approach to integration is a masterclass in selective boundary-drawing.
Apple designs its own silicon, operating systems, and retail experience, but outsources final assembly to Foxconn and chip manufacturing to TSMC. Why don’t they just do everything? The reason they don’t is a masterclass in selective boundary-drawing: Apple integrates where tight coupling between hardware and software creates differentiation that cannot be achieved through arms-length contracts, and outsources where the bottleneck is commodity-scale manufacturing.
Apple’s decision to design its own processors is the Transaction Cost Economies framework in action. In 2008, Apple acquired PA Semi, a small processor design house, for $278 million. In 2010, Apple launched the iPad with the Apple-designed A4 chip—the first Apple system-on-chip. By 2020, Apple had introduced the M1 chip, which ended their reliance on Intel for Mac processors. Steve Jobs captured the strategic logic by quoting Alan Kay: “People who are really serious about software should make their own hardware.” It’s not that Apple could make better chips than Qualcomm or Intel in some abstract sense—but that a chip designed for a single entity can optimise for things a general-purpose chip cannot: power consumption, security architecture, and hardware-software co-design.
Amazon’s boundary was drawn not from selective integration—but sequential expansion.
In his 2023 shareholder letter, Andy Jassy described building something called “primitives“—core capabilities like warehousing, picking, packing, and shipping that can be turned into new products. Internal logistic primitives turned into Amazon Prime, an online store selling books turned into an everything store, and internal infrastructure needs turned into AWS. This boundary strategy wasn’t selective, but sequential. Each internal need was externalised into a product that fed into the Amazon flywheel.
The pattern at Amazon is consistent: solve an internal problem, build the infrastructure, then externalise it as a platform—expanding the firm’s boundary and feeding the flywheel simultaneously.
It’s easy to look for companies that have demonstrated a masterclass in not only designing their boundaries but executing them. Let’s now look at two that got it terribly wrong.
Boundary decisions gone wrong
Boeing’s “efficiency” boundary design ends up being a plane crash.
The Boeing 787 Dreamliner programme is an example of catastrophic boundary failure. Boeing outsourced up to 70% of the aircraft’s design, engineering, and manufacturing to more than fifty tier-one suppliers across the globe. The budget ran to an estimated $40 billion, dwarfing their original estimate of $5 billion. As well as being 40 months late, Boeing’s final assembly line was held together with temporary bolts purchased from Home Depot after supplier Alcoa could not ramp up fastener production fast enough.
How could a company like Boeing—known for engineering excellence—get it so wrong? UCLA Anderson Professor Christopher Tang had an answer:
“You only know what’s going on with your tier 1 supplier. You have no visibility, no coordination, no real understanding of how all the pieces fit together.”
I would say it’s pretty critical that you have a deep understanding of how the parts of an aircraft should fit together.
Boeing’s mistake was expecting to receive complete, flight-ready assemblies, but instead receiving sub-assemblies requiring extensive rework. They had fallen into a classic systems trap: information delay. They had modelled their system on Toyota’s supplier-development approach, but failed to build the relational infrastructure that makes Toyota’s system work. Toyota shares information, conducts joint improvement activities, and treats key suppliers as extensions of itself—relationships built over decades of investment in human asset specificity. Boeing handed off complete design control to suppliers based on price, not relationships. These are exactly the high-specificity transactions that Williamson predicted would fail under market governance. In his Nobel lecture, Williamson used the outsourcing of Boeing’s highly specialised fuselage to Vought Aircraft Industries as a perfect example of something requiring significant investment in specific assets that should have been kept in-house.
Boeing’s 787 Dreamliner is a case study in misaligned organisational boundaries—a prime example of how prioritising cost structure over quality control can massively backfire.
How Kodak had the future of photography in its hands and let it slip.
