This is a teaching corpus, not one person's theory: partners who have sat with thousands of companies and alumni founders who lived the specific failure they are describing. What they converge on is a tight loop, ship something embarrassing to a customer who is desperate, charge them, and let one weekly number tell you what to do next. Where the lectures contradict each other, which is on pivoting, on data versus conversation, on competitors, on co-founders, and on how to set the first price, the contradiction is written out below instead of smoothed into a single house line.

Do

Ship something embarrassing within weeks. Airbnb's first version had no in-app payments and no map view, Twitch launched as one person's low-resolution personal livestream with zero games, and early Stripe had no bank deals and its founders personally flew out to integrate each new user. Paul Graham's threshold is a quantum of utility: ship as soon as one real person is genuinely glad the thing exists, rather than waiting until it stops being humiliating. [Source: "Michael Seibel - How to Plan an MVP"] [Source: "A Conversation with Paul Graham - Moderated by Geoff Ralston"]

Run your own sales and your own user interviews, personally. Alströmer's framing is that talking to customers and doing sales are the same skill, so startups that wait for a good product to sell itself almost never take off. At Clever, one of three co-founders peeled off to sell almost full time starting before the product existed. Migicovsky names talking to users as one of only two things YC insists a founder cannot outsource. [Source: "How to Get Your First Customers | Startup School"] [Source: "How to Sell by Tyler Bosmeny"] [Source: "Eric Migicovsky - How to Talk to Users"]

Extract facts about what already happened, never opinions about the future. The five reliable questions all point backward: what is the hardest part, tell me about the last time you hit this, why was that hard, what have you already tried, what do you dislike about what you tried. Escher Reality mistook enthusiasm from marketers for demand, and the polite excitement evaporated the moment they asked directly whether anyone would pay. [Source: "Eric Migicovsky - How to Talk to Users"] [Source: "Startup Experts Reveal Their Favorite Pivot Stories"]

Charge from day one, and charge more than feels comfortable. Two Segment engineers nervously started charging $10 a month and heard from customers that they wished it cost more, since a cheap vendor seemed less trustworthy with their data; a later enterprise quote of $120,000 a year, even after being negotiated down, landed roughly 150 times the original price. Koomen quoted $10,000 a month, got negotiated to $2,000, and the deal still closed. Bosmeny's rule is that if a sale closes easily, roughly double the price on the next prospect. [Source: "Startup Business Models and Pricing | Startup School"] [Source: "Enterprise Sales | Startup School"] [Source: "How to Sell by Tyler Bosmeny"]

Pick one primary metric that lags real value, and set a weekly target on it. Cheung's four tests are that it quantifies value delivered, captures recurring rather than one-time value, is a lagging indicator (revenue received, not signups that might convert), and is fast enough to inform this week's decisions. Growth among recent demo-day companies clusters at five to ten percent week over week. Doshi's warning from the other end is that even thousand-person companies over-measure, and twenty-five metrics split across teams produces paralysis rather than clarity. [Source: "Adora Cheung - How to Set KPIs and Goals"] [Source: "Suhail Doshi - How to Measure Your Product"]

Prove retention with a cohort curve before you do any growth work. Chart repeat usage of your one value metric over time: a curve still declining toward zero at twenty-four months is not product-market fit, while one that flattens at a real floor is, and Alströmer's benchmark is Netflix retaining roughly seventy percent of payers after a year. Doshi names the failure pattern the shark fin, a viral acquisition spike followed by churn outpacing new signups, after which reactivation campaigns are reliably weak. [Source: "Gustaf Alstromer - How to Get Users and Grow"] [Source: "Suhail Doshi - How to Measure Your Product"]

Calculate default alive off your own bank statements, weekly. Three numbers taken straight from the statements (balance, money in, money out) give burn, runway and growth rate without a bookkeeper or software, and the question is binary: at your current growth rate, do you reach profitability before the cash runs out if no further money ever arrives. Justin.tv ran the exercise with about $500,000 left against $750,000 a month of revenue and $1 million a month of expenses, told the whole company it might die, and reached breakeven in roughly two months. [Source: "Kirsty Nathoo - Managing Startup Finances"] [Source: "Save Your Startup During an Economic Downturn"]

Grade every task on impact and complexity, and always take high impact and low complexity first. Real progress is narrowly defined as anything that grows the primary metric, which in practice reduces to talking to users and building product. Low-impact, high-complexity work should essentially never be started, because there is no path where finishing it was worth the week. [Source: "Adora Cheung - How to Prioritize Your Time"]

Choose a co-founder for how they handle stress, and put the equity and vesting in writing before anything goes wrong. Seibel says he has never seen a company die because a co-founder was excellent but lacked the right skills, and has seen many die because a co-founder became unbearable. The structural advice is to get as close to equal as possible while avoiding a literal 50/50 deadlock, giving the CEO one extra tiebreaker share agreed in writing in advance. [Source: "Co-Founder Mistakes That Kill Companies & How To Avoid Them"] [Source: "How to Find the Right Co-founder"]

