Almost every startup that dies, dies of the same thing: it scaled, hired, raised, and marketed before the market actually wanted what it made. Product-market fit is the line between those two worlds, and Marc Andreessen named it bluntly in 2007: it is "the only thing that matters." This is the long, horizontal version. Not a guide to one industry, but the universal shape of fit, illustrated by the companies that found it and the ones that faked it, across SaaS, marketplaces, consumer, fintech, hardware, deep tech and AI. Every claim is sourced; every example is named. If you take one idea, take this: before fit, you are pushing; after fit, the market is pulling, and your whole job until then is to get the market to pull.
The single most useful reframe in the whole topic comes from Andy Rachleff, the investor who coined the term: when your product is not working, the instinct is to add features. That almost never helps. "Adding more features doesn't turn someone into someone who's desperate," he says. "You have to change the audience." Fit is found far more often by changing the WHO than by piling onto the WHAT. Keep that in your pocket; it reappears in every chapter below.
What it is, and why it is the only thing that matters
The definition, and the lineage
Marc Andreessen popularized product-market fit in his 2007 essay "The Only Thing That Matters," and his definition is still the cleanest: "Product/market fit means being in a good market with a product that can satisfy that market." He credited Andy Rachleff (co-founder of Benchmark, later Wealthfront), who in turn credited Sequoia's Don Valentine. Valentine's whole philosophy was that the pull of a big market has to be strong enough to overcome an inexperienced team. Rachleff "put a name to that of product-market fit. Don didn't call it that."
Rachleff sharpened it into two hypotheses you prove in order. The value hypothesis is what you build (the WHAT), who desperately wants it (the WHO), and how the business works (the HOW). Product-market fit is the value hypothesis, proven. Only then do you move to the growth hypothesis: can you acquire those customers cost-effectively and repeatedly? Founders who chase the growth hypothesis first, pouring money into acquisition before the value hypothesis is real, are filling a leaky bucket. His one-line test cuts through it all: "What do you uniquely offer that people desperately want?"
Why the market wins
Andreessen ranks the three forces, team, product and market, and is unsentimental about the order: the market dominates. "In a terrible market, you can have the best product in the world and an absolutely killer team, and it doesn't matter, you're going to fail." And the flip side that founders hate to hear: the product "doesn't need to be great; it just has to basically work." A great market pulls a product out of a startup. VMware was so transformative it created the operating-system-virtualization market around itself; demand made the category, not the reverse.
One more durable truth: first to fit beats first to market. Facebook was not the first social network (Friendster, MySpace and Orkut came first), Google was not the first search engine, Apple and Intuit were not first in their categories. Being first to durable fit, and then to scale, mattered more than being first overall. The takeaway for a founder is liberating and demanding at once: you do not need to be early or perfect, you need to find the audience that is desperate, and if you have not found it, change the audience before you change the roadmap.
How you know: pull, not push
The two states Andreessen described
Andreessen's description of life before fit is so accurate it reads like a diagnosis: "customers aren't quite getting value, word of mouth isn't spreading, usage isn't growing that fast, press reviews are kind of blah, the sales cycle takes too long, and lots of deals never close." If that is your company, you do not have a marketing problem or a sales problem. You have a fit problem, and no amount of pushing fixes it.
Life after fit is just as recognizable: "customers are buying the product just as fast as you can make it, or usage is growing just as fast as you can add more servers. Money from customers is piling up in your company checking account." Slack's preview launch drew about 8,000 signup requests on day one and 15,000 within two weeks. That is the pull. You feel it; the question is only whether you have built the instrumentation to confirm it before you bet the company on it.
A moment, or a spectrum?
It is tempting to treat fit as a binary switch, and for the purpose of "do not scale before it" that framing is useful. But in practice fit is a spectrum and a moving target. You can have weak fit (some pull, fragile retention) or strong fit (relentless pull), and you can lose fit as a market shifts under you. Brian Balfour's "four fits" model is the honest extension: durable growth needs market-product fit, then product-channel fit (the product must work in the channels that can actually reach the market), then channel-model fit (the channel's economics must match your business model), then model-market fit (your model must match the market's willingness and size). Fit is not one fit. It is a chain, and the chain is what most "we have PMF" claims are missing.
