To Robert Ikwudinso; for all the bad songs you taught me, coolest big brother.

Somewhere in the early 2000s, a software engineer at a small avatar-chat startup watched six months of work evaporate. His team had built an elaborate system for importing contacts from other messaging networks – the whole product hinged on the assumption that people would want to chat with their existing friends using 3D avatars. They’d architected the import pipeline with real elegance, handled edge cases for three different protocols, built a sync engine that quietly delighted the engineers who made it. Nobody outside that room cared. When they finally talked to users – actually sat across from them and watched their faces – they discovered the opposite: people wanted to meet strangers, not import the social lives they already had. The assumption was wrong, the six months were gone, and the team stared at a codebase that was technically excellent and fundamentally useless. The lesson lodged itself so deep that it eventually reshaped how an entire generation thinks about building things.

And the lesson was reasonable enough. Don’t spend years constructing something nobody wants. Build the smallest version that can test your core assumption, put it in front of real people, measure what happens, learn from it, adjust, repeat. Build-Measure-Learn. The Minimum Viable Product. Ship fast, fail cheap, iterate relentlessly. This loop – and the book that codified it in 2011 – sold over a million copies, spawned meetups in more than a hundred cities across seventeen countries, colonized accelerator programs and MBA curricula, and achieved something rare for a business methodology: it became common sense. To reject it felt not just contrarian but reckless, the way questioning germ theory might feel to a first-year med student. You don’t argue with what works. For a certain class of products, it does work beautifully. The story of how a single software company’s painful lesson became every founder’s operating religion – applied indiscriminately to restaurants, hardware, luxury goods, games, regulated industries, and complex platforms – is also the story of how good advice, stripped of its original context, can quietly destroy the very things it promises to save.

The term itself predates the bestseller by a full decade. In 2001, a product development consultant coined “Minimum Viable Product” with a definition far more careful than the one that eventually went viral. His version was an optimization problem: the product big enough to cause adoption, satisfaction, and sales, yet not so big as to be bloated and risky. A balancing act plotted on a risk-versus-return grid. The word “minimum” was doing real work in that sentence – it meant the minimum that could succeed, not the minimum that could exist. What happened next is what always happens when a subtle idea meets an anxious audience. The nuance fell away. “Minimum” collapsed into “as little as possible.” “Viable” collapsed into “functional enough to demo.” The methodology’s own architect later protested that “MVP, despite the name, is not about creating minimal products” – a clarification that arrived years too late to matter. The intellectual scaffolding had already been stripped down to a bumper sticker: ship something, anything, and let the market tell you what to build.

A parallel thinker, whose work on customer development fed into the same movement, acknowledged that his framework was originally designed for enterprise software – a context where you could prototype interfaces, demo to prospects, and iterate on contracts before building the full system. Web startups that deploy code fifty times a day occupy a similar sweet spot: changes are cheap, feedback is instant, and the cost of being wrong approaches zero. No other industry operates this way. Not restaurants. Not hardware companies. Not game studios. Not anyone building a product whose value depends on the totality of the experience rather than the function of its individual parts. The methodology spread anyway, carrying its assumptions like invisible baggage into contexts where they didn’t hold. By the mid-2010s, “MVP,” “pivot,” and “iteration” had become mandatory vocabulary in pitch decks worldwide – applied to products where the minimum viable version was, by the founder’s own vision, a fundamentally different thing than the intended product. And therein lies the problem that nobody seemed to want to name.

An illustration made the rounds. You’ve probably seen it – it’s been projected in pitch meetings and tacked to coworking space walls for a decade now. Two rows of images. The top row shows the wrong way to build a car: a tire, then a chassis, then a body shell, then finally a complete vehicle – nothing usable until the very end. The bottom row shows the right way: a skateboard, then a scooter, then a bicycle, then a motorcycle, then a car. Each step is a complete, usable transportation device. The underlying need – getting from A to B faster – is met at every stage. The illustration was created around 2013 by an agile coach working at a music streaming company in Stockholm. It went viral instantly. Founders printed it, pinned it to walls, projected it in pitch meetings. It became the visual scripture of iterative product development. It contained a flaw so fundamental that its own creator spent the next several years trying to explain what he actually meant. He said it was a metaphor, not a literal prescription for building cars. The illustration assumed the customer’s real need was general mobility, not a car specifically. If you’re selling “getting around,” then yes, a skateboard is a valid starting point. If you’re selling a car – with a roof, climate control, comfort, storage, safety – a skateboard is not a stripped-down car. It is a different product entirely. Different customers. Different use cases. Different business model. Different supply chain. Different everything.

