I keep watching people position things from the wrong end, and it took me a while to see why it bothered me.

They start with the words. What should we say? What word do we want to own? I understand the pull. Words are the part you can argue about in a room and walk out feeling like you did something. I’ve sat in those rooms, and for a long time (in my early years) I thought that was the work.

The longer I look at companies that came to mean something, the less I believe a position is written at all. It forms on its own, off to the side, out of everything the company kept doing, refused, and paid for over years. The company doesn’t get the last word on what that adds up to. The market does. By the time there’s a word on the thing, someone else has already chosen it, off the back of what they watched the company do.

So there are two words in play, and most of the confusion comes from treating them as one. There’s the word a company claims, and there’s the word it gets handed. When those line up, the claim reads as true. When they come apart, the words are decoration, and the real answer is sitting in the P&L waiting for someone to read it.

There’s a mechanism under that I didn’t appreciate until I’d seen it a few times. When people sense they’re being sold a claim, they brace and start hunting for the catch. Tell someone you’re the trustworthy one and you’ve sent them looking for the reason you’re not. A refusal buried in your cost structure asks nothing of anyone, and it gets believed precisely because it was never pointed at them. The words draw the scrutiny, and the proof slips in underneath it.

Where it starts, every time I trace it back, is a belief most people thought was wrong.

Peter Thiel has that interview question I keep returning to. What important truth do very few people agree with you on? A real answer has a particular shape. Most people believe one thing, and the truth runs closer to the opposite. When I meet a founder who can’t answer some version of that, I usually find a preference rather than a belief, and a preference gets you a product, rarely a whole category.

The language that helped me name it came from Chris Argyris. He talks about the governing variables an organization runs on, the assumptions it keeps inside safe limits without ever examining them. Most learning happens inside those limits. You adjust what you do to keep things working. Every so often, someone questions the limits themselves, and that’s the rarer move. The day a company gets founded is usually the first time someone looked at the assumption everyone else takes for granted, the one baked into the shape of the industry, decided it was wrong, and started acting from the corrected one.

Netflix got named for streaming years before streaming worked, and Reed Hastings has been clear they didn’t call it DVDs-by-mail, because the belief was about where this was all going, and the discs were never the point. Stripe started with the idea that taking money online should be seven lines of code instead of seven weeks of procurement, which quietly shifted the comparison from payment features to how fast you got your first dollar. Figma started from the belief that design was already something people did together, while every tool still treated it as a solo act, and that the browser was the only honest place to build it. Dylan Field said later that launching it felt like heresy.

What I think is happening is that an incumbent gets trained by its own wins to keep confirming the patterns that produced them. A founder walks in carrying none of that and sees the thing the incumbent’s whole organization has learned to look past. It’s why almost none of the strongest positions I can think of were built by anyone holding a positioning playbook. They got built by people who didn’t know there was one yet, so they weren’t following it.

None of that matters, though, until the belief starts costing something.

This is the part most positioning work skips, and I think it’s the whole game. A belief that stays cheap is just a claim, the kind a competitor can lift by Monday. It becomes an operating model the moment the company has to give something up to hold it. So when I want to know what a company actually believes, I stop reading the words and look at the decisions. What they built, what they refused, how they priced it, who they hired, what it cost them, what they walked away from. That’s the position, and the market reads it long before it reads the brand deck.

Roger Martin gives me the cleanest way to hold it. He treats strategy as a set of choices that have to reinforce each other: where to play, how to win, what you’d need to be good at, what has to sit underneath. You can’t make those one at a time and lock them. You run them against each other until they cohere. An operating model is all of that answered at once, and the honest version of the answer is just where the money goes.

The decisions that tell me anything are the expensive ones, the choices that cost the company something a sensible competitor would have kept. Costco has sold the same hot dog and soda for $1.50 since 1985 and loses money on each one. When the price came up, Jim Sinegal’s answer to his CEO, as it’s been told, was that if he raised it, he’d kill him, so figure it out. That’s forty years of swallowed margin, and it buys a position nobody reaches by rewriting a homepage. Patagonia still runs its service through a phone line and email and won’t add chat, because someone decided a real conversation was the right answer and the cost of holding it is the proof. None of that shows up in the messaging. You find it in the staffing plan and the cost structure.

The part I find most underrated is that this is the only way a belief survives the founder. A belief that lives in one head dies the first afternoon that person isn’t in the room. The operating model is the one form it can take that outlives them, because once the costly decisions are wired into the structure, others carry them out without being told. That’s why the early hires aren’t really staff. Brian Chesky set Airbnb’s values before he hired anyone and screened for belief over raw ability, on the logic that an early hire makes calls when he’s not there and then hires the next people in their own image. His way of putting it stuck with me. Set aside integrity and honesty, since everyone says those. The values that matter are the things you hold to be true that other companies don’t.