How to measure it
The Sean Ellis 40 percent test
The most famous quantitative proxy comes from Sean Ellis: survey your active users with one question, "How would you feel if you could no longer use this product?" If 40 percent or more answer "very disappointed," you probably have fit. Ellis derived the threshold by comparing roughly 100 startups: those below 40 percent consistently struggled to grow, those above it tended to take off. It is a proxy, not proof, and it works best once you have a few hundred genuine users to ask, but it is the cheapest credible read you can get. Benchmarks make it concrete: Slack scored 51 percent "very disappointed" when it already had around half a million paying users.
The Superhuman engine: making the test a system
Rahul Vohra of Superhuman turned that one survey into a repeatable PMF engine (documented in a First Round essay worth reading in full). In summer 2017 Superhuman's score was just 22 percent, well below the bar. Instead of guessing, Vohra segmented the "very disappointed" users to find his high-expectation customer (a persona he called "Nicole," a busy executive processing 100 to 200 emails a day), then split the roadmap roughly 50/50: half on amplifying what fans already loved (speed, keyboard shortcuts), half on converting the "somewhat disappointed" fence-sitters by removing their specific blockers, while deliberately ignoring the users who would never love it. The score climbed to 33 percent, then to 58 percent over three quarters. The lesson generalizes: fit is not only measured, it can be engineered, by narrowing to the people who already feel the pull and widening from there.
Retention is the truth-teller
Surveys can flatter you; retention cannot. The deepest signal of fit is a cohort retention curve that flattens instead of decaying to zero, the "smile" where a stable core keeps coming back (and, in the best products, lapsed users resurrect). What "good" means depends entirely on the model, which is the whole point of the next chapter, but the principle is universal: viral reach without retention is not fit, it is a leak with good marketing. Viddy rode Facebook's Open Graph to tens of millions of users with no retention underneath; when the viral channel closed, usage collapsed and it sold for almost nothing. Traffic is not fit. The curve that flattens is.
Talking to users without lying to yourself
The qualitative half matters as much as the quant, and it is where founders fool themselves. Two tools keep you honest. Rob Fitzpatrick's The Mom Test teaches you to ask about the customer's life and past behavior, not your idea ("how do you do this today, what does it cost you" rather than "would you use my app?"), because people lie to be nice. Clayton Christensen's jobs-to-be-done reframes the question entirely: people "hire" a product to make progress on a job (his famous finding that customers were hiring a milkshake to make a boring morning commute bearable). Find the job, find the desperate customer, and the survey numbers tend to follow.
What fit looks like across business models
The same idea, twelve different shapes
This is the horizontal heart of the topic. "Demand pull" is universal, but the signal of it, and the bar you must clear, is wildly different by model. Reading fit in a marketplace the way you read it in B2B SaaS will get you killed. Here is what the pull actually looks like, model by model, with the companies that show it.
B2B SaaS: retention and expansion
In B2B SaaS, fit shows up as logos that stay and accounts that grow: high logo retention, and net revenue retention above 100 percent (existing customers spending more over time), plus customers who give references without being asked. Slack, Zoom, HubSpot and Datadog all crossed into fit when usage and seats expanded inside accounts faster than churn pulled them out. The trap is mistaking a busy free trial for fit; the signal is renewal and expansion, not signups.
Product-led / bottom-up: the aha moment
In product-led products, the user self-serves, so fit lives in activation (reaching the "aha moment" fast) and the slope of organic, bottom-up adoption. Dropbox's aha was a file syncing across devices; Figma's was two designers in the same file at once; Notion's and Calendly's and Loom's were each a single moment where the value became obvious without a salesperson. The metric is time-to-value and the share of accounts that reach the aha, then come back. When the product spreads team to team on its own, that is the pull.
