Every profitable business attracts imitators. The only question that matters for strategy is what happens next: do the imitators take your customers and your margin, or do they bounce off something they cannot replicate? That something is a moat, and being clear-eyed about whether you actually have one is the difference between a strategy and a wish.

The quick version

  • A moat is a structural advantage that lets you keep earning above-average returns even after rivals try to copy you. A great product, a clever campaign or a price cut is not a moat, competitors match those in months.
  • The durable sources are few and well-named: scale economies, network effects, switching costs, a unique resource, counter-positioning, brand and process (Helmer's 7 Powers). Most real moats are one or two of these, not all of them.
  • Moats erode. Buffett's own framing is active, you "widen the moat" every day, or it silts up. Defensibility is a verb, not a trophy.
  • The move: name your one real source of advantage in a sentence, test it against the question "could a well-funded rival neutralise this in a year?", and route your roadmap toward the answer that is no.

The idea in depth

The metaphor is Warren Buffett's, and he has been remarkably consistent about it. In his 1996 Berkshire Hathaway letter, writing about GEICO, he put it plainly: the company's cost advantage meant "the economies of scale we enjoy should allow us to maintain or even widen the protective moat surrounding our economic castle" (Berkshire Hathaway, 1996). Nearly a decade later, in the 2005 letter, he sharpened the point into a discipline: "When our long-term competitive position improves as a result of these almost unnoticeable actions, we describe the phenomenon as 'widening the moat.'" The castle is the business; the moat is whatever keeps the rest of the market from storming it.

What Buffett leaves as a metaphor, the academic literature treats as a structural fact. Imitation, new entrants and substitutes erode almost all competitive advantages over time, that erosion is the default, and persistent above-average returns are the exception that demands an explanation (a thread running through Porter's work and the strategy-persistence research that followed, e.g. McGahan & Porter on how industry and firm effects persist). The working assumption, then: treat your current margin as guilty until proven innocent. Assume it decays, and go hunting for the specific mechanism, if there is one, that is actually holding it up.

Porter: a moat is a position, not an efficiency

Michael Porter's 1996 Harvard Business Review essay "What Is Strategy?" draws the sharpest line in the field. Operational effectiveness, doing the same activities better, faster, cheaper, is necessary but not defensible, because the techniques diffuse: rivals benchmark, consultants spread best practice, and everyone converges. Strategy is choosing a different and valuable position, defended by deliberate trade-offs (the things you choose not to do) and by fit, activities that reinforce each other so that copying one in isolation does nothing (Porter, HBR, 1996).

Fit is the underrated half. A rival can copy Southwest's point-to-point routes or its single aircraft type; copying the whole interlocking system, fast turnarounds, no meals, no seat assignments, no baggage transfers, secondary airports, means giving up the business they already have. So audit your system, not your feature list. Write down the activities a copycat would have to take on together to beat you, and the commitments they'd have to abandon to do it. The harder that trade-off, the deeper the moat.

flowchart TD
  A(["A profitable business"]) --> B{"Rivals copy it. What stops them keeping your margin?"}
  B -->|"Nothing structural"| C(["Operational lead
(erodes, not a moat)"])
  B -->|"A structural barrier"| D(["Defensibility / moat"])
  D --> E(["Scale & cost"])
  D --> F(["Network effects"])
  D --> G(["Switching costs"])
  D --> H(["Unique resource"])
  D --> I(["Brand & counter-positioning"])
A good lead is not a moat unless a structural barrier survives imitation. Leaders Loop

Helmer's 7 Powers: the moats actually have names

The most useful catalogue is Hamilton Helmer's 7 Powers (2016). Helmer defines Power as "the set of conditions creating the potential for persistent differential returns", and argues a power only counts if the benefit is simultaneously superior (it improves cash flow), significant (materially) and sustainable (immune to competitive arbitrage) (7powers.com). His seven are: scale economies, network economies, switching costs, branding, cornered resource, counter-positioning and process power. The discipline is the filter, not the list: if your supposed advantage isn't sustainable against a determined rival, it isn't a power, however nice it is.

The internet era narrows the field further. The venture firm NFX argues that of the four defensibilities that travel online, brand, scale, embedding and network effects, "Network Effects is by far the most powerful," because each new user makes the product more valuable to the others, so a late copycat faces an empty network (James Currier, NFX, 2021). The practical read: pick the one power you are genuinely building toward and pour everything into it, rather than dusting a feature with all seven and hoping. Real moats are usually one mechanism dug deep, not a checklist touched lightly.

