♛
14.07.2026
Shape vs Moat
Read Time: 12 Minutes
Hello
How to Know If Your Company Survives the Day It Gets Copied
The Three Engines framework, eight case studies, one forty-year-old counterexample, and the four tests I run before I believe any defensibility story.
Over my career, I have sat in rooms where companies decided what made them safe, and I helped win over $2 billion in strategic deals by watching what actually protected the winners. In almost every one of those rooms, someone pointed at the product and called it the moat. In almost every case, they were pointing at the wrong thing. This piece is the correction, written so you can run it on your own company this week.
Read it as a working session, not an essay. There is a framework, a set of case studies to steal from, and a worksheet at the end. By the last section, you will be able to answer one question with evidence instead of hope: if a competitor with more money ships a copy of your product tomorrow, what do your customers do on Thursday?
The framework is called The Three Engines. Here is the claim it rests on, stated once, plainly.
Product shape is everything about your product a rival can observe: the interface, the workflow, the output format, the pricing page, the demo. A moat is whatever makes customers stay when a rival offers the same thing cheaper. Shape is visible. Moats are not. And anything a rival can see, a rival can copy. So shape is never the moat. It is the paint on top of three engines: segment knowledge, workflow entrenchment, and a data loop. Companies with the engines survive being copied. Companies with only the paint do not.
That is the whole thesis. The rest of this piece is the evidence, the psychology of why you will resist it, and the tests that tell you which side of it you are on.
Before the evidence, one picture to hold. A medieval castle’s silhouette, the towers and crenellations, is what every traveller could see and every rival lord could sketch. Castles were not defended by their silhouettes. They were defended by the granary that outlasted a siege, the well inside the walls, and the sworn men who showed up when the horn sounded. Plenty of lords built impressive silhouettes over empty granaries, and history does not remember them, because the first serious siege ended the story. Shape is the silhouette. The engines are the granary, the well, and the men. Everything that follows is that picture with proof attached.
I. The Story You Have Been Sold
Start with the pattern, because you have heard this story recently even if you do not remember where. For four years, AI application companies have rotated through defensibility stories to raise money. First, fine-tuning was the moat. Then evals. Then model routing. Each story expired when the underlying capability became a commodity, and a new story slid into its place. A founder of a legal AI company recently named this cycle a shell game, and he was right. Then he offered his own answer: the real moat is product shape, the fit-for-purpose arrangement of screens and workflows a focused team builds for one vertical. A great kettle cannot be a great toaster. Model providers cannot maintain a thousand shapes. Shape yourself tightly to one job, and you are safe.
It is a clean argument. It reads well. It feels true the way a rhyme feels true. It is also the fifth story in the very shell game it describes.
Strip it to the testable claim: fit-for-purpose product design is a durable defence. For that to hold, two things must be true. Companies with strong shape should survive. Companies without it should bleed. That is what necessary and sufficient mean once you take the Latin off.
So the test is simple. Find one company with a famously bad shape and a fortress business, and the “necessary” half dies. Find companies with beautiful shapes that got run over anyway, and the “sufficient” half dies. Both exist. In quantity.
Do this before reading on: write down, in one sentence, what you currently believe your own moat is. You will test that sentence against everything that follows.
II. Not Necessary: The Ugliest Product in Finance
The Bloomberg Terminal is a black keyboard and an interface that has looked broadly the same since the 1980s. Amber text. Cryptic function codes. A learning curve firms pay to train people through. By every rule of fit-for-purpose design, this is a shape disaster. If shape were the moat, Bloomberg should have been dismantled decades ago by any of the dozens of competitors that shipped cleaner, friendlier, cheaper screens. Many tried. Money was spent.
Instead, a single seat costs $31,980 a year as of January 2025; the price has only ever moved up, and there were about 325,000 subscribers as of 2022 generating more than 85% of Bloomberg’s revenue.