Marketplaces: liquidity is the fit
Two-sided marketplaces have the hardest, slowest fit because of the chicken-and-egg problem: no buyers without sellers, no sellers without buyers. Fit here is liquidity, the probability that a listing transacts (or a search finds supply) quickly. Airbnb hand-photographed listings in New York to seed the supply side; Uber subsidized drivers to guarantee short wait times; DoorDash, Etsy and Faire each had to solve one side first in a narrow geography or category before the flywheel turned. The signal is not total users, it is match rate and repeat transactions in a specific market.
Consumer subscription: habit and churn
Consumer subscriptions live or die on retention and habit: does the product earn a recurring place in someone's week, and is monthly churn low enough that the cohort survives? Netflix, Spotify, Duolingo and Calm all show fit as durable engagement and low churn, not as install counts. Duolingo's obsession with streaks is, in PMF terms, a habit-formation machine. The bar is high because the credit card is a monthly referendum.
Consumer social and free: frequency and virality
Free, ad-supported consumer apps face the highest retention bar of all, because attention is the product and there is no payment to signal commitment. Fit means frequency (daily or near-daily use) and organic virality. Instagram, TikTok and Snap cleared it; the cautionary tale is the fade, where a product catches fire on novelty and then cannot hold the habit (BeReal is the recent example of fit-then-questions). Here, traffic without D30 retention is the classic false positive.
The rest of the spread, in brief
- E-commerce / DTC: fit is repeat purchase and CAC/LTV that works, because the product literally is the product. Warby Parker, Dollar Shave Club, Glossier and Gymshark each proved people came back, not just bought once on a launch spike.
- Fintech: the bar is trust and "the money has to work." Fit shows as retained deposits, balances and transaction frequency. Stripe (developers integrating payments in an afternoon), Wise, Revolut, Chime and Nubank all earned habitual financial use, which is stickier and harder-won than any app install.
- Developer tools / APIs: fit is developer love, measured in time-to-first-successful-call and bottom-up spread. Stripe, Twilio, MongoDB and Vercel grew because individual developers adopted them and pulled them into companies. Great docs are a fit lever, not a nicety.
- Enterprise / top-down: the signal is the fuzziest and slowest. With long sales cycles you read fit through successful pilots, internal champions, expanding ACV and reference customers long before you have mass retention data. Palantir, Snowflake and Workday show fit as land-and-expand inside accounts.
- Hardware: the cycle is long and capital is at risk, so the early fit signal is pre-orders, waitlists and crowdfunding (real money committed before shipping), then attach and repeat. Oura, Peloton, Tesla and GoPro each had demand visible before mass production, which is exactly why pre-orders matter: they de-risk the inventory bet.
- Deep tech / biotech / climate: often you have "market fit" before you have a finished product. The signal is letters of intent, design partners and milestones, not retention curves, over very long horizons. SpaceX's manifest of signed launches and Moderna's platform partnerships were fit signals years before a "product" in the consumer sense existed.
- AI products: the live frontier (2024 to 2026). The trap is the "wrapper" that wins a huge spike of curiosity and cannot keep anyone, so retention is the real test and the gap between "tried it" and "kept it" is the whole story. ChatGPT, Cursor, Perplexity and Midjourney show durable use; others rode novelty and saw engagement fade once the wow wore off. In AI especially, day-one signups are meaningless and week-four retention is everything.
Takeaway: there is one PMF, but a dozen dashboards. Before you claim it, name your model, then name the specific signal that model demands, then look at that number, not the vanity one next to it.
Verticals and edges
Going deep on one industry
Vertical SaaS finds fit by owning a narrow industry so completely that the product becomes the way that industry runs: Toast for restaurants, Procore for construction, Veeva for pharma, ServiceTitan for the trades. The fit is often stickier than horizontal software because it fits the messy specifics of one workflow and embeds payments and operations, raising switching costs. The wedge (one painful job done better than anyone) is how these companies enter, then they expand across the vertical.
Prosumer and the creator economy sit on the blurry line between consumer and business, and find fit through bottom-up love plus monetization: Canva, Substack and Figma each won individuals first, then turned that love into paid teams or paid audiences. Emerging markets reach fit by solving constraints the West already solved differently: M-Pesa (mobile money without banks), Nubank (a bank for the underbanked), Gojek and Mercado Libre (super-apps and commerce built for local payment and trust realities). And B2B2C platforms, Shopify, Stripe, Plaid, Twilio, prove fit when other businesses build their own products on top, the deepest kind of pull, because your customers' success becomes your retention.