An honest limitation. Moats are clearer in hindsight than in the moment. Much of the persuasive writing on the topic, Buffett, NFX, even crisp case studies, is built on winners, which invites survivorship bias: the graveyard of companies with "network effects" that never reached critical mass is quiet, because it doesn't publish. And the empirical strategy research is humbler than the metaphor suggests, firm-specific advantage explains real variance in returns, but so does industry structure and plain luck, and the share attributable to each is contested. Treat the frameworks as a lens for asking better questions, not as a guarantee that naming a power confers one.

A worked example

Picture a B2B scheduling tool for dental clinics, call it ChairFill (illustrative; figures below are made up to show the reasoning). It has a clean product and is growing nicely, and the founder is convinced the moat is "we're just better." Run the test.

Better is operational effectiveness, Porter's warning. A funded rival could match the feature set in two or three quarters, so "better" is not the moat. But three structural things are quietly forming. First, switching costs: each clinic has loaded two years of patient history, recall rules and insurance templates into ChairFill, and a receptionist who knows it cold. Ripping that out mid-quarter is genuinely painful, say churn runs at roughly 4% a year while a feature-equivalent rival's trials churn at 30%+. Second, the beginnings of a network effect: clinics that share locum dentists can pass shifts to each other inside ChairFill, so the tool is worth more where more nearby clinics already use it. Third, a cornered resource: an exclusive data-sharing deal with the regional dental association that a competitor cannot simply buy.

Now the roadmap writes itself. The founder's instinct, "build more features to stay better", pours effort into the one thing rivals can copy. The moat-led move is to deepen the three things they can't: make onboarding migrate even more clinic-specific data (raising switching costs), make the locum-sharing network the default behaviour (so density compounds), and extend the association deal to neighbouring regions before anyone else courts it. Same team, same budget, pointed at defensibility instead of parity.

flowchart LR
  A(["Deeper data onboarding"]) --> B(["Higher switching costs"])
  C(["Locum-sharing default"]) --> D(["Local network density"])
  E(["Association data deal"]) --> F(["Cornered resource"])
  B --> G(["Margin survives imitation"])
  D --> G
  F --> G
Illustrative: routing the same roadmap toward the things a rival can't copy. Leaders Loop

This is also where moats meet adjacent tools. A clear-eyed read of ChairFill's position is really an exercise in the resource-based view and VRIO, which of its resources are valuable, rare, hard to imitate and organised to exploit, and the threat it's defending against is best mapped with Porter's Five Forces, where a moat is simply a high barrier to the rivalry, entry and substitution those forces describe.

Frequently asked questions

Is a strong brand a moat?

Sometimes, but less often than people assume. Brand is a real power in Helmer's list and a defensibility in NFX's, but only when it changes what customers will pay or accept, not merely whether they recognise you. The test: would a blind, identical product from an unknown rival lose to yours on price? If yes, the brand is doing structural work. If customers would happily switch for 10% off, you have awareness, not a moat.

Do startups need a moat on day one?

No, and demanding one too early kills good ideas. Early on you need a wedge, a reason to win a first beachhead, and a credible path to a moat as you scale (network density, accumulated switching costs, a resource you can corner). The mistake is the opposite: scaling hard on an advantage that's pure operational effectiveness, then watching margins collapse the moment a funded competitor notices the category.

Does AI destroy moats or create them?

Both, and it's early. If a capability becomes a cheap API call, advantages built on it erode fast, that's the destructive side. The durable side tends to sit around the model, not in it: proprietary data, distribution, switching costs and workflow embedding. Treat "we use AI" as operational effectiveness (everyone will) and ask where the defensible accumulation is. This is a live, contested area, hold conclusions loosely.

Can you have too deep a moat?

A moat is protection, not permission to coast. Porter's trade-offs cut both ways: the commitments that lock rivals out can lock you out of a shift in the market, the classic incumbent caught by counter-positioning, unable to adopt the disruptor's model without cannibalising its own. A moat buys time to adapt; it does not excuse you from adapting.

How do I tell a moat from a good quarter?

Ask one question: "If a well-funded, competent rival set out to neutralise this specific advantage, could they do it within a year?" A good quarter, a launch, a campaign, a price move, fails that test instantly. A moat survives it. If you can't articulate the mechanism that would stop the rival, you're looking at a good quarter.

Related in the Toolkit

Where to go next