If you find a company that is going to be a good company long-term, then you should hold on to it, because there is a persistency of these barriers to entry.
— Chris Hohn
Download the Slides Deck: The_Mechanics_of_Cash_Flow
Listen to the deep-dive discussion – The Mechanics of Business Tolls
A law practice and Visa are both tolls. Each collects a fee at a point others must pass through — the lawyer on a dispute that must be resolved, Visa on a payment that must be made. Call them both an A, the cash collected now, and the word is accurate for each. It is also nearly useless, because almost nothing else about the two is alike. One is re-won file by file and dies with the person who holds it; the other arrives on its own, widens on a wave it did not build, and would outlive anyone. If a single word covers both, the word must be taken apart before it can be used. A toll is not one thing. It is a family of things, and knowing which kind you are looking at is the first act of reading a business — prior to any question about its growth, and often decisive on its own.
I. The A and the B
Two durations run through every business. The A is the toll: the cash the company collects now, at a point others must pass through. It is short duration — money in the present. The B is the growth of the base that will make tomorrow’s toll larger than today’s. It is long duration — value in the future. A company is always some arrangement of the two, and compounding is the conversion of one into the other: the toll’s cash poured into a base that will earn more, again and again.
The B comes in two kinds, and the difference runs through everything that follows. A mechanical B is growth that must be bought — funded out of the toll, paid for at market price, each increment of base acquired. The business is an A plus a B: two things added, the second purchased. Bought growth is not in itself poor growth — a mechanical B built behind a strong moat keeps its margins and its return protected, because the moat, not the price paid, is what defends them. It is only growth bought on open ground, outside any moat, whose return is competed away toward its cost. The question is never merely whether growth is purchased, but whether it is purchased inside a moat or beyond one. A free B is the rare other kind: growth carried in from outside by a secular wave, at little or no capital cost, and captured by the moat so that it passes through the toll rather than around it. There the business is an A times a B — the toll and the wave fused, multiplying, growth that asks for almost no capital because the company did not build the wave and does not fund it.
This piece is about the A — the toll itself, before any question of the B. For the toll is not one thing either, and which kind it is decides more than any projection of its growth.
II. The Strength of the Toll
The first axis along which tolls differ is protection: how well the collection point is defended, and for how long. This is the moat, and its role here is narrow and specific — it does not change what a toll is, only how long the toll keeps its strength. A reminder matters before going further: the A is short duration by definition. It is cash now, not the base that earns later. The moat never makes a toll long; it makes a toll’s shortness persist. A well-defended toll collects its near cash for decades; an undefended one collects until the next competitor arrives.
At the weak end sits the toll with no moat at all. The coffee shop on the corner collects a real fee, but anyone may open another across the street, and the margin is competed away. The commodity producer, the undifferentiated manufacturer, the retailer undercut by a larger one — these are weak tolls, exposed to competition, commoditization, and displacement. They collect, but they cannot defend what they collect, and their shortness is brief: the cash is here today and gone when someone cheaper appears.
A step up is the personal moat — a defense that is real but tied to a person and untransferable. The law practice has this: reputation, expertise, a licence, relationships that keep competitors out for a while. It is stronger than the coffee shop, because the barrier exists; it is weaker than a true franchise, because the barrier walks out the door when the person does. The toll is defended only as long as one individual is present to defend it. Above this sits the powerful toll — defended by something that outlives any person: a network, a brand, a standard, switching costs, scale. This is the franchise, and its shortness persists the longest, because the defense is structural rather than personal.
And the nature of the moat itself matters, because not all defenses are equally durable. A natural moat strengthens with use — the payment network that grows more entrenched with every transaction, the index that anchors deeper as more assets track it. These are self-reinforcing, and the toll they protect has the longest-lasting shortness of all. A regulatory moat is the opposite kind of danger: it looks solid, but it is granted, not earned, and it can vanish at a stroke when the rule changes. A toll protected only by regulation carries a persistence that is high until the day it is zero. The source of the protection, not merely its presence, decides how durable the toll’s strength really is.