Michael Porter explains why any of this is defensible. He separates being good at operations, which anyone can copy, from strategy, which is a system of activities that fit together. The fit is what locks competitors out, because a chain is strongest at its weakest link, and a rival has to match the whole chain at once. So the test I keep coming back to is simple. If a competitor wanted the position you have, what would they actually have to rebuild? If the answer is the homepage, there’s nothing under it. If the answer is the hiring bar, the supply chain, the cost structure, and the way the product is built, there’s real weight holding the word in place. That weight is the only part worth much.

What the market does with all that, given enough time, is hand you a single word.

The words that stick are nouns. Safety is one. Scarcity is another. The adjectives never hold the same way, because anyone can wear “innovative” or “customer-first” for a week, and nobody can stop them. The test I use is whether you can name your category and then the one concept you own inside it, and whether that concept is yours alone or a word your competitor would reach for too. You own it when the market can’t make sense of you without it.

The jar can’t read its own label from the inside, and that’s the uncomfortable part for whoever built the thing. The founder spent the whole climb one layer up, acting on the belief, while the position quietly set behind them, which leaves them about the worst-placed person to see it clearly. They keep telling the story that got them here and miss the position their own decisions laid down.

Tesla is the one I keep coming back to, and it’s worth not flattening into a slogan. The mission in the filings is to accelerate the world’s transition to sustainable energy. That’s the frame the company claims. The frame the market reaches for, the one that moves the cars off the lot, is future. The claimed word and the assigned word aren’t the same, and the gap between them isn’t a flaw. It’s what it looks like when a company’s decisions earn it a meaning the founder never set out to own.

Volvo runs the whole thing backwards from the way it usually gets sold. Nils Bohlin developed the modern three-point seatbelt in 1959, and Volvo left the patent open for other carmakers to use free of charge, in the name of public safety. The claim came decades after that. The proof piled up first and the word followed, and by the time anyone said safety out loud it was already true. That’s the version that convinces me most, because nobody wrote it.

There’s a distinction in here I kept tripping over until I forced myself to separate it out.

A company headed somewhere right by a wrong route looks, at the moment it dies, identical to a company whose belief was just wrong. They’re different deaths, and they need different diagnoses.

The destination is the concept the company set out to own. The route is the particular run of decisions, timing bets, and technology assumptions that it made to get there. You can be dead right about the destination and dead wrong about the route, and most early pivots I’ve seen are someone hunting for a route the market will actually walk this year, rather than giving up on where they were headed.

Webvan and Instacart are the same destination down two routes. Webvan launched in 1999, betting on groceries delivered to your door, built automated warehouses before anyone had proven there was demand, and went under in 2001, having burned through more than $800 million. The destination was right. Every grocery chain delivers now. The route was heavy in a year when smartphones and live routing didn’t exist, and no one felt comfortable buying groceries on a screen. Instacart took the same destination down a light route a decade later, using existing stores as warehouses and people’s own cars for deliveries, and the phones that hadn’t existed in 1999 made the whole thing coordinate. The insight had nothing to do with groceries. It was a better read on which route the destination needed, given what actually existed at the time.

There’s a third version that’s crueller. Right destination, right route, wrong decade. General Magic built something close to the smartphone in the early 1990s, touchscreen and a software modem and all, something like fifteen years before the world could hold it. It died, and the people who built it went on to build the iPod, the iPhone, Android, and eBay. Being early is impossible to tell apart from being wrong while you’re living through it. Only the clock separates them, and the clock never rules in time to save the company that was right too soon.

The way I tell the two apart is to watch which way the evidence points. A company wrong about the route keeps having its destination confirmed by factors outside it, even as its own bets keep failing. A company wrong about the destination collects the opposite. The product works as it was meant to; the delivery is clean, and people still don’t want it, even under the exact conditions the belief said would turn them. The phrase that hides this from a founder is “we’re just early,” because it fits both a real early bet and a dead-wrong one equally well. So I stopped asking whether the thing could be true someday, since almost anything could. I started asking what the belief would have to show, and by when, and whether it’s showing it.

There’s one more wrinkle that nearly broke my own rule, so I want to be careful with it.

Sometimes the market doesn’t just fix your route. It hands you a different destination, a word nobody set out to own, and the whole win turns out to be taking it.

Slack came out of a failed game. Stewart Butterfield’s company built an online game that didn’t work, and the little tool the team had made to talk to each other became the actual product, and Salesforce eventually paid $27.7 billion for it. Twitter came out of a podcasting company that Apple made pointless overnight. Instagram was a cluttered check-in app until the founders noticed people only used the photos, cut everything else, and relaunched it; after that, Facebook paid a billion for it. PayPal started as cryptography for PalmPilots before it found its real word on eBay. In none of those cases did the founders stop believing they could make something real. They figured out the real thing wasn’t the thing they’d named.