The coach himself eventually recommended abandoning the term MVP altogether, preferring three stages: the Earliest Testable Product, the Earliest Usable Product, and the Earliest Lovable Product. He acknowledged that his model worked especially well for software, because software is – his word – “soft.” You can morph it as you go. Hardware, and anything else whose value depends on physical or experiential completeness, demands rebuilding from scratch at each stage. Critics were less diplomatic. One product leader wrote that no matter how good your engineering, a functioning car almost certainly doesn’t contain the components of a skateboard. Another put it more bluntly: if you were hired to build a car and delivered a skateboard as your first working prototype, you would be fired. A self-driving car researcher captured the deepest flaw when he compared the illustration to saying that if you work hard enough at jumping, one day you’ll be able to fly. Incremental progress within one category of solution does not, at some threshold, magically become a different category of solution. A really excellent skateboard does not approach being a car. A thousand iterations of a skateboard produce a spectacular skateboard. The most devastating observation is the simplest one, and it rarely gets said aloud in startup circles: the learning you gain from skateboard users tells you almost nothing about what car buyers need. A skateboarder is seventeen, unafraid of rain, willing to fall, excited by the wind. A car buyer is forty-three, has two kids in car seats, needs climate control and crash ratings. Putting a skateboard in front of the car buyer doesn’t test whether they want a car. It tests whether they want a skateboard. When they decline, you’ve learned precisely nothing about the viability of the car.

The historical record is littered with products that died not because they were bad ideas, but because they were exposed before they were ready – and the exposure became the verdict. A major tech company unveiled what may have been the most ambitious collaboration tool ever conceived – a platform that merged email, instant messaging, wikis, and real-time co-editing into a single interface. It debuted as a developer preview at a flashy conference in 2009, and the tech press lost its collective mind. Invitations became status symbols. Then people actually opened the thing. The interface was a vast glittering dashboard of possibilities with no clear entry point – powerful the way a cockpit is powerful if nobody teaches you to fly. Within weeks, the memes started: “Got an invite. Now what?” By September, barely a quarter of invited users remained active. When the full version finally launched a year later, the world had already made up its mind. Nobody came back for a second look. The bitter irony: almost every feature in that platform eventually became the foundation of tools people now use daily. Real-time collaboration, threaded conversations, document co-editing – these ideas weren’t wrong. They were early and incomplete. The vision was correct. The launch killed it.

A software giant’s mobile operating system tells an even more expensive version of the same story. It arrived in 2010 without copy-and-paste – a feature so basic that its absence became a punchline. Third-party multitasking was missing. The app ecosystem was barren. Developers who’d spent months building for the platform found themselves selling to ghost-town storefronts. None of these were permanent limitations; every one of them was eventually fixed. The engineers working on it were talented, the design language genuinely innovative in places. None of that mattered. The initial impression created a death spiral: fewer users meant fewer developers, which meant fewer apps, which meant fewer users. At peak, the platform held barely three percent of the global market. The parent company poured billions into acquisitions trying to reverse the trajectory. Nothing worked. One executive later reflected that they’d built a beautiful car but forgot to build the roads. The metaphor is apt, though incomplete: they didn’t forget to build the roads. They opened the car dealership before the roads were paved, and customers who test-drove on dirt never came back to check whether anything had changed.