Why? Not the screens. The moat is the other 324,999 people. Traders message each other inside the terminal. They trade through it. Their contacts, their chat history, their muscle memory live inside it. Leaving Bloomberg does not mean switching software. It means leaving the room where your counterparties are. That is workflow entrenchment plus a network, and it has carried the worst shape in professional software for forty years.
One case is all you need to kill a “necessary” claim. A company with terrible shape built one of the deepest moats in software. Whatever is defending it, it is not the paint.
Do this: name the ugliest product in your own industry that refuses to die. Then write down what is actually holding its customers. That answer is a moat you can study for free.
III. Not Sufficient: Three Bodies
Now the other direction. Companies that had the shape and lost anyway. Three bodies, three decades.
Jasper, the AI writing tool. The first big AI application company had exactly what the shape thesis prescribes. Templates for fifty marketing jobs. Brand voice controls. Team workflows. A real, opinionated layer for marketers, built years before anyone else. Then the model provider shipped a plain chat box at $20 a month, and revenue fell from roughly $120 million to $88 million. The shape stayed. The customers left. What the company never built was the thing underneath: proprietary data the model provider lacked, lock-in that made leaving painful, switching costs beyond a saved template.
Stories. Snapchat invented a genuinely new product shape: ephemeral, vertical, tap-through stories. Instagram copied it feature-for-feature in 2016, and within eight months Instagram Stories passed Snapchat’s entire user base, reaching 250 million daily users against Snapchat’s 166 million by mid-2017. Same shape on both sides. The winner was the one with 700 million users who did not have to download anything.
Slack. The best-shaped workplace product of its decade, loved and polished. Microsoft copied the shape, bundled Teams into Office at no extra cost, and passed Slack in daily users within two years. Slack’s shape was better the entire time. The fight was never about shape. It was about who already owned the buyer.
Notice what decided all three: distribution, bundling economics, network, switching cost. Every deciding force was invisible in a screenshot. And notice the logic of the Stories case especially: when the same shape appears on both sides, the shape cancels out of the equation, and whatever remains determines the outcome. A variable that cancels out cannot be the cause.
Do this: take your one-sentence moat from Section I. If a copy would put that same feature on both sides of the fight, cross it out. It was never the moat.
IV. The Experiment Running Right Now
This is not a historical debate. The test is running live this quarter, in the exact market where the shape thesis was published.
The thesis assumes model providers cannot maintain a thousand product shapes, so they will not come after yours. The first half is true. The second half does not follow. They do not need a thousand shapes. They need the ten biggest verticals, and legal, finance, and design are on every version of that list.
The record: Anthropic has shipped thirteen vertical products since February 2025, covering coding, design, financial services, small business, and law. Claude for Legal launched in May 2026 with 80-plus legal agents, contract review, Word integration, and global firms already in production, priced around $20 per user per month. Three weeks later, OpenAI hired the founder of a $3.2 billion contract-management platform to build its own legal vertical. Read that hire carefully: a model lab recruited the person who built the defining contract-management shape of the last decade, specifically to build shapes.
Two more forces compound the problem. The labs control the layer everyone builds on, and at least one has stopped consistently warning partners before launching competing products. Your supplier is your competitor, and it can see the demand data. And the shape layer is being unbundled from below: the document platform at the core of most large law firms has shipped an open-protocol server that lets any AI agent access governed legal content, permissions intact. When the data layer opens to all agents, the proprietary container ceases to be a container. The shape becomes a skin any agent can wear.
None of this means every vertical AI application dies. It means the ones that live will not be saved by their screens.
Do this: name your own supplier-competitor. Everyone has one. Write down what they can see about your business from the traffic you send them.
V. The Copy That Had Every Blueprint
The strongest historical case is also the oldest, and Michael Porter wrote it up in “What Is Strategy?” in Harvard Business Review in 1996.
Southwest Airlines had a fully public shape: point-to-point routes, one aircraft type, no meals, fast gate turnarounds, low fares. Continental could see all of it. Continental launched Continental Lite to copy it exactly, blueprint by blueprint. The copy failed and was shut down.