This complicates the idea that you can’t run the sequence backwards, and I’d rather sit with that than smooth it over. You still can’t start with a word and reverse-engineer the decisions that earn it. The thing that can move is the destination. The belief doing the work is a belief in making something true, and the specific word it settles into gets discovered along the way, read back off whatever the market keeps reaching for once the thing is in their hands. Clinging to the first word after the market has offered a better one is its own kind of failure. The discipline has nothing to do with loyalty to where you thought you were going. It’s reading which word the market is actually handing you, and being humble enough to take it.

Here’s the part that actually matters to me, because anyone can line up winners and call it a pattern.

I went looking for the evidence that would kill this. I wanted the company that believed something true, built a coherent model around it, and still died, because if I couldn’t find one, then I wasn’t describing a mechanism; I was reading tea leaves. They weren’t hard to find.

Quibi believed something true about the world and wrong about the offer. Jeffrey Katzenberg and Meg Whitman raised $1.75 billion for premium, professionally made, mobile-only short video, and the operating model carried that belief faithfully. The short mobile video did show up at a scale almost nobody predicted. Premium, paid, and mobile-only didn’t. Against a first-year plan of more than seven million subscribers, they had a few hundred thousand, and they shut it down about six months in. The farewell letter put the failure down to the idea itself not being strong enough, or to timing. I think it was the first one. The belief was wrong, and raising that much before launch had wired it in so tight there was no room left to move.

Then there’s the belief that’s right and still breaks under its own weight. Friendster was early, the demand was real, and the social network it was pointing at is exactly the one that arrived. The pages buckled under the load, the whole thing got slow and painful to use, and MySpace and then Facebook walked off with the word it had reached first. Being right and being first didn’t survive a machine that couldn’t carry the traffic.

And there’s the belief that’s right but leans on people you can’t command. Better Place built a genuinely coherent answer to range anxiety: a network where you swapped your electric car’s battery instead of waiting to charge it, and raised nearly a billion dollars to build it. It only worked if carmakers standardized their batteries around it, and only Renault ever did. The belief was right, and the model was coherent, and the company still went under in 2013, because the rest of the industry was under no obligation to play along.

Which is why I keep looking at the planes that didn’t come back. If I only study the ones that landed, I’ll conclude that belief and coherence are enough, and I’ll be wrong, because the companies that had both and died aren’t around to argue. They had conviction and a coherent model, and they still lost to timing, to the money running out, to a dependency, to a better-funded copy. So the most I’ll claim is that the thing explains how a position forms. It doesn’t promise the company lives.

There’s a harder problem under all of this, and it points back at me, not just at the founder.

The whole thing assumes the position can be read, and the reading is shaky in a few specific ways. Customer language comes after the fact and splits by who you ask, so different customers hand you different words. The refusals that matter are often invisible from outside, and a refusal made out of conviction is hard to tell apart from a corner the founder was backed into. And then there’s me. I’m the one supplying the word. Hand the same company to two careful readers, and they’ll read two different positions off it. Tesla is sustainable energy for one and “the future” for the other, and both can defend it.

So I lean on a rough hierarchy of evidence.

What a company says about itself is the weakest thing in the pile, and its campaigns are barely better. Founder interviews help without settling anything. Customer language helps and then gets rewritten in memory after the fact. The strongest evidence by a distance is a sacrifice you can see in the structure, a refusal sitting on the books, because that’s the thing a company can’t fake and wouldn’t carry unless it meant it.

This is the trap I have to name out loud, or none of this is worth anything, because a reader who always finds a clean pattern is just doing astrology. The thing that saves the reading is when it converges. The word the refusals point at has to be the same word the customers keep reaching for. When only one side carries it, what I’ve found is either an aspiration the company never paid for, or a belief that hasn’t reached the market yet. And the refusals, I trust, are the ones you can see from the outside. Patagonia handing the whole company to a trust whose only beneficiary is the planet. Volvo opening the seatbelt patent. Those aren’t read off a deck. You can see them on the books, and that’s what makes the word they earn something I can believe, rather than something I imagined.

So the honest version of all this is smaller than what usually gets sold, and I think it’s stronger for being smaller. A belief most people called wrong, wired into decisions that cost something, run long enough that the market reads a word off the behaviour and starts using it without thinking, held in place by how deep those decisions are buried. That’s the whole of it. It isn’t a guarantee you win. Companies with the right belief and the right model still run out of cash, still split at the top, still get caught by a regulator, still get out-spent by a copy. The thing reads backward far better than it predicts forward, and I’ve made my peace with that.

Which leaves the one question I’d actually put to a founder, and it isn’t a kind one. Take the words away. Look at what the company has genuinely refused to do, and where the money goes when nobody’s watching the homepage. Name the word those decisions have earned. Then ask whether it’s the word you think you own.

The market already knows the answer. It settled on the word a while ago, while you were still upstream, busy acting on the belief, the last person in the building who could read your own label.