In 1975, a twenty-four-year-old Kodak engineer named Steven Sasson built the world’s first portable digital camera—a 3.6-kilogram device capturing 100×100 pixel black-and-white images onto a cassette tape. Management’s reaction, as Sasson later told the New York Times, was: “That’s cute — but don’t tell anyone about it.” It’s not that Kodak was ignorant of the new technology; they filed a patent in 1977 and earned billions of dollars in licensing revenue from it. The problem was the fence line Kodak had drawn around the company—their boundary. Kodak’s business was film, paper, chemicals, and processing. Digital photography didn’t fit their model or their identity. They treated it as exogenous—something to be curious about, rather than an endogenous capability to be developed from within.
When Kodak’s leadership did diversify, they moved into adjacent markets by acquiring Sterling Drug for $5.1 billion in 1988. Their logic seemed sound: “we know chemistry, drugs involve chemistry, therefore this is adjacent.”
The problem wasn’t their logic—it was that they weren’t following where their customers were going: digital. The Sterling Drug investment was sold off in pieces by 1994, and in 2012 Kodak filed for bankruptcy, thirty-seven years after one of their own engineers had built the future.
Kodak restricted the expansion of their boundary by actively deciding not to meaningfully pursue digital photography. This illustrates how inside-view thinking (we’re good at chemistry) can crowd out outside-view thinking (the market is moving somewhere we aren’t).
These four examples—two good, two bad—lead us into the theoretical framework that explains why Apple and Amazon succeeded, and why Boeing and Kodak didn’t. Variety.
Ashby’s Law and the question of variety
W. Ross Ashby’s Law of Requisite Variety, published in An Introduction to Cybernetics in 1956, provides the theoretical foundation for understanding why boundary decisions matter so much. The law states, in Ashby’s own words, that “only variety can destroy variety”—meaning that for a system to remain stable, its control mechanism must be capable of generating at least as many states as the disturbances it faces. A thermostat with only on/off settings cannot regulate temperature precisely; a manager with a single response to every problem cannot govern a complex organisation.
The implications for boundary-setting are direct. If you draw your system boundary too narrowly, your system will lack the requisite variety to handle disturbances from the environment. If you draw it too broadly, you waste resources managing internal complexity that does not contribute to matching external variety.
Organisations must be able to adapt to variety. Kodak’s boundary was drawn too narrowly by not assigning resources to digital—which was not only a clear market trend but a signpost from the future that chemical would be obsoleted by digital. The conglomerate’s boundary is often drawn too broadly, its variety impinged by size and coordination costs that offer little flexibility.
So, for you—the business operator or systems designer—how do you move from understanding Ashby’s Law to putting it into practice?
Stafford Beer’s Viable System Model (VSM) is essentially Ashby’s Law operationalised into an organisational architecture. Ashby says a viable system must match the variety of its environment. Beer asked a follow-up question: how do you actually design an organisation to do that?
The Viable Systems Model Framework
The Viable System Model (VSM) answers this by splitting the variety-handling problem across five subsystems:
System 1: the operational units absorbing frontline variety
System 2: the coordination needed between units to prevent confliction
System 3: internal control and resource allocation (the “inside and now”)
System 4: external intelligence gathering (the “outside and future”)
System 5: identity and policy, which sets the purpose and balance of the organisation
Through this framework, Beer recognises that you can’t centralise all variety-handling at the top. The environment moves too fast and is too complex, with informational delays inevitably distorting what’s happening in real time. A viable system should absorb local variety. System 1 units handle the complexity of what’s happening on the ground. System 4 is focused on information gathering—not just to plan for the future, but to protect it. Each system offers a buffer.
Learning about the VSM, you begin to see why Warren Buffett and Charlie Munger preferred to run Berkshire Hathaway the way they did. Their decentralised model continued to allow exceptional management to run and absorb the complexity of day-to-day operations—something both Warren and Charlie would gleefully admit was outside their Circle of Competence. Warren and Charlie’s role was clearly Systems 4 and 5: looking for external opportunities by reading the market and focusing on increasing shareholder value.
If you've found value in this essay so far, what follows is where the theory meets the ground. I walk through how I've applied Beer's five systems to my own construction business and the tools I've invested in, what they cost, what they saved, and the questions I'd now ask of any business before drawing its boundary. If that's useful to you, consider upgrading to TSOK+.
How I used the VSM framework to build my business’s systems