A celebrated game studio released its most anticipated title in December 2020 in such a broken state on older hardware that a major console manufacturer pulled it from its digital storefront – unprecedented in the industry’s history. Streamers loaded the game on launch night and the excitement curdled in real time: character models popping in and out of existence, vehicles vanishing mid-drive, frame rates staggering through what were supposed to be cinematic set pieces. Years of patches eventually salvaged the game’s reputation to some degree. It never recaptured the position it would have held with a polished launch. The permanent scar wasn’t the bugs themselves. It was the gap between the promise and the first experience – a gap wide enough to swallow all the goodwill the studio had spent a decade building. These are the visible cases, the ones big enough to make headlines. The invisible cases are, by definition, uncountable: the startups that shipped a sliver of their vision, got a lukewarm market response, concluded there was no demand, and shut down. Some of those ideas would have worked. We’ll never know which ones, because the shadow was tested in place of the thing, and the shadow’s failure was projected onto the thing itself. The founder walks away believing the market rejected their product. The market never saw their product. What it saw was a moving hand, a warm heart, a crawling leg joint connected to nothing – and recoiled, not because the complete human being the founder imagined wouldn’t have been compelling, but because disembodied parts are disturbing, not delightful.

Against this graveyard stands a roster of companies that succeeded precisely because they refused to ship until the experience was whole. A phone manufacturer spent roughly three years developing a device in absolute secrecy, from 2004 to its unveiling in January 2007. No public beta. No developer preview. No lean iteration with early adopters. When the product appeared, it was so fully realized – so obviously finished – that it instantly obsoleted every competitor on the market. People who held one for the first time kept trying to describe the sensation: it was the feeling of realizing you’d been tolerating something terrible without knowing it, the way you only notice how loud a room was after someone turns off the machine that’s been humming in the corner. This wasn’t luck or arrogance. It was a calculated bet that the value of the product lay in its integration: the way the hardware, software, interface, and ecosystem worked as a single experience. You could not have tested that experience by shipping a subset of it. A phone without the touchscreen keyboard wasn’t a simpler version of the product – it was a completely different, worse product. Every subsequent device this company launched followed the same pattern: years of invisible development, then a polished debut. The approach made it the most valuable company on earth. A workplace messaging tool is often celebrated as an MVP success story, but the timeline tells the opposite tale. Born as an internal communication system inside a game studio whose actual game failed in 2012, the tool was used and refined internally for years. The team lived inside it daily, experienced every friction firsthand, argued about notification sounds and threading models and the exact shade of a status indicator. A design agency was brought in to polish the interface. Friends at other companies were gradually invited to try it – not as beta testers filling out feedback forms, but as colleagues whose actual work would expose whether the thing held up under real pressure. By the time it reached the public in early 2014, after an extended preview period, it was not a minimum viable product. It was a maximum lovable one. Eight thousand people signed up on the first day. Within eighteen months, daily active users surpassed a million. The eventual acquisition price was nearly twenty-eight billion dollars. The conventional narrative treats this as “iterative development.” The reality is closer to the restaurant model: years of preparation behind closed doors, followed by a debut of the fully realized experience.

Perhaps the most instructive case involves a productivity platform whose first version was, by its founders’ own admission, a genuine MVP – and a disaster. The original product crashed constantly, built on a tech stack that couldn’t support the vision. What the co-founders did next is something most startup advice would have talked them out of. They laid off their entire team of four employees. They relocated overseas. They borrowed a hundred and fifty thousand dollars from family – from people who loved them and trusted them and had no particular reason to believe this second attempt would go differently. Then they rebuilt the product from scratch over many months of eighteen-hour days. Not iterating. Not patching. Rebuilding. The rebuilt version hit number one on its launch platform, earned national press coverage calling it the only productivity app anyone needed, and eventually reached a two-billion-dollar valuation. Their philosophy distilled to a single conviction: build a great product above all else. There’s also the inventor who built 5,127 prototypes of his bagless vacuum cleaner over five years – each one changing a single variable – before ever selling a unit to a customer. When the product reached market, it outsold competitors three to one. The iteration was extensive, even obsessive. It was entirely internal. No consumer ever saw prototype number 2,000. No one was asked to vacuum their apartment with a half-working cyclone chamber and then fill out a satisfaction survey. The inventor iterated relentlessly on the product, not on the market’s patience. An electric vehicle company took a similar approach, publicly articulating its master plan: a luxury sports car would fund development of a mid-range sedan, which would fund development of an affordable mass-market vehicle. Each stage delivered a complete, desirable product for its specific market segment – not a minimal experiment to test whether people wanted electric cars. The sports car wasn’t a skateboard. It was a sports car. People drove it and felt something. That feeling – not a survey response, not a conversion metric – was the product.