What did the copy miss? Southwest’s advantage was not the list of practices. It was the fit between them. One aircraft type made turnarounds fast, and every mechanic knew how to repair them. Fast turnarounds made point-to-point economics work. The culture made it all cheap to run. And doing it meant refusing everything a full-service airline does. Continental copied the visible practices while still running a full-service airline underneath. It got the shape, and none of the system, and the system was the moat.
Toyota ran the same experiment from the other side, on purpose. For decades, it gave competitors tours of its factories. Detroit executives walked the lines, photographed the kanban cards, went home, and installed kanban cards. The gap did not close. Toyota could afford the tours because the cards were the shape. The moat was thousands of workers trained to stop the line when something looked wrong, and a management culture that treated every defect as a system to fix rather than a person to blame. You cannot photograph a habit shared by fifty thousand people.
This gives you the tour test, and it is worth memorizing: if giving your competitor a complete tour of your product would destroy you, you have shape, not a moat. Toyota passed. Southwest passed. Most software companies would fail.
Every restaurateur already knows this test, even if they have never named it. A hit restaurant opens. The room, the menu, the plating, the prices are all public. Within a year, three imitations open nearby with the same look and the same dishes, and most are gone in six months. The copies got the shape for free. What they could not copy was the supplier who saves the best fish for a twenty-year relationship, the kitchen that runs the menu on instinct, and the regulars who feel the room belongs to them. Fit-for-purpose design told diners where to look. Something else made them come back. Software is not different. It only feels different because the copying is faster.
Do this: imagine giving your top competitor a full, honest tour. Every screen, every workflow, every doc. Write down what they still could not take home. That list is your actual asset register.
VI. The Moat Behind the Complaints
Go back to Bloomberg one more time, because the deepest lesson is not that it survived with an ugly product. It is how. Bloomberg kept making paid choices its own users complained about, and those same choices are what made the terminal impossible to leave.
Users said it was too expensive. Bloomberg held premium pricing anyway; procurement analysis puts standard subscriptions in the $24,000 to $27,000 range with limited negotiating room. The price became a filter for serious professional use. The terminal is not judged against cheap tools. It is judged against the cost of being wrong.
Users said the interface was hard. Bloomberg refused to flatten it into consumer software; its executives treated the distinctive look as a selling point, and one CEO said the company had to stay “religiously consistent” because users became attached to its look and feel. Difficulty became fluency. Fluency became status. A user who knows the codes has earned membership, and people do not walk away from earned membership.
Users asked why chat was housed within a data product. Bloomberg kept communication on a professional surface; Instant Bloomberg sits at the center of the terminal experience, with built-in surveillance and compliance so that regulated firms can actually use it. The terminal became the room where the other side of the trade already is.
And when its own newsroom was caught accessing client terminal data in 2013, Bloomberg called the practice “inexcusable,” cut off journalists’ access, and then built compliance controls around restricted data. It gave up an internal information advantage to protect the thing the whole business stands on: client trust.
Here is the pattern under all of it. The moat is not the feature. The moat is the refusal behind the feature. Bloomberg refused cheapness in favour of seriousness, refused consumer ease in favour of expert fluency, refused generic tools in favour of a governed room, refused its own newsroom’s advantage in favour of trust. Every refusal cost money in the short run. Every refusal deepened an engine.
Which gives you the distinction that makes this usable: adoption friction versus seriousness friction. Adoption friction stops a new user from getting started. Remove it ruthlessly. Seriousness friction is the weight that creates trust, fluency, and auditability for professionals doing consequential work. Defend it, name it, teach it. The most dangerous roadmap habit in software is treating every complaint as a bug. Some complaints are proof the product owns a serious job.
Do this: pull your last ten customer complaints. Sort them into the two piles. If you cannot tell which pile a complaint belongs in, you do not yet know what job your product owns.
VII. Why Your Brain Buys the Shape Story
If the evidence is this lopsided, why does the shape story keep selling? Three mechanisms, all documented, none flattering. Knowing them is the vaccination.
