Tales of startups that have raised staggering sums with little more than marketing hype have become part of entrepreneurial lore. Consider WeWork, which secured billions of dollars from top-tier investors on the strength of audacious claims about its plans to "elevate the world's consciousness" through the "reinvention of work." The founders insisted it wasn't simply another real estate venture and deserved a technology-style valuation because of its proprietary WeOS. That argument succeeded — until the company suddenly came crashing down.
Headline-grabbing stories like that point to a deeper and more pervasive rot in startup culture: the belief that the rules of business — especially the need for a viable business model — somehow don't apply to entrepreneurial ventures. Yet even companies with enviable top-line growth often collapse under the weight of their own economics. In 2025 we studied more than 1,000 startups that had achieved unicorn status over the past decade and found that fewer than 150 had become prosperous public companies. In many cases the problem wasn't bad luck — it was bad advice.
Over the past decade, we've studied high-growth entrepreneurship in depth, conducting more than 250 interviews with founders, investors, and board members across consumer marketplaces, online fashion, drones, clean technologies, financial services, and edtech. Although our insights emerged from studying venture-backed startups, we've seen the same pattern — and the same mistakes — play out in small and midsize firms and in large companies launching new businesses.
Paradoxically, we've realized, the key to long-term, profitable growth across all these contexts isn't chasing growth. It's the opposite. Enduring growth is usually the result of hewing to a business fundamental: solid unit economics. That often requires intentionally slowing down, taking the time to build the right foundation, and only then scaling up. In contrast with the dominant "growth first" approach, we call this approach "foundation first."
The Growth Trap
Why does chasing growth sabotage startups' long-term success? Because in their rush to reach scale, entrepreneurs often settle on flawed business models with weak unit economics. They may find a path to sales growth or product-market fit. Yet since they neglect the underlying economics, their businesses may never be viable.
Technically speaking, unit economics are the incremental revenues and costs associated with each additional sale. They go beyond traditional accounting margins to consider a variety of per-unit costs, including customer acquisition and customer support. When unit economics are ignored, startups often focus on expanding the market for a product that customers love — but do so at a price below what the product costs to produce and sell. Take MoviePass, a company that offered subscribers virtually unlimited movie tickets for a flat monthly fee. It attracted millions of customers. The downside? The firm lost money on nearly every transaction. Despite rapid growth, there was no viable path to profitable unit economics. Customers would pay only so much, and theaters would supply tickets only at a certain price. The company ended up filing for bankruptcy. (It was later bought by one of its cofounders, who substantially reworked the business model.) MoviePass illustrates what Warren Buffett had in mind when he wrote, "The worst sort of business is one that grows rapidly, requires significant capital to engender the growth, and then earns little or no money."
In their defense the executives and investors behind MoviePass may have assumed that they could fix poor unit economics later — by raising prices or cutting costly services. But this "grow and then gouge" strategy rarely pans out, because product-market fit and unit economics are deeply intertwined. We can think of very few startups that were able to substantially raise prices or redesign a costly business model after they began to grow. Caught up in the thrill of rapidly rising sales, founders may not even have a plan for fixing unit economics. Backed by venture investors who are willing to fund loss-producing companies, many forget about business fundamentals.
Founders also sometimes assume that network effects (when a product or service becomes more valuable to its users as the number of customers using it increases) will save a growth-first business. Once the business grows very large — the thinking goes — it will become an essential service. While that logic is alluring and might work for some companies, it often rests on a misunderstanding of unit economics. Consider a peer-to-peer marketplace for kids' secondhand clothing that we studied. Parents loved its ease and convenience. Not surprisingly, people kept joining it, and powerful network effects drove increasingly attractive revenue growth. But its founders forgot about their high unit costs for each transaction, and parents simply would not pay more for children's clothing. Instead of solving the startup's challenges, the powerful network effects just accelerated its losses. As one investor told us, "Growth solves many problems, but unit economics isn't one of them."
To be sure, there are examples of loss-producing startups that have gone on to financial success: Amazon and Airbnb burned through enormous sums before becoming profitable. Yet such examples miss the critical distinction between unit economics and firm-level profitability. Amazon is legendary for its close attention to unit economics, especially costs. Transactions in its initial core retail business were efficient and simple, and its prices covered the costs of such elements as payments, warehousing, and shipping. Amazon chose to be unprofitable at the firm level in order to invest in new businesses like Amazon Web Services. Over the years the company deployed teams of economists to evaluate the unit economics of potential new businesses. Other companies have followed this playbook too, sacrificing firm-level profitability by reinvesting profits, especially in fast-growing markets, but retaining a strong foundation with a vigilant focus on unit economics.
