How to build a startup.
Start with the decisions.

Most startups die of self-inflicted wounds, not competition

Building a startup is a different job from building a product. The product question is "does this work". The startup question is "can this keep going long enough to matter", and the answers that kill companies are usually structural decisions made in the first month, casually, by people who assumed they could revisit them later. Equity splits, co-founder arrangements, what you spend before revenue, what you promise a first customer.

The order matters. There's no point optimising a funnel for a product nobody wants, and no point picking a legal structure before you know whether there's a business. But there is also a set of decisions that get dramatically more expensive to change the longer you leave them, and those deserve attention early even though they feel premature.

How to build a startup, step by step

Going from idea to startup is not one step, and treating it as one is why so many attempts stall at the same place. An idea is a hypothesis about someone's problem. A startup is a repeatable way to solve that problem and get paid for it. Between the two sits a sequence, and it has an order most founders discover backwards.

1. Pressure-test the idea before anything else. Can you name the buyer without using the word "everyone"? What do they do today instead, and what does that cost them? Six honest tests will tell you whether to keep going, and should I start a startup? covers the personal side of the same decision.
2. Size the market from the bottom up. Count the buyers you could actually reach and multiply by what they'd pay. Top-down numbers from analyst reports flatter every idea equally. The TAM, SAM and SOM breakdown and the bottom-up sizing calculator keep it honest.
3. Talk to buyers about the past. What they last did about the problem, what it cost, who decided. Customer discovery is this step done properly.
4. Settle the company basics. Co-founders, equity and runway, covered below, because they get harder to change every week you wait.
5. Decide price and scope together. Scope decided before pricing has no budget to fit inside. Price decided before you know the buyer is a guess.
6. Build the smallest thing that tests it. With AI builders this is days, not months, which makes the steps above the slow, valuable part.
7. Win the first ten customers by hand. Then measure whether they'd be disappointed to lose it.

Each answer constrains the next. Most idea-to-startup advice hands you the steps as a checklist you can do in any order. You cannot, and the ones people skip are almost always the early ones, because they're the ones that risk a "no".

Co-founders, equity, and runway

Equity is the clearest example. Splitting it evenly on day one because the conversation is uncomfortable is the single most common structural mistake, and it's close to unfixable once someone's contribution diverges from their share. Vesting exists precisely so that a split made on incomplete information doesn't become permanent. A four-year schedule with a one-year cliff is standard, and skipping it is how founders end up with a departed co-founder holding a third of the company. The equity split calculator gives you a defensible starting point based on contribution rather than an argument.

Runway is the other number worth knowing before you need it. Runway is how many months you can operate before money runs out, and founders consistently overestimate it by forgetting that costs rise as the company does. The practical version of the question isn't "how long can I last" but "what has to be true by the time I'm down to three months, and is that achievable?" Work it out with the runway calculator, check monthly spend with the burn rate calculator, and revisit both whenever you add a recurring cost.

If it's SaaS, the economics are the product

Recurring-revenue businesses have a failure mode that one-off sales don't: they can look healthy while quietly dying. Signups grow, the dashboard trends up, and underneath it customers are leaving nearly as fast as they arrive. Churn compounds against you exactly the way growth compounds for you, and at 5% monthly you're replacing your entire customer base every 20 months just to stand still.

Which is why MRR and ARR are the numbers that matter rather than cumulative signups, and why the ratio between what a customer costs to acquire and what they're worth over their life decides whether the business scales or just gets more expensive. Run the CAC-to-LTV numbers on your assumptions before you build. If the ratio doesn't work on paper with generous assumptions, it won't work in reality with real ones. The valuation calculator shows what those numbers are worth if you do raise.

Your first ten customers are research, not revenue

The first handful of customers are worth more for what they teach you than for what they pay. They tell you which feature they actually opened the product for, what nearly stopped them signing up, and what they'd have used instead. That's information you cannot get from a survey, because people are unreliable narrators of their own future behaviour: they'll tell you an idea sounds great and then never use it. The Mom Test is the standard method for asking questions that produce facts about the past instead of predictions about the future.

Get them one at a time, by hand, from wherever those people already are. Founders reliably skip this in favour of building more, because building is under your control and talking to strangers isn't. It's still the step that decides whether the next six months are spent on something real. When you have a few dozen active users, Superhuman's PMF engine turns "do they like it?" into a number you can track, and how to find product-market fit covers what to do with it.

Where ShipFit fits in the plan

ShipFit covers the part of this plan that happens before you write code or sign anything. It runs your idea through nine decisions in a fixed order (market verdict, buyer, pains, positioning, v1 scope, pricing, a pay-or-not test, launch, and export), each checked against live market signal and a named framework. A decision locks before the next one opens, which is the ordering discipline this page argues for, enforced rather than recommended.

What it does not do is the company-building above: the equity conversation, the runway maths, the first ten calls. Those stay yours. A Quick Take gives you the market verdict in about 2 minutes, and the full playbook takes 15 to 20. The free tier includes 3 credits a month and paid playbooks start at $5. Read the 9-step playbook, get a free idea score, or see pricing.