No Michelin-starred restaurant in history has iterated from a hot dog stand to fine dining. Say that again, because the tech industry needs to hear it. No chef has ever opened a food cart, validated that people like their food, upgraded to a food truck, validated that people will eat their food sitting down, opened a small sit-out, validated that people will eat their food with a roof over their heads, and then – finally, triumphantly, having proved demand at every stage – opened a multi-course tasting-menu restaurant with a wine program and tablecloths. The idea is absurd. And yet this is precisely the mental model that MVP methodology prescribes to founders whose products are, in every meaningful sense, restaurants. The path to a great restaurant follows a completely different logic. A chef trains for years in established kitchens, developing a point of view – a culinary philosophy that is theirs and nobody else’s. The way a particular chef approaches acid, or how she thinks about texture and temperature in the same bite, or what he believes a meal should leave behind in the person eating it. They raise capital from investors who believe in that vision. They secure a space. They design every element of the experience – the menu, the room, the service style, the lighting, the pacing, the wine list – as an integrated whole. The menu doesn’t exist without the room. The room doesn’t work without the service. The service doesn’t land without the pacing. The pacing doesn’t hold without the food. It is a system, and the value lives in the totality. The plate is one thought. You don’t serve half a thought.

One celebrated American chef spent nineteen months raising over a million dollars from more than sixty investors before opening his now-legendary restaurant. He’d developed his culinary philosophy through years of training with masters in France. The restaurant earned its first major national recognition within a year. In Scandinavia, a chef opened his restaurant in a reclaimed warehouse with carefully commissioned interiors – a complete concept from the first evening a guest sat down, which he then evolved artistically over the years, not incrementally. The evolution was about deepening, not about adding missing pieces. Across the Atlantic, a restaurant launched as a complete neoclassical brasserie; its subsequent transformation into one of the world’s highest-ranked dining experiences involved comprehensive reinventions of the entire experience, not piecemeal additions. Each reinvention was a new thesis, not a patch. Restaurants do have something resembling a test period: the soft opening. For one to four weeks before the grand opening, invited guests are served the full menu in the finished space with the trained staff. The energy of a soft opening is unmistakable – a kitchen running hot with nerves, servers who know the wine list cold but haven’t yet found their rhythm with the room, a pastry chef hovering behind the pass to see whether the dessert lands the way she imagined it at 2:15am on a Tuesday six months ago. This differs from an MVP in every respect that matters. The product is essentially complete: food ordered, staff trained, design finished, systems in place. The purpose is stress-testing execution under real conditions – can the kitchen produce eighty covers at pace? do the servers know the wine list? does the reservation system hold? – not testing whether people want a restaurant. No restaurateur soft-opens with paper plates to validate whether diners will pay for food before investing in ceramics.

One of the world’s most celebrated hotel brands employs a famous seven-day countdown for every new property: an intensive, orchestrated training process where staff from around the globe converge to rehearse every element of the guest experience. Every employee is empowered to spend up to two thousand dollars resolving a guest issue without managerial approval. You cannot MVP this level of empowerment. It requires a comprehensive cultural framework – values, training, hiring, authority structures – functioning as an integrated system before a single guest checks in. Ship the hotel with untrained staff and limited authority to “test demand,” and what you’ve tested is demand for a mediocre hotel, which is a different product than the one you intended to build. The distinction isn’t just practical; it’s conceptual. Business researchers distinguish between functional attributes and experiential attributes. Functional attributes are processed deliberately and piece by piece – you can evaluate a spreadsheet feature by feature, checking boxes on a list. Experiential attributes are processed holistically, as a gestalt. The warmth of a restaurant, the feeling of a luxury hotel, the arc of a great game – these cannot be decomposed into independent variables and tested individually. The whole is not merely greater than the sum of the parts. The whole is the product. Remove half the variables and you haven’t created a simpler version of the same experience. You’ve created a different experience entirely – one that might repel the very audience you’re trying to reach. One restaurateur’s formulation captures this precisely: service is the technical delivery of a product, but hospitality is how that delivery makes its recipient feel. You cannot A/B test a feeling by removing half of what produces it.