Easy ideas feel true. The brain grades claims partly on how smoothly they process, a bias Daniel Kahneman and the fluency researchers have documented across decades: statements that are easier to read, easier to say, or delivered as rhymes get judged as more accurate than identical claims stated awkwardly. “A great kettle cannot be a great toaster” is a fluency machine. You can picture it. You can repeat it in a partner meeting. The smoothness of the sentence is doing the work the evidence should be doing.
You only autopsy the survivors. “Look around: every mature product you own is fit for purpose.” True, and empty. The products you own are the survivors. The graveyard is full of beautifully shaped products that died, and you do not own them, so they never enter your sample. Every dead product had a shape too. Distinctive shape is common. Survival is rare. A common trait cannot explain a rare outcome.
Builders overvalue what they built. Behavioural researchers call it the IKEA effect: people assign inflated value to things they assembled themselves, even wobbly ones. A founder’s craft is the artifact, the screens and flows. When that founder reasons about defensibility, the artifact is what they control and what they love, so the artifact becomes the moat. That is attachment wearing analysis as a costume.
And under all three sits the incentive. Defensibility stories get published by founders, into a fundraising market that rewards exactly one thing: a story that fits in a sentence. The genre demands the shell game. Fine-tuning, evals, routing, shape. Same shell, new pea.
Do this: next time a defensibility claim feels instantly true, slow down and ask which of the three biases fired. The feeling of obviousness is not evidence. It is fluency.
VIII. The Three Engines
So what actually protected the companies that survived contact with a bigger player? Strip away the narratives and the same three mechanisms appear every time. None of them are glamorous, which is partly why they do not trend.
Engine one: segment knowledge. Not “we know lawyers.” Knowledge that is expensive to acquire and invisible from the outside. Which clause disputes actually block deals. What a claims adjuster checks before escalating. The cleanest case is the company that built a CRM for pharmaceutical companies on Salesforce’s platform. It rented its shape from the platform it should have feared. What it owned was the regulatory workflow: compliance rules, validation requirements, audit trails that life-sciences sales legally require. Salesforce could see the shape perfectly and still could not cross the knowledge gap economically. FICO runs the same play in credit: the score is a three-digit number anyone can imitate, but the criteria behind it were refined over decades against real default outcomes and held as trade secret, and the entire lending industry wired them into its underwriting. The test for this engine: could a smart outsider with the same tools and the same public data produce the same judgment? If yes, you have effort, not knowledge.
Engine two: workflow entrenchment. The product becomes the place where the work lives, not a tool the work passes through. Bloomberg’s chat and contacts. The dentist who holds five years of your X-rays. Apple’s ecosystem, which survived Android matching the iPhone’s shape within two years because every year of use raises what you lose by leaving. In enterprise software, this engine runs on a brutal, boring truth: buyers are not optimizing for the best product; they are minimizing blame. Loss aversion runs procurement. “Nobody got fired for buying IBM” was never a joke about IBM; it was a statement about how organizations price risk. Entrenchment converts your product from a choice into a default, and defaults do not get re-litigated every budget cycle. The test: what exactly does the customer lose on the day they leave? If the answer is “nothing, they export their data,” this engine is not running.
Engine three: the data loop. The product generates data that improves it, and that data cannot be bought elsewhere. The legal AI company currently winning the enterprise market reached $190 million in annual recurring revenue and an $11 billion valuation without training its own foundation model; the market reads its defence as depth of firm relationships and the proprietary usage inside them. Waze is the consumer-pure form: every driver improves every route. This is the only engine that gets stronger while you sleep. The test: is unit one thousand better than unit one because of the first nine hundred ninety-nine? If each sale is as good as the last but no better, you have output, not a loop.
Now the relationship between the engines and shape, stated once. A team with deep segment knowledge will almost automatically ship a fit-for-purpose shape, because they know the job cold. So across the market, good shape and survival move together. That is the correlation the shape thesis noticed. But the shape is the exhaust of the knowledge, not the engine. Copy the shape without the knowledge, and you get the forgery: same brushstrokes, worth nothing at auction. The counterfeit watch is the purest version of this. A fake can be visually perfect, sell for two percent of the price, and barely dent the real brand, because the buyer of the real thing was never buying the shape. They were buying provenance. Whenever a copy fails, ask what the copy could not carry. Whatever it could not carry is the moat.