III. How the Toll Collects
The second axis is independent of the first: not how well defended the toll is, but how it arrives. A recurring toll collects on its own, period after period, without being re-won — the network’s fraction of every transaction, the railroad’s levy on freight that must move, the brand’s earning on every unit sold. A one-off toll must be regained each time: the consulting mandate that ends, the legal file that closes, the deal that must be replaced by the next deal.
This axis crosses the first. The coffee shop is a weak toll that is also, in a sense, recurring — customers return, but without protected loyalty. The law practice is a one-off toll with a personal moat. Visa is a powerful toll that is also recurring. A commoditized manufacturer is a weak toll that recurs. Strength decides how long the toll’s shortness persists; the manner of collection decides how stable it is from one period to the next. A one-off toll is the more fragile on this axis, because even its present cash is not secure — it must be produced again before it can be collected again, and it rests, in the case of the practice, on the continued presence of particular people.
There is a trap that hides on this axis, and it wears the look of growth. A one-off toll can take its cash and reinvest it into buying more one-off tolls — a firm acquiring other firms, a practice adding partners. This reinvests without building anything durable: it stacks short tolls on short tolls, more fees to be re-won, no base that compounds. The revenue climbs and the economic nature does not change, because stacking the short upon the short never produces the long.
IV. The Two Summits
At the strong, recurring end of the map, two kinds of toll overflow — they generate more cash than they can reinvest, and return the excess — and they are easily confused, because from the outside both look like superb, cash-gushing businesses. They are opposites in one respect that decides everything: whether the wave behind the toll is still alive.
The magnificent A is a toll whose wave has died. Its base is full; the world does not consume meaningfully more of what it sells each year. Coca-Cola is the pure case — a protected levy on the world’s thirst, with margins and returns that are superb, and almost no volume growth left to find. It overflows because it has finished growing. Its reported growth, where it has any, comes not from a living wave but from efficiency: price increases, cost discipline, share repurchases, the optimization of an existing base. This is real, and it is finite — one can raise prices and cut costs only so far — and it is why the magnificent A, for all its perfect metrics, has a shorter economic life than its numbers suggest. There is a quiet warning in this: the most flawless metrics are sometimes the sign not of vitality but of saturation, of a toll that has stopped investing in growth and now simply extracts. A business still spending to ride a living wave often shows less perfect numbers, because it is paying to grow.
The paramount A is the A times a B made flesh: the toll and a free wave fused, the toll that is the wave. Here growth is not bought but received — carried in by a secular movement and captured by the moat — so the business compounds as the wave widens, at almost no capital cost. Visa is the case — the digitization of payments is a secular movement Visa did not build and does not fund, and it widens still, with much of the world’s cash yet to convert. The paramount A overflows too, but for the opposite reason: not because its base is full, but because its return on capital is so high that it cannot reinvest most of what it earns at that rate, so it grows at a sober pace on the wave and returns the rest. Its rate of compounding is capped — a dozen or so points, not thirty — but its duration is long, lasting as long as the wave arrives. The cap is not a limit to lament; a sure return that compounds for decades is the rarest and best thing an owner can hold. One must never mistake the ceiling on the rate for a limit on the duration. The magnificent A earns a higher rate for a shorter time; the paramount A a steadier rate for far longer. Which is why only the free-wave toll — the Freesurfer, where A and B are one — reaches the paramount.
V. When the Toll Is Only Half Protected
One more form deserves naming, because it is the most easily misread and because it differs from a defect that lives elsewhere. The imperfect A is a toll whose revenue passes only partly through its moat. Some of what it collects is protected and defended; some is earned on open ground, where margins are competed away. Amazon is the plain case: a company widely taken to own a powerful toll, and it does — Amazon Web Services collects behind formidable protection, scale and switching costs and infrastructure others cannot match. But its retail arm is the opposite, a thin-margin business fought on open ground, competing on price with little durable defense. The two live under one roof, and a single figure blends them. Its economic duration is composite — a protected part whose shortness persists, bolted to an exposed part whose shortness is brief — and reading the company as one toll, strong or weak, misses that it is both at once.