There is a deeper problem with MVP testing, one that goes beyond the practical objections and into how human minds actually process incomplete experiences. Behavioral economists identified something called the peak-end rule: people judge an experience not by its average quality but by its most intense moment and its ending. Duration barely factors in. You remember the spectacular view from the summit and the twisted ankle on the descent, not the three hours of pleasant hiking in between. An MVP, by definition, delivers a stripped-down experience where the peak moment is likely mediocre and the end is likely abrupt or buggy. Users don’t form a “partial” evaluation – they don’t think, “well, this is just the MVP, so I’ll mentally adjust upward.” They form a complete evaluation based on what they actually experienced. And that evaluation is, predictably, poor. Worse: a 2024 study found that first impressions become more heavily weighted the longer the gap between experience and evaluation. The bad impression doesn’t fade. It calcifies. Anchoring compounds this. When someone encounters a minimal product, their initial experience sets a reference point for all future evaluations. Even after dramatic improvements, users who saw the MVP version judge the enhanced product relative to their anchor, not on its own merits. This plays out with dispiriting regularity: a founder ships a design tool with janky export and one font family, then releases a polished version eight months later. The early users who suffered through version one say “it’s better” in the tone you’d use when someone finally fixes a leak you reported in January. New users who never saw the first version are enthusiastic. Same product, different anchor, different verdict. The persistence of anchoring is well-documented and alarmingly robust – it operates even when people are explicitly warned about it and financially incentivized to override it. The MVP doesn’t just produce a bad first impression. It produces a gravitational pull that warps every subsequent impression.

Now layer on the technology adoption lifecycle. The users willing to try an MVP are, by definition, innovators and early adopters – roughly the first sixteen percent of any market. These are people who tolerate bugs, enjoy rough edges, and value novelty over polish. The early majority – a full third of the market, and the segment where real business viability lives – wants what one technology strategist called “the whole product”: a complete, proven, reliable solution. Positive signals from early adopters do not predict whether the early majority will buy. The two groups want different things. They shop differently, evaluate differently, and abandon for different reasons. This means MVP testing is structurally capable of producing both false positives (enthusiastic early adopters don’t predict mainstream adoption) and false negatives (mainstream users who would love the complete product reject the incomplete version). The methodology promises to reduce risk by generating data. It does generate data. The data is systematically distorted in ways that neither founders nor investors typically account for. In luxury and premium markets, the distortion becomes almost total. Testing a cheap version of a luxury product tells you precisely nothing about demand for the luxury product. The entire value proposition – craftsmanship, exclusivity, emotional resonance, status – is exactly what the minimum version strips away. A golf cart on the highway does not test demand for a luxury sedan. It tests demand for a golf cart. And when nobody wants to drive a golf cart on the highway, the conclusion should not be “there is no demand for this type of vehicle.” One brand strategist put this plainly: the MVP route may boost short-term sales, but it erodes brand relationships built over years. It reduces a brand to a commodity. The thing you’re testing isn’t the thing you intend to sell – and the market’s response to the test object tells you nothing about its response to the intended object.

Woven into MVP dogma is an anxiety about originality – the belief that your idea must be validated because someone else might already own it. If a competitor exists, the thinking goes, you need to prove your differentiation cheaply before committing resources. This anxiety rests on a premise that most industries quietly disprove every day: the premise that markets are winner-take-all competitions for a single correct idea. Every town, even small ones, has a profusion of restaurants. The third Thai place doesn’t close because two already exist. The new Italian spot doesn’t need to “validate” that people eat Italian food. It needs to be good. And “good” is a function of execution, not idea novelty – the quality of the food, the warmth of the service, the personality of the space, the consistency of the experience over dozens of visits. Ideas are cheap. Markets accommodate many expressions of the same idea, differentiated not by concept but by execution. The academic evidence here is striking. A comprehensive analysis of roughly five hundred brands across fifty product categories found that nearly half of market pioneers fail, only about one in nine end up as long-term category leaders, and the average pioneer’s market share settles around ten percent. First movers fail at rates approaching fifty percent; late entrants who improve on the concept fail at rates closer to eight. One strategic analysis published in a leading business journal argued that first-mover advantage is one of those management concepts with such intuitive appeal that its validity is nearly taken for granted – and that the reality is far more complicated.