Don't

Don't spend money to look professional before you have fit. The recurring shape is buying what a bigger company buys: a FAANG-caliber hire whose old productivity depended on infrastructure you do not have, PR agency retainers running $10,000 to $50,000 a month that produce generic pitches from junior staff, event sponsorships as cargo culting, $50,000 burned having a first-year associate learn employment law on your dime, and equity-hungry advisors, one of whom dropped an ask from thirty percent to one percent simply by being challenged. [Source: "Top Ways Startups Waste Money"]

Don't hire your way to product-market fit. Nathoo calls scaling headcount before fit the single mistake there is no coming back from: more salespeople and engineers do not create fit, and weak sales or a thin feature set are product problems rather than staffing problems. Every hire also costs roughly twenty-five to fifty percent more than salary alone once equipment, benefits and space are counted. [Source: "Kirsty Nathoo - Managing Startup Finances"]

Don't spend your time on whoever is easiest to reach. Koomen names this as the major founder anti-pattern, because easy contacts supply plenty of feedback and never become customers, producing an illusion of progress. Seibel is blunter: founder friends, YC peers and investors are close to useless for product feedback because they almost never have the problem and will lead you astray out of good intentions. Friends and family are too kind to be worth much either. [Source: "Enterprise Sales | Startup School"] [Source: "Michael Seibel - Building Product"] [Source: "Kat Mañalac - How to Launch (Again and Again)"]

Don't read a launch spike as fit. Every major feature at Weebly produced a burst of signups from TechCrunch, Newsweek or Time that decayed back to the prior baseline, for a year and a half. The real inflection only showed up when daily new users stopped decaying and started compounding, which no press hit ever caused. [Source: "David Rusenko - How To Find Product Market Fit"]

Don't offer a long free trial. Bosmeny says he has never used one at any company and prefers an annual contract with a thirty-day no-penalty opt-out, since a free trial secures neither commitment nor revenue. Alströmer's version for B2B is a money-back guarantee or month-to-month opt-out instead, and a prospect's refusal to pay anything is a signal to find a better-fit customer rather than to lower the price. Seibel adds that free products mostly attract casual users who are not the desperate customers you need to learn from. [Source: "How to Sell by Tyler Bosmeny"] [Source: "How to Get Your First Customers | Startup School"] [Source: "Michael Seibel - Building Product"]

Don't build a custom feature for a prospect who has not committed. The line "we will buy if you just add X" is frequently a soft pass in disguise; the correct handling is to build it only after they commit, or once several separate prospects have asked for the same thing. [Source: "How to Sell by Tyler Bosmeny"]

Don't treat implementation as the customer's job. Optimizely closed six-figure deals where the customer never ran a single test, because nobody on the vendor side helped get it installed, and the fix was a shared implementation roadmap project-managed as seriously as an internal priority. Onboarding is also the funnel step founders most often forget entirely, which is how a successful close still ends in a quiet churn. [Source: "Enterprise Sales | Startup School"] [Source: "How to Get Your First Customers | Startup School"]

Don't let work that does not move the metric count as progress. Conferences, awards and optimizing the wrong number are named as fake progress specifically because they feel productive; low-value work creeps into a week unnoticed because it is the easiest thing available and gives the satisfaction of checking something off a list. [Source: "Adora Cheung - How to Prioritize Your Time"]

Where they disagree

When is changing the idea discipline, and when is it flinching? Michael Seibel's rule is to hold the customer and the problem tightly while holding the solution loosely, treating iteration as the normal mode and a pivot as rare, and he says teams that read two months of slow progress as grounds to pivot are moving far too fast, since real fit realistically takes closer to two years. Rusenko's own timeline backs him: Weebly took about eighteen months of flat or declining growth and roughly four years overall. Against them stands a lecture that is nothing but counterexamples. Brex abandoned a virtual reality headset within three to four weeks once conversations with real hardware experts made clear the founders were not equipped to build it, and went back to the payments expertise they already had. Tom Blomfield's bill-splitting app became GoCardless after two weeks of cold-calling UK sports club treasurers at three and four in the morning produced exactly one new user. Clipboard Health spent years in the idea maze before direct hospital selling revealed that the acute pain was last-minute nurse absences rather than nurse hiring. The corpus supplies its own tiebreaker, and it is not time elapsed: what separates a healthy pivot from pivot hell is having a single primary number. Creative Market spent about a year and a half after YC running unrepeatable monetization experiments without ever noticing they were not working, precisely because there was no daily figure to watch. [Source: "Michael Seibel - Building Product"] [Source: "David Rusenko - How To Find Product Market Fit"] [Source: "Startup Experts Reveal Their Favorite Pivot Stories"]