This is distinct from the imperfect B, and the distinction is worth holding. A company like Alphabet or Microsoft has an imperfect B: its toll is fully protected — search advertising, the enterprise software franchise, collected entirely behind a strong moat — but its growth is mixed, part free and part bought, as it spends on contested ground in cloud and artificial intelligence to build a new base. The defect there is in the B, the growth. The imperfect A is the other case: the defect is in the toll itself, part of its collection protected and part exposed. One is a protected toll buying uncertain growth; the other is a half-protected toll. They are read differently, and conflating them hides where the weakness actually lies.
VI. Why the Spreadsheet Cannot Tell Them Apart
All of this is invisible to a standard discounted-cash-flow model, and the clearest proof is to put the magnificent A and the paramount A through one. Take Coca-Cola and Visa, and give each the kind of assumption an analyst reaches for by habit — five percent growth for ten years, a three percent terminal rate, chosen to be conservative. The model applies the same shape to both, because it cannot see where the five percent comes from. And the five percent means opposite things.
For Coca-Cola, is that five percent real compounding, or is it efficiency — price, cost, buybacks, the extraction Christensen described, the optimization of an existing product rather than the creation of a new market? The base is saturated; the conversion of cash into a growing base has stopped. So the growth is extracted, not absorbed, and extraction has a floor it approaches: one cannot raise prices forever. Five percent for ten years may be generous; the durable organic figure is lower. The model, assuming the growth persists, overstates Coca-Cola’s duration. For Visa, the same five percent — in reality ten to twelve — comes from a living wave that the moat captures and that keeps widening. The ten-year window is arbitrary: the wave has decades left, and the honest horizon is twenty-five years, not ten. The model, cutting the wave short, understates Visa’s duration. Put six or seven percent over a longer horizon and the valuation inverts.
So two discounted-cash-flow models, built on nearly the same inputs, can both conclude that the stock is cheap, or both that it is dear — and mislead in opposite directions, because the number carries no information about the source of the growth or the life of the wave. What distinguishes the two businesses — conversion dead and absorption saturated in the one, conversion alive on a free wave in the other — is exactly what a single growth rate agglomerates and loses. It is qualitative and quantitative at once: the components can be measured, but the nature that separates a magnificent A from a paramount one cannot be reduced to a figure without destroying the distinction. This is why the kind of toll is read, not computed. A model that treats a saturated extractor and a living free-wave toll as the same asset is precisely wrong; a reading that names the one magnificent and the other paramount is approximately right.
VII. Reading the A First
The lesson is one of order. Before asking what a business will grow into — before the B, the conversion, the discounted flows — one asks what kind of toll it already is. What defends it, and how durably: nothing, a person, a franchise, a self-reinforcing network, or a rule that can be revoked. How it collects: once, or again and again on its own. Whether its revenue passes wholly or only partly through its moat. And, at the strong end, whether the wave behind it is dead or alive — magnificent or paramount. These questions are answered by looking, not by modeling, and they sort most businesses before a single projection is made. The word toll tells you a company collects a fee. It tells you nothing else until you have asked what kind — and what kind is most of the answer.
And decomposing the toll reveals something about duration itself. A business is dual, short and long, the A and the B — but these are not two fixed lengths, one simply short and one simply long. Each has a duration of its own that varies. The A is short by nature, yet how long its shortness persists ranges from a single season, for the toll with no moat, to decades, for the toll behind a self-reinforcing network. The B is long by nature, yet how long its wave runs ranges from a base that saturates within a decade to one that widens for a generation. The economic duration of a business is therefore not the sum of a short block and a long block. It is the combination of two variable durations — the persistence of the short and the life of the long — each moving on its own axis. The piece on economic duration looked mostly at the life of the long; reading the A reveals that even the short has a length, and that the length of the short is, as often as not, where the reading begins.
Watch the video