One more distinction before you score yourself, because founders conflate these two constantly: being useful is not being irreplaceable. A lab test is useful. It produces accurate output on demand. A physician is harder to replace, because the physician holds your history, sees this result against the last five, and knows which anomaly matters for you specifically. A gym is useful; the trainer who has logged your injuries and adjusted your program for three years is the reason you do not switch gyms. Products that behave like lab tests are swapped whenever a cheaper test becomes available. Products that behave like physicians accumulate the customer’s history within the relationship, and each month of accumulation raises the cost of leaving. Ask which one your product is. The honest answer is usually the lab test, and the roadmap question becomes: what would we have to remember about each customer to become the physician?
Do this: score your company from 1 to 10 on each engine, with written evidence for each score. No evidence, no points.
IX. The Positioning Read
There is a positioning lesson beneath all of this, and it is the part most operators are never told. A position is not what you say about yourself. It is what the market can verify about you when your words are removed. Run that test on “our moat is our product shape.” Delete the words. What remains is a set of screens any funded competitor can study in a free trial and rebuild in two quarters. Nothing verifiable survives. Now run it on Bloomberg. Delete every word Bloomberg has ever published about itself. What remains is 325,000 professionals who cannot do their jobs without the box, and a price that has risen for forty years without churn. The position survives with the words deleted because it was never in the words.
Coca-Cola is the same lesson in theatre form. The company keeps its formula in a literal vault and talks about the vault. Blind taste tests have long shown most people cannot reliably pick Coke from rivals; treat the specific figures as directional, but the pattern is old and stable. If the formula were the moat, private-label cola would have won decades ago. The real moat is the trucks and the memory: physical availability in more places than nearly any product on earth, and a century of consistent cues that make it come to mind first. The vault is a story about shape. The trucks are the moat. Companies often dramatize a visible asset precisely because the real moat is boring. Watch for that move in every deck you read, including your own.
And the deeper error: treating positioning as an output, something you present to the market. A durable position is an operating filter, not a presentation layer. It is the list of things a company refuses to build because they would dilute the knowledge advantage. It is Southwest refusing meals and hubs. It is Bloomberg refusing to become pretty. The companies with real moats can usually state, in one sentence, the thing they know that nobody else has paid to learn. The shape follows from that sentence. It never precedes it.
One more trap in this family, because it is fashionable again: naming a category is not a defence either. Founders spend real money coining a new market label, planting it in analyst decks, and calling the label a moat. A category name is a shape made of words. It is fully visible, free to adopt, and the moment the category looks lucrative, the largest player in the room takes the label and outspends you on it. Category creation is worth doing when it directs attention toward an engine you already run, the way Salesforce’s “no software” pointed at a genuinely different delivery model. Coined as a substitute for an engine, it is a press release with a burn rate.
A moat, meanwhile, is a claim about future margins, and it shows up in exactly one place: the numbers. Retention that does not budge when a cheaper copy launches. Pricing power that holds through a platform shift. If a defensibility story arrives with analogies and no retention curve, that is not an oversight. That is the tell.
Do this: write the one sentence. “We know ___ that nobody else has paid to learn.” If you cannot finish it, that is the work, and no interface redesign will do it for you.
X. The Worksheet: Four Tests and a Ledger
Everything above can be compressed into a diagnostic you can run in one sitting. Block ninety minutes. Answer in writing. Vague answers are failing answers.
Test one, the removal test. Delete every word of your deck, your homepage, your category name. List what the market can independently verify about your company. Rank each item by how expensive it would be for a rival to replicate. If the list is empty, you have marketing, not a position.
Test two, the tour test. Your strongest competitor gets a complete tour: every screen, workflow, and internal doc. Write down what they still cannot take home. Only accumulations survive a tour: trained judgment, earned trust, stored history, running loops.