Trust the experiment, or trust the conversation? Alströmer's case for A/B testing is that intuition reliably loses once a company is past its most obvious decisions: he walks an audience through real Airbnb experiments, a redesigned share sheet that lifted shares forty percent and a button copy change that lifted signups fourteen percent, and a majority of the room guesses wrong every time. Against that, Doshi says that under fifty users you should skip formal analytics entirely and just talk to people, and describes personally chatting with all ten to twelve of Mixpanel's earliest customers over instant messenger to decide between a vertical and a horizontal funnel visualization. Collison calls pre-fit metrics relatively unhelpful next to direct granular inspection, and early Stripe routed every single API request to an internal email with a high-priority alert on any error so someone could follow up within fifteen minutes, sometimes about the user's own typo. Migicovsky's fix is to collect phone numbers at signup so that when aggregate data is confusing you can call a specific human. What decides it is volume, and it is worth noticing that every one of Alströmer's examples comes from Airbnb or Facebook at a scale where a single percentage point is a real number. [Source: "Gustaf Alstromer - How to Get Users and Grow"] [Source: "Suhail Doshi - How to Measure Your Product"] [Source: "Running Your Company by Patrick Collison"] [Source: "Eric Migicovsky - How to Talk to Users"]

Are competitors proof of demand, or proof of a trap? Jared Friedman argues that having competitors is usually a good sign rather than a red flag, pointing at Dropbox launching as roughly the twentieth cloud storage company into a market where nobody had solved the actual user experience problem. Paul Graham goes further, comparing startup competition to two small planes flying blind through clouds, the market space large enough that collisions are unlikely, so the correct response to news of a competitor is to keep executing. But Friedman sits on both sides of this inside a single lecture: his tar-pit category is precisely an idea with many prior attempts and a hidden structural reason it has stayed unsolved for years, with yet another app for coordinating plans with friends as the canonical example, and Dalton Caldwell says a team with no technical co-founder working on a known tar pit has the lowest odds of anything YC sees. Blomfield's first company was exactly that bill-splitting tar pit. The distinction the corpus actually draws is not how many competitors exist but why the earlier attempts failed: a field of bad user experiences is an opening, a field of structurally identical failures is not. [Source: "How to Get and Evaluate Startup Ideas | Startup School"] [Source: "A Conversation with Paul Graham - Moderated by Geoff Ralston"] [Source: "How to Apply And Succeed at Y Combinator | Startup School"] [Source: "Tom Blomfield: How I Created Two Billion-Dollar Fintech Startups"]

Do you actually need a co-founder? Harj Taggar's answer is close to yes: Apple, Facebook, Google and Microsoft all started with co-founders, and solo working is conditional on extremely high personal conviction (usually from living the problem) plus being technical enough to build a first version alone. Graham makes the harder version of the case, that being solo is unusually difficult not because of the workload but because there is nobody to keep morale up during the many moments when the evidence genuinely suggests the idea will not work. Against that, Caldwell and Seibel argue that what founders are usually told they lack is not a person at all: Google, Nvidia, Microsoft, Stripe and Dropbox all had all-technical founding teams, and when an investor says you need a business co-founder that is typically a comment on thin go-to-market thinking or a rough pitch rather than a structural gap. Drew Houston is the case both sides can claim, rejected by YC for having no co-founder, who kept building alone, found Arash, and was funded on the second application. What the corpus does settle is the ordering: the missing skill is learnable on the job, the missing trust is not. [Source: "How to Find the Right Co-founder"] [Source: "A Conversation with Paul Graham - Moderated by Geoff Ralston"] [Source: "Do Technical Founders Need Business Co-Founders?"] [Source: "Co-Founder Mistakes That Kill Companies & How To Avoid Them"]

How do you set the first price? The B2B pricing method is to build a written value equation with your champion inside the customer, quantifying cost savings, time savings or revenue impact in numbers they can defend to their own CFO, then price at roughly a quarter to a half of it, keeping the majority of the value with the buyer. The worked example is a hundred-person support team costing $10 million fully loaded, where cutting twenty percent of query time is $2 million of value and the vendor charges around $700,000. Against that method, Graham says to just guess from domain knowledge or ask trusted early users directly what they would pay, on the reasoning that prices can always be lowered without complaint and raised later while grandfathering existing users in. Bosmeny agrees it is informed guessing and says Clever's original $100,000 contract price started as a guess and was iterated deal by deal. What separates the two is whether there is a champion inside a buying organization who has to justify the number upward; with no such person, there is nobody for the equation to persuade, and guessing then raising is the only instrument left. [Source: "How To Price For B2B | Startup School"] [Source: "A Conversation with Paul Graham - Moderated by Geoff Ralston"] [Source: "How to Sell by Tyler Bosmeny"]

The one line

Put something embarrassing in front of a customer desperate enough to pay for it this month, then let a single weekly number, not your own optimism, decide what you build next.