Finding it: the path, and the pivots
The path
The route is well-worn even if the destination is uncertain: problem-solution fit first (confirm a real, painful problem and a solution people want, through customer discovery), then a deliberately small MVP, then fast iteration against real usage, until the pull appears. Steve Blank's customer-development discipline ("get out of the building") and Eric Ries's build-measure-learn loop are the operating system. The mandate before fit, per Sam Altman, is narrow: do whatever it takes to live as long as possible and iterate as fast as possible. Hiring before fit slows you down; hiring after fit speeds you up.
The pivots
A huge share of great companies found fit by pivoting, usually by changing the WHO or the use case rather than grinding on the original idea. The canonical list is a master class in not being precious about your first plan: Slack was the internal chat tool of a failed game studio (Glitch). Instagram was carved out of a cluttered check-in app (Burbn). Twitter emerged from a podcasting company (Odeo) made obsolete by iTunes. Shopify was built to sell snowboards (Snowdevil) before the team realized the commerce software was the product. YouTube started as a video dating site. The pattern: the team kept the part users actually pulled on and threw away the rest. If one narrow feature or audience shows disproportionate pull, that is not a distraction from your roadmap. That is your roadmap.
How it goes wrong: false fit, premature scaling, losing it
The ways fit lies, and the ways it leaves
The deadliest mistake is premature scaling: hiring, spending and expanding before the value hypothesis is proven. The Startup Genome study found roughly 70 percent of startups scaled prematurely on some dimension, a leading contributor to the high failure rate, and that companies need two to three times longer than founders expect to validate their market. Viddy is the monument to this: enormous viral reach, no retention, near-zero exit.
The other failure modes rhyme. False or fragile fit: a vanity metric (signups, page views, a launch spike) that masks weak retention. Building for everyone: trying to please a broad lukewarm audience instead of a narrow desperate one (the opposite of Rachleff's rule). Founder denial: explaining away the flat retention curve. And finally, losing fit at scale: a market shifts, a platform changes, a better option appears, and a company that had fit no longer does. Fit is not a trophy you win once; it is a position you have to keep defending, and the "second product" problem (finding fit again for the next product) is its own hard, separate fight. The discipline is the same throughout: trust the retention curve over the story you want to tell.
From fit to growth: when to step on the gas
The handoff, and the checklist
Fit is the permission to scale, not scale itself. Once the value hypothesis is proven, the job shifts to the growth hypothesis: finding repeatable, cost-effective acquisition, which is product-channel fit (the product has to be built to spread through a channel that can actually reach the market) and then the go-to-market and sales machinery. That handoff is its own discipline; we cover it in EquityFlow's go-to-market and sales deep dives. The danger at this seam is scaling a channel before product-channel fit, or hiring a sales team before founder-led selling has shown the motion works.
The practical checklist, stripped down:
- Name your model and its signal. Retention for subscription and AI, liquidity and match rate for marketplaces, NRR for B2B SaaS, repeat purchase for DTC, pilots and expansion for enterprise, pre-orders for hardware, LOIs for deep tech. Watch that number.
- Run the 40 percent test once you have a few hundred real users, and look at the cohort retention curve. If it does not flatten, you do not have fit, no matter what the survey says.
- If you are stuck, change the WHO before the WHAT. Narrow to the most desperate audience; serve them completely before you widen.
- Do not scale before the curve flattens. Before fit: live long, iterate fast, stay small. After fit: hire and spend behind the pull, fast.
- Keep talking to users honestly (the Mom Test), and trust retention over your own story.
Product-market fit is not a tactic, a metric or a slogan, it is the moment the market starts doing your selling for you. Everything before it is the search; everything after it is the build. Find the people who are desperate, prove they keep coming back, and only then pour fuel on the fire. This article is general guidance with sourced examples, dated to 2026; the named companies and numbers are illustrative, not investment advice.