Test three, the copy test. A frontier lab or your biggest incumbent ships your exact shape at one-tenth of your price on Friday. For each of your top ten customers, write the specific reason they are still yours the following Thursday. “Our UX is better” is not a reason. “Their three years of benchmark history lives with us” is.
Test four, the P&L test. Find your moat in the numbers or admit it is not there yet. Cohort retention against cheaper alternatives. Price increases that held. Sales cycles that shortened because risk questions stopped coming up. If you are early and the numbers do not exist yet, say “mechanisms that should become a moat,” and name the date by which each mechanism must show evidence.
Then build the refusal ledger. Two columns. Column one: every point of friction customers complain about. Column two: for each, the verdict: adoption friction or seriousness friction. Remove the first pile this quarter. For the second pile, write down what each friction protects: trust, fluency, auditability, or accumulation. Those entries are your Bloomberg refusals. A company that cannot name a single costly refusal lacks a strategy. It has a backlog.
XI. What Would Change My Mind
Rigour requires naming the falsifiers; here they are.
If, three years from now, the labs’ vertical products have stalled in legal, finance, and design, and independently shaped applications have held pricing and retention without deep data or entrenchment advantages, the shape thesis gains real support. Early enterprise adoption of a lab’s product could be tourism; production logos are not renewed contracts.
If it turns out that shape is where the data loop starts, meaning you cannot collect proprietary workflow data without first shipping the opinionated surfaces that capture it, then shape gets upgraded from decoration to precondition. This is the strongest version of the shape argument, and its authors should make it: shape as the intake valve for the moat, rather than the moat. I would sign that essay.
And if agent-driven interfaces genuinely fail in daily enterprise use, if humans insist on stable, hand-built screens indefinitely, then the shape layer retains more value than I have granted here. People do hate change. Whether that inertia survives a generation of workers who grew up delegating to agents is an open question, and I hold it as one.
XII. Thursday
The title promised you would be able to answer one question with evidence. Here it is, one last time, with the standard attached.
A competitor with more money ships a copy of your product tomorrow. What do your customers do on Thursday?
If your honest answer names an engine, you are defensible: the judgment a rival cannot reconstruct from public data, the history a customer will not abandon, the loop that makes your thousandth unit better than your first. If your honest answer is the shape, you now know exactly what you have: a head start with an expiry date. Bloomberg had ugly screens and kept the whole market. Slack had beautiful ones and lost it. The AI writing company had the perfect vertical shape and watched a plain chat box take its revenue. The pattern held for forty years of terminals, thirty years of airlines, and eighteen months of AI applications, and it will hold for whatever you are building now. That is not a death sentence. A head start is the window in which engines get built. But the window is the plan, and the paint is not.
Stop polishing the part they can copy. Score your engines tonight. Sort your complaints into the two piles this week. Write the refusal ledger this month. And put the one sentence you owe yourself at the top of the next board deck: we know ___ that nobody else has paid to learn.
If this changed how you see your own product, forward it to one founder who is currently calling their interface a moat.
Access 30+1 Positioning Reports
If any of this was useful
There are three ways to go deeper, depending on where you are.
1. Read and follow.
Everything I publish is free. The Gravity Reports, the Digest, the positioning frameworks — all of it is on the blog and in your feed and inbox.
LinkedIn or X
2. Use the systems.
I’ve built a set of diagnostics for founders and brand leaders who want to do this work on their own company. Start with the free Analyzer — it shows you the gap between what you think you sell and what customers actually buy, and tells you whether AI recommends you when buyers are searching in their language. If you want to go further, the Clarity Kit and Monopoly are built for the same diagnostic, at a deeper level.
Start with the free Analyzer
CEO Clarity Starter Kit
Monopoly
3. Work together.
I take on a small number of advisory clients each year. These are founders and leadership teams who want an outside read on their position — the same method as the Gravity Reports, applied to their own company. If that’s relevant, the best place to start is a conversation. Better call Paul.
DIGEST
