The Role of a CEO - Part II - Limit Theory

Operating in the Limit

In the previous piece, I argued that a CEO is responsible for converting the firm's specialized capabilities into compounding returns on the capital entrusted to the firm. That is the rule. The question this piece is meant to answer is how, in practice, that rule actually gets executed day to day and year to year.

The shortest honest answer I can give is that a CEO has to operate in the limit. By that I mean something specific, borrowed from calculus, and I want to spend a few pages explaining what I mean and why I have come to think it is the only useful frame for the work.

What a Limit Is, and Why It Matters Here

In calculus, a limit is the value that a function approaches as its input is pushed toward some extreme. The function may never quite reach the limit, but its behavior — the closer you get to that extreme — is increasingly dominated by what happens at the edge rather than what happens in the middle. The shape of a curve at its asymptotes tells you more about its long-run behavior than any point along its middle does.

Business has the same property. The durable advantages, the meaningful margins, and the companies that survive long enough to compound all live at the extremes of one distribution or another. The middle of any serious distribution in business is, over time, the place where companies get quietly eliminated. Once that observation sinks in, a great deal of what passes for corporate strategy starts to look like wishful thinking dressed up in slide decks.

Warren Buffett's language about moats is one form of this argument. A moat is, in plain terms, a structural feature that keeps a company at the profitable end of an industry's distribution and prevents it from being dragged back toward the middle. Companies with deep moats earn returns above the cost of capital for long stretches of time. Companies in the middle of their industry earn the cost of capital, roughly, and only for as long as competitive pressure permits.

The CEO's job, then, in operational terms, is to figure out which limits their company needs to operate at, and to push the company toward them with consistency.

Limits at the Level of the Company

The clearest example is strategic position. A company can earn outsized returns by being a high-price, high-margin producer with a defensible reason to command a premium. It can also earn outsized returns by being the low-cost producer at meaningful scale. Both models have produced some of the most durable companies of the last hundred years. What does not work, over any real period of time, is sitting in between. A company stuck in the middle has neither the pricing power of the premium end nor the cost structure of the volume end, and it tends to be ground down between them.

Cost structure is its own version of the same idea. Two companies producing the same product at different fundamental cost positions are not really in the same business. The lower-cost company has more capital available to reinvest, more room to absorb a bad year, more flexibility to cut price when it needs to, and more cushion to fund the next investment cycle. Over enough years, that gap stops being a competitive nuance and becomes a structural fact. Lower operating cost compounds into faster growth, which compounds into more scale, which compounds into still lower unit cost. The math of compounding does not reward proximity to the cost frontier. It rewards being on it.

Talent has the same shape. A genuinely great hire at almost any level of the company produces output that a stack of average hires cannot match. Distributions of talent in most functions are heavily skewed rather than normal. A CEO who staffs the bench with median talent is, knowingly or not, choosing the middle of the distribution and accepting the returns that come with it. The cost of holding out for someone excellent is real, but it is almost always smaller than the long-run cost of filling the seat with someone passable.

Capital allocation behaves the same way. A handful of decisions, made over the life of a company, will drive most of its long-run outcome. Whether to acquire. Whether to reinvest. Whether to return cash. Whether to enter a new market. Whether to exit an old one. A CEO who treats those decisions with the same level of attention as routine operating choices is misweighting their time. The outcomes are not evenly distributed across decisions, and the work should not be either.

Limits at the Level of the Person

The same logic that governs the company governs the person running it. Every dimension of a CEO's own conduct, examined honestly, behaves the same way the company's economics do. Returns to operating in the middle are modest. Returns to operating at the limit, sustained over years, are not.

Work ethic is the most misunderstood of these. The right metric is not hours logged. The right metric is effective output per hour of focused effort, sustained across years without breaking the human producing it. A CEO who runs themselves into exhaustion to maintain a long hour count is degrading the very faculty — judgment under uncertainty — that the job most depends on. The discipline is to do the minimum number of hours required to achieve the result, and to make sure those hours are spent on the highest-leverage problem available. Everything outside that — sleep, family, exercise, time to think — is part of the system that produces the next decade of decisions.

Trustworthiness operates at the limit in a particularly powerful way. A CEO who is reliably honest with employees, customers, lenders, partners, and shareholders gets information that less trustworthy CEOs never see. They get capital on better terms. They get the benefit of the doubt in moments when it matters. Trust compounds slowly, but it compounds, and it does so along an asymptote that someone who is occasionally dishonest can never reach.

The same applies to leading with bad news. A CEO who consistently delivers bad news first, plainly and before it is asked for, creates an information environment in which good decisions can actually be made. A CEO who hides bad news, or softens it, or delays it, degrades that environment in ways that are hard to see in the short run and decisive in the long run. The information system is the most important asset most companies have, and the CEO sets its standard.

Humility belongs in the same list because it determines learning rate. A CEO who is genuinely open to being wrong updates faster than one who is not. Updating faster, over enough years, is the difference between a CEO whose model of the world tracks reality and one whose model has quietly drifted out of alignment with it. Pride is a tax on learning, and the CEOs I respect most pay the least of it.

Persistence is the limit no one writes about, because it sounds like a cliché until you watch it operate. Most real advantages — in product, in cost, in distribution, in brand — take years to compound. The curves bend slowly, and they bend most steeply on the far end. A CEO who quits in year three on a project that needed seven is, mathematically, not operating in the limit. They are operating in the early part of a curve they never let mature.

Each of these dimensions, taken on its own, sounds obvious. The point is not that any of them is novel. The point is that they all share the same shape. Returns to character, in this job, are non-linear. Being slightly more honest, slightly more humble, slightly more selective in hiring, or slightly more efficient in time use is not slightly better. Across enough years, it is the difference between a company that compounds and one that does not.

The Logic Underneath All of This

The reason this works is the same reason the previous piece's three economists work. Specialization compounds. Companies are the structures that organize specialization. CEOs are responsible for the returns those structures produce. And every meaningful axis of that responsibility — strategy, cost, talent, allocation, and character — exhibits non-linear returns to operating at the extreme.

A CEO who lives in the middle on any of those axes is, slowly and without drama, ceding ground to whoever does not. The math is not interested in intentions. It is interested in where, on the curve, you are actually spending your time.

In the next piece I will narrow this further into the daily practice that follows from it. Once you accept that the work has to happen at the limit, the question of which limit you are working on, today, becomes the only operating question worth asking.