The most transformative products of our era confirm this. The dominant search engine was the twenty-first to market. The dominant social network followed at least three predecessors. The phone that revolutionized mobile computing arrived years after several smartphone competitors. The console that revitalized home gaming entered a market already occupied by established players. In every case, the winner entered a proven category and executed the idea better – often radically better – than anyone had before. The idea was free. The execution was everything. A founder who sold his independent music distribution company for twenty-two million dollars distilled this with memorable economy: ideas are worth nothing unless executed. They are just a multiplier. The most brilliant idea, with no execution, is worth twenty dollars. With great execution, twenty million.

The coffee industry provides a living laboratory. When a global coffee chain opens a new location, independent coffee shops in the same neighborhood don’t die – they multiply. One economic study found that neighborhoods gaining a location of the dominant chain experienced five to eighteen percent increases in new business registrations annually, sustained over seven subsequent years. Independents collectively held a larger share of the U.S. coffee market than the chain itself. In London, independent coffee shops grew by eight hundred percent in a single decade, vastly outpacing chain expansion. The chain didn’t capture a fixed pie; it expanded the pie, and independents thrived on the newly created demand. The independent down the block that roasts its own beans and plays São Paulo jazz at a volume just loud enough to notice didn’t need to validate that people drink coffee. It needed to make coffee worth crossing the street for. This matters for MVP thinking because it undermines the urgency that drives premature launches. If the idea were truly a race – first to ship wins, everyone else is foreclosed – then shipping fast would be rational even at the expense of quality. Markets don’t work that way. The twenty-first search engine won. The fourth social network won. Speed of launch matters far less than completeness of vision, and the anxiety that drives founders to ship before ready is often based on a competitive model that doesn’t match reality.

What’s remarkable is how many of the methodology’s own architects have tried to walk it back. One of the most visible investors in technology wrote that leanness is a methodology, not a goal, and that making small changes to things that already exist might lead to a local maximum but won’t help find the global maximum. He characterized the lean approach as code for “unplanned” and asked the question that should be obvious: why should you expect your business to succeed without a plan? A founder known for building profitable software companies without venture capital published a piece titled “Validation is a Mirage,” arguing that you cannot take a slice of a product, ask people how they like it, and deduce they’ll like the rest once it’s complete. The metaphor is precise: what you’re testing isn’t a smaller version of the finished thing. It’s a cross-section. And a cross-section of a human body is not a thinner human. It’s a tissue sample. It tells you something about cell structure. It tells you nothing about whether the person can dance. A respected product leader declared bluntly: an MVP should never be an actual product. The investor who literally coined the term “product-market fit” stated: you cannot customer-development your way into a massive success. Even the agile coach whose skateboard illustration became the movement’s visual gospel recommended abandoning the term MVP entirely.

The game industry, where the tension between vision and iteration is sharpest, has been particularly vocal. One analysis called the MVP approach “the kiss of death for game projects,” because a great game requires passion, artistry, and a willingness to experiment with mechanics that won’t make sense until dozens of systems interlock – qualities that cannot be squeezed from a team under the constraints of ruthless efficiency. A legendary game designer said that releasing MVPs “shows contempt for the audience.” This isn’t hyperbole. A half-built game isn’t a simpler game. It’s a broken game. And the experience of playing a broken game isn’t “this has potential” – it’s betrayal. You trusted someone with your time and money and they handed you scaffolding dressed up as architecture. The market’s response to that – frustration, abandonment, reviews that live forever on storefronts – tells you nothing about its response to the finished one. Alternative frameworks have emerged from within the lean ecosystem itself. The concept of a Minimum Lovable Product, introduced around 2013, shifted the emphasis from viability to desirability – arguing that a merely viable product is so bad it’s almost irrelevant in a market where a dozen alternatives already exist. Another framework proposed the acronym SLC: Simple, Lovable, Complete. The key insight is the third word. Each version should be complete – a small yet whole product with autonomous value. A bicycle has autonomous value. It doesn’t become worthless when someone invents a motorcycle. The progression isn’t from broken to less broken to functional. It’s from small-and-whole to larger-and-whole.