The Foundation-First Startup
So what does a foundation-first startup look like? An excellent example is Rent the Runway. This firm was launched by Jennifer Hyman and Jennifer Fleiss to rent out designer dresses for special events like weddings, galas, and proms. Early discussions with fashion designers helped Hyman and Fleiss assess the feasibility of securing wholesale pricing, which substantially reduced their unit costs. In parallel they also ran a series of experiments to test demand and pricing, obtaining 100 dresses and holding rental pop-ups targeting sororities and social clubs. Those moves helped them understand the drivers of economics like unit revenues (product-market fit, willingness to pay) and unit costs (purchasing dresses, cleaning costs, the number of times each outfit could be rented before replacement). As they slowly rolled out their service to early users, they gained further insight into customer acquisition costs. All of this meant that when Rent the Runway was ready to accelerate its growth, its business model was already on the path to unit profitability.
Another telling example is Starbucks. Rather than rushing into rapid expansion early on, Howard Schultz and his team meticulously analyzed the potential profitability of each possible store location before moving forward. In other words, unit profitability mattered. As Schultz noted in his memoir, "Starbucks' ability to build and operate profitable stores had succeeded for years because we had adhered to a simple yet ambitious economic model: a sales-to-investment ratio of 2 to 1. During a Starbucks store's first year in business, it needed to bring in $2 for every $1 invested to build it." While that exact ratio is specific to the company's business model and industry, the disciplined approach to unit profitability helped Starbucks avoid the trap of geographic expansion at the expense of store-level economics.
In sum, unit economics are not just metrics — they're a discipline. Founders who put this discipline into practice follow three major tenets: They experiment broadly, simplify ruthlessly, and deliberately delay growth.
Experiment Broadly to Understand Deeply
Founders are taught — correctly — to run rigorous experiments to size up demand, pricing, and willingness to pay. What's missing is the same rigor on the cost side. The foundation-first approach treats cost structure not as something to optimize later but as a set of hypotheses to be explored early on, alongside revenue potential.
In practice this means conducting tests across the entire business model. Which activities truly require human labor, and which can be simplified, automated, or delegated to artificial intelligence? Which design choices quietly lock in high variable costs? And which cost drivers improve — or deteriorate — as volume increases? Most founders already know they need to run tests to assess minimum viable products and potential pivots, but their experimentation is focused almost exclusively on demand and growth. Foundation-first entrepreneurs instead probe both revenue and cost drivers with an eye on unit economics. This broader approach is slower and more laborious, but it typically yields a simpler, more profitable business model with real product-market fit and a clearer strategy. By experimenting across many elements of the business — including customer acquisition channels, service levels, operations, and target segments — founders uncover not only what works but also how different levers interact. These interdependencies between revenues and costs are critical: They reveal which changes improve unit economics and which quietly erode them. Once entrepreneurs get these insights right, scaling up becomes less about discovery and more about replicating a well-designed model.
Consider a prosperous online fashion venture we'll call by the pseudonym StyleCo. The founders began with a vision of bundling personal styling services with e-commerce. One of them described the first year simply: "It was just about proving this works. Do the economics work?" The founders tested whether algorithmic tools could replace some human stylists, lowering costs. They tried different brands and discovered that boutique labels, which could be bought at wholesale prices, still resonated with customers. They experimented with an up-front styling fee, credited toward any purchase, which both reduced theft and unexpectedly attracted higher-value customers. They even learned that full retail pricing did not deter buyers who valued convenience — but did attract more-desirable suppliers. Through this series of experiments, StyleCo discovered powerful interdependencies and reshaped its business model. Growth then became a matter of repeating what already worked and culminated in a successful initial public offering.
A near twin, FashionCo, launched around the same time with a similar concept and price point but took a growth-first path. The founders aggressively pursued sales and poured money into customer acquisition. They quickly found a product-market fit, but almost all their experimentation was aimed at boosting revenues. As the business grew, they layered on expensive features — such as commission-based stylists and a brick-and-mortar channel — without testing lower-cost alternatives or understanding how their choices affected overall economics. As one of them later admitted, "We didn't know if we could ever bring down our costs, but our customers loved their shopping experience." Revenues rose, but costs rose just as fast. FashionCo's weak unit economics never recovered, and the business model ultimately failed.
Simplify to Scale
Broad experimentation also pushes ventures toward simplification, not just toward improved unit economics, product-market fit, and strategic clarity. Simpler business models are easier and cheaper to scale, have fewer failure points, and are more intuitive for customers and employees alike. Experimenting widely forces entrepreneurs to stay close to the details of their business long enough to become genuine experts — able to see patterns, think strategically, and strip the model down to what really creates value.
