The Five Ways a Moat Breaks

Nine companies. At their peak, every one of them held a position that looked as defensible as anything in business. Eight of them lost it. The ninth did not, and the ninth is the most instructive of the set.

This is the last piece in The Moat That Broke, the series that began in August with Nike walking away from its own retail shelves and Intel handing TSMC the most valuable position in technology.

Before it, this newsletter mostly published winners. The series went the other way on purpose, because the mechanism that destroys an advantage is more teachable than the one that builds it.

Nine autopsies later, the useful finding is that moats do not break in nine different ways. They break in five.

The Number Tells You Late

Every company in this series had the data. Not one of them was flying blind.

Intel could see its 10 nanometre schedule slipping in 2015. Nike could see wholesale revenue falling, because it was falling on purpose. Under Armour could watch its own discount rate climb quarter after quarter. Peloton had the best subscriber retention data in consumer fitness.

The problem is that a metric reports the state of an advantage after the advantage has already changed. Revenue is the last thing to move. By the time the number is bad enough to force a meeting, the mechanism that produced it has been running for years, and the cost of reversing it has compounded the whole time.

So the number is not the diagnostic. The structure is. And structure has to be examined deliberately, because unlike revenue it does not report itself on a quarterly cadence.

What follows is the examination.

The Five Ways

1. You were renting it

Some advantages compound while you sleep. A network effect gets stronger with every user. A switching cost deepens every year a customer stays, which is why Oracle can ship worse software and keep the account.

Manufacturing leadership does none of that. It is re-earned every eighteen to twenty-four months, at enormous capital cost, with execution risk that does not go away because you succeeded last time. Intel missed one cycle and never recovered the position, and the reason it could not recover it is that the rent had already been paid to somebody else.

Boeing rented too, though it would not have used that word. Its advantage was engineering judgement held by engineers, which is a capability that has to be continuously funded, staffed and defended. Stop paying for it and it leaves, slowly, in a way no quarterly metric captures until something falls out of the sky.

The test: does this get stronger if we do nothing for two years, or weaker?

2. Somebody else was holding it

This is the most common failure in the series and the least understood.

Nike’s shelf space was held by retailers. Nike could withdraw from it, but could not take it with it, and the shelf did not sit empty waiting for a change of heart.

Bud Light’s position was held in the customer’s head, as the beer you ordered when you were not choosing, which is a real advantage that appears on no balance sheet and can be revoked without notice. Zoom’s access to the enterprise was held by Microsoft, which could include an acceptable version of the same feature in a suite the customer had already bought.

In all three cases the company read a borrowed asset as an owned one. Market share flattered the mistake, because share looks identical whether you own the position or merely occupy it.

The test: who else would have to agree for this to keep working?

3. It was weather, not climate

Peloton read a pandemic-shaped demand spike as proof of structural advantage and built permanent fixed costs against it. WeWork funded ten and fifteen year leases with thirty day cancellable memberships, which is a bet that growth never pauses.

Both failures share a shape. A temporary condition produced a real number, the number was treated as the new baseline, and fixed obligations were committed against it. The obligations outlived the condition. Notably, neither company was wrong about demand.

Flexible workspace was a genuine need and still is. Connected fitness retained better than almost anyone expected. The error was not the demand read, it was the durability read.

The test: what happens to this number if the reason for it goes away, and what have we already committed against it?

4. It dissolved while you were winning

Under Armour did not lose its position in a bad quarter. It lost it across a decade of good ones.

The brand began as performance apparel for serious athletes, a sharp answer to a specific question.

Then it broadened, one defensible decision at a time, each supported by a growth case, until it no longer answered the question at all. Once a brand is undifferentiated the only lever left is price, and discounting teaches a customer never to pay full price again.

This is the hardest failure to catch because every individual step passes review. There is no bad decision to point at in the postmortem, only a cumulative direction that nobody owned.

The test: could a customer say who this is for in one sentence, and is it the same sentence as three years ago?

5. You defended it with no stopping rule

Meta is the case where the moat did not break, and it earns its place in the series for that reason.

The diagnosis was correct. Apple’s tracking changes proved that Meta did not own the platform its business ran on, and a company that does not own its platform is renting its entire distribution.

Responding to that was right. What was missing was a checkpoint: a defined point at which the bet could be judged and stopped, and a person permitted to say so.

The bet worked out.

It worked out because the core advertising business was strong enough to absorb tens of billions of dollars in Reality Labs losses, which is a statement about the core business rather than about the quality of the bet. The same playbook is now running again at larger scale on AI infrastructure. A correct diagnosis licenses a response, not an unbounded one.

The test: what would have to be true for us to stop, and who is allowed to say it?

What a Moat That Holds Looks Like

Set these against the ones that held, because the contrast is where the definition lives.

Costco’s membership gets stronger with scale and renews itself annually in cash. Bloomberg’s terminal is worth more the more people are on it. Ferrari’s scarcity is enforced by the company itself, which means no competitor can take it away and no customer can vote it down.

And Nintendo is the sharpest case of all, because it ran the experiment on itself. It built its advantage out of mature components nobody was fighting over, abandoned that doctrine for the Wii U and sold 13.56 million units, then restored it for the Switch and sold over 156 million. Same company, same market, one variable changed.

What the survivors share is that their advantage is owned rather than rented, held by them rather than by a partner, structural rather than seasonal, and specific enough that broadening it would obviously destroy it.

Run the Moat Audit

1. List advantages, not metrics

Write down what you believe protects you, in plain language, without numbers. “Customers cannot easily leave” is an advantage. “Retention is 94%” is a metric that may or may not be evidence of one. Most teams cannot produce this list, which is itself the finding.

2. Put every item through the five tests

For each advantage: is it rented, is somebody else holding it, is it weather, is it dissolving, and are we defending it without a stopping rule. Most advantages fail at least one. The ones that fail three are not moats, they are current conditions.

3. Schedule the rent

For every rented advantage, name the cycle and the next payment. Process leadership has a node cadence. Engineering culture has a hiring and promotion cadence. Put those dates in the same calendar as the debt maturities, because they are the same kind of obligation.

4. Map the counterparties

For every advantage somebody else is holding, name them. The retailer, the platform, the distributor, the customer habit. Then ask what it would cost to take it back after they have stopped depending on you, which is the number Nike is paying now.

5. Write the stopping rule before you commit the capital

Any bet large enough to matter gets a written condition for abandonment, agreed before the money moves, and a named person with the standing to call it. A stopping rule written after the fact is not a stopping rule, it is a negotiation.

Run this once a year, at board level, and keep the previous year’s answers so the drift is visible.

Every advantage you hold is either compounding or decaying right now. The companies in this series were not badly run and were not short of information. They simply never asked which of theirs was which, until the number asked on their behalf, years late and at full price.

Know which of your moats you own, and which you are only renting.