This is closer to what most industries have always practiced. A food cart is a complete product. So is a food truck. So is a sit-down restaurant. So is a Michelin-starred tasting-menu experience. They’re different businesses with different customers, different economics, and different operational demands. A food cart is not an MVP of fine dining. It’s a finished product in a different category. A founder who starts a food cart “meaning to become fine dining someday” on the same customer base will discover, painfully, that the transformation demands not iteration but reinvention – new customers, new pricing, new location, new staffing, new everything. Each leap requires starting over, not leveling up.

MVP methodology works. It works for a specific and identifiable category of products: those whose core value proposition can be isolated into a single testable feature, where the market is genuinely unknown, where the product can be improved incrementally without architectural overhaul, and where users tolerate rough edges because the utility is immediately apparent. An explainer video testing demand for a file-syncing service. A founder personally fulfilling shoe orders from her apartment to test whether people will buy footwear online. A couple renting out their spare room on an air mattress to test whether strangers will stay in each other’s homes. These are legitimate uses of lean methodology, and they worked because the minimum version preserved the essential nature of the product being tested. The shoe delivery was still a shoe delivery. The air mattress was still a place to sleep. The explainer video was, admittedly, not the product – but it was testing a clean, decomposable hypothesis: will people want to sync files across devices? The methodology breaks down – and it breaks down hard – for experiential products where quality is the value proposition. For platforms requiring ecosystem completeness before the product makes sense. For safety-critical systems where correctness is non-negotiable from day one. For luxury and premium brands where an incomplete version destroys the very thing being tested. For narrative products like games and films where the whole is irreducibly greater than the sum of parts. For any product where the customer is buying the gestalt – the totality of the experience, the feeling of completeness, the sense that someone cared enough to get every detail right – rather than a discrete function.

There is a quiet survivor’s bias at work in startup culture. We celebrate the MVPs that worked and hold them up as proof of the methodology. We never count the good ideas killed by bad MVP signals – the founders who shipped a shadow, watched the market shrug, and concluded the market didn’t want the thing. The shadow is not the thing. The market’s response to the shadow is not its response to the thing. And the number of viable products that were strangled in their cradles by premature exposure is, by definition, unknowable. We can only see the successes. The dead companies tell no tales, and their founders carry a lesson that may be the wrong lesson entirely: “I tested the market and there was no demand.” You didn’t test the market. You tested a tissue sample. The emerging wisdom – quiet, not yet popular in every circle, gaining traction among founders who’ve lived through the failure mode – is not “MVP versus vision.” It’s a recognition that different products demand different approaches. Start with a strong, opinionated picture of the future. Identify the riskiest assumptions within that picture. Then decide honestly: can this assumption be tested with a minimal version that preserves the essential character of the product? Or is the assumption itself bound up in the completeness of the experience – meaning the only honest test is the full product? For many of the most interesting, most ambitious, most category-defining products, the riskiest assumption isn’t “does anyone want this?” It’s “can we build the complete experience well enough to show what this actually is?”

The restaurant industry has always understood this. The question is whether the rest of the economy can learn to distinguish between products that can be sliced and products that must be served whole. Between a business where shipping a skateboard teaches you something about cars, and a business where shipping a skateboard teaches you only about skateboards – while the car, unbuilt, remains a ghost haunting the founder who almost brought it to life. Building the whole thing takes longer. It costs more. It requires patience that the “fail fast” culture openly disdains. It demands that founders trust their vision enough to endure the silence of a long build – no user metrics, no growth charts, no validation dopamine hits. Just the work, and the faith that the completed thing will speak for itself. If the vision is worth having, it is worth building whole. A baby born in parts is not a baby. It is an emergency. And the world is full of founders staring at Frankenstein assemblages of disconnected features, wondering why nobody loves what they’ve made – never suspecting that what they made was never, in fact, what they imagined. The thing they imagined is still in their heads, intact and beautiful and untested. What the market saw was something else entirely. The shadow is not the real product. We never quite met the real product. Maybe we should.

Featured song:

This song made me soo uncomfortable as a kid… it still does today for other reasons

Similar Essay: Dying Forward
Image Source