Instagram is a vivid example. It began as Burbn, an overloaded app combining check-ins, photo sharing, and rewards. Through experimentation, the founders realized that while everything "worked," photo sharing was what users truly cared about. They ruthlessly cut the rest and then simplified the core action to just three taps. That clarity of focus turned a cluttered product into a breakout hit.
Foundation First in the Age of AI
Artificial intelligence (particularly generative AI) creates new urgency for foundation-first thinking. Consider a major U.S. service provider that recently explored using AI to handle customer service interactions. Rather than racing to deploy the technology, it took a foundation-first approach — running careful experiments to understand both unit economics and service quality before inserting AI into the workflow and then scaling.
The experiments revealed unexpected insights. For high-reliability tasks like offering customers concrete options to choose from, traditional code outperformed AI. But with conversational elements, gen AI excelled. Interweaving both delivered the best experience and enabled rapid responses when call volumes spiked. And while AI was expensive per interaction, it was still substantially cheaper than human representatives. Only after establishing these unit economics and assessing the customer experience did the company begin exploring broader deployment.
Several AI dynamics make building a foundation first especially critical:
Unit costs accumulate fast. Large language models charge for every word (or more accurately, token) processed and generated. Video, images, and especially agents become exponentially more expensive. These costs compound quickly across millions of interactions.
Broad experimentation reveals crucial trade-offs. For any AI use case, critical choices — which model to use, how to structure relevant context for AI, when to combine AI with traditional code or humans — have interdependent effects on quality, speed, and costs. Those effects are hard to unpack without experimentation.
Simplification is powerful — and often necessary. AI performance can degrade quickly with complexity. If instructions for AI are too long or too many tools are available, so-called "context rot" happens. AI agents lose focus, get confused, and often fail to follow instructions. Complexity also harms humans in the loop who are likely to be overwhelmed by AI output and miss where their judgment actually matters.
The message is clear: AI's complexity, power, and cost structure make it especially foolhardy to chase growth before understanding unit economics, interdependencies, and usage patterns. The transformative potential of AI makes slowing down feel risky, but scaling a fundamentally unprofitable business model is far riskier.
Businesspeople tend to solve problems by adding features, channels, and options, which is exactly what we saw happen at FashionCo. A foundation-first approach counteracts that bias by making subtraction part of the design. StyleCo did this when it experimented with the number of items to ship to each customer. The initial plan called for 10 pieces, but testing revealed that five performed better. Halving the number of items reduced shipping and inventory costs and, counterintuitively, increased sales — customers bought more when choices felt curated rather than overwhelming.
A key question is: When should entrepreneurs stop broad experimentation? In our experience startups benefit from it until they develop a relatively complete business model with promising unit economics, including product-market fit. That can require six to 12 additional months after the first indication of a favorable opportunity.
Go Slow to Go Far
Perhaps the most counterintuitive aspect of building for long-term, profitable growth is actively delaying it. This means deliberately holding back on adding revenues and customers while broad experimentation continues. The reason is simple: Growth creates immediate operational demands — hiring, infrastructure, customer support — that soak up attention and capital. It's far easier to refine unit economics when you're not simultaneously scrambling to build capacity. Staying small a bit longer preserves the ability to spot, implement, and unwind changes before complexity and customer expectations harden into constraints.
One tactic for delaying growth is to throttle customer acquisition — limiting sales, hiring, and marketing spending to "just enough" to get traction and validate product-market fit, while freeing time and money to experiment with both revenues and costs. Klaviyo, a customer data and AI marketing platform, did exactly this. Its two cofounders iterated for three years before raising significant capital. One focused on the product, the other on early customers. In the first year alone, they pivoted from an onboarding tool to a data-integration platform to an email-marketing platform. Even as sales picked up, they intentionally avoided building a sales team, keeping customer acquisition modest. That restraint gave them space to test and tune their business model and unit economics. Only once the economics — including customer acquisition costs — were highly attractive did they lean into growth and raise major funding, ultimately going public in 2023.
A second tactic is using waiting lists, which make sense when demand runs ahead of capacity. Midjourney, the AI image-generation company, used a closed waiting list even as interest in its offerings surged. By capping the number of users, the team could concentrate on improving the offering rather than on growth-induced problems. As a result, the venture never needed to raise outside capital, despite building a business now believed to generate hundreds of millions in annual revenues.
Some founders go further. Notion, a $10 billion productivity platform, initially raised about $2 million to create a no-code app-building product. After burning through much of that cash and nearing failure, the founders chose not to chase more customers or funding. Instead, they laid off most of the team and completely rebuilt the product using foundation-first principles. That pause let them reimagine Notion as an intuitive workspace for documents, databases, and projects. By delaying growth and returning to broad experimentation, the founders created a product that now serves more than 100 million users.
Delaying growth, of course, runs against the grain of many investors. Most are happy to fund unprofitable startups — until, abruptly, they aren't. A deliberate pause can complicate early efforts to raise capital, especially Series A funding, because it demands investors who value discipline over top-line growth and founders who can stretch cash. But it can also reduce the total capital required from investors. Above all, actively delaying growth raises the odds that when founders do scale a business, they're building on a solid foundation rather than speeding toward a cliff.
Ventures Beyond Startups
Regardless of organizational context, the foundation-first approach can help new enterprises succeed. After all, small and medium-size businesses (SMBs) launching new offerings and large companies incubating ventures face the same fundamental risk as startups: scaling before unit economics are understood.
At large companies, corporate initiatives often are justified by potential synergies with the core business and enjoy enviable advantages — existing customers, a sales force, brand recognition, and a parent willing to absorb early losses. Yet many still fail — for the same reasons that startups do: weak unit economics and fuzzy product-market fit, compounded by wishful thinking about synergies.
A foundation-first approach forces corporate entrepreneurs to validate those synergies instead of just assuming that they exist. Which assets actually lower customer acquisition costs? Where does brand trust really translate into pricing power? Which internal capabilities truly improve risk or operational efficiency? Answering those questions early, before scaling, can prevent billions in write-offs.
Goldman Sachs's consumer venture Marcus is a case in point. It began with a strong thesis: Use Goldman's risk-management expertise to offer unsecured personal loans, then expand into credit cards and high-yield savings. The product resonated with customers, and the business grew quickly. But on the personal-lending side, the anticipated synergies never translated into solid unit economics. Marcus's loss rate for these products remained far higher than that of its peers, and the business ultimately racked up roughly $3 billion in losses before Goldman pulled back on personal lending, keeping only the savings business. Marcus shows how even within a world-class institution, growth without proven unit profitability is fragile.
A foundation-first approach can also propel a small business into a highly profitable medium-size business — what is termed the "mighty middle." If a company lacks the high revenue potential that attracts VC investors, strong unit economics are necessary not just for survival but to generate cash to invest in growth. For SMBs, the danger is often not rapid growth around poor unit economics — but settling into poor unit economics that never enable growth in the first place.
. . .
The gospel of "growth at all costs" has dominated startup thinking for the past two decades. Our research suggests that it's time to reconsider this orthodoxy. The foundation-first model — focusing on unit profitability, broad experimentation, and delayed growth — offers a more reliable path for most ventures, whether they're VC-backed startups, new corporate businesses, or SMBs. Pursuing growth at all costs may have worked for a few outlier successes, but for too many others it's a trap. The future of entrepreneurship lies in the recognition that the key to long-term, profitable growth isn't chasing growth. In fact, it's just the opposite — intentionally slowing down, building the right foundation, and only then scaling aggressively.
Is Foundation First Always the Right Choice?
The foundation-first approach works particularly well for companies selling physical products and for business models with significant variable costs (for example, inventory) and operational complexities (for example, logistics) that make unit economics challenging. Yet even products with low variable costs and little operational complexity still require learning about product-market fit and often coping with some major variable costs (for example, customer acquisition) that make unit economics relevant. Building the foundation first also works for business models with potential network effects and economies of scale. Why? Network effects cannot compensate for negative unit profitability, while the timing and size of scale economies can be difficult to predict. Rather than aiming for illusory first-mover advantages, developing a simpler, better business model with real product-market fit usually makes more sense, as successful late movers like Google in search and Apple in smartphones illustrate.
Some point to startups backed by investors willing to subsidize massive, prolonged losses as proof that the foundation-first approach isn't universal. But these startups are like adult kids living with their parents: The rules feel different until the parents cut them off.
Deep tech startups, however, do have different rules. These ventures make long-term and expensive bets on creating pathbreaking technologies that often address societal issues like climate change or that open entire new industries like space exploration. Raising massive funds from many sources — governments, corporate strategic investors, venture capitalists, sovereign wealth funds, and even philanthropists — is usually essential. Working with governments — as regulators, customers, and investors — also is often key. But the foundation-first approach will come into play when these startups transition from scientific discovery to commercial products.
Copyright 2026 Harvard Business School Publishing Corporation. Distributed by The New York Times Syndicate.
