How to Start a Startup: An Evidence-First Path to First Sale

To start a startup, pick a problem you understand, test it with the people who have it, build the smallest version that solves it, sell that version yourself, and only then scale what worked. Paul Graham summarised the requirements in 2005: start with good people, make something customers actually want, and spend as little money as possible.

Graham added: "Most startups that fail do it because they fail at one of these. A startup that does all three will probably succeed" (Paul Graham, "How to Start a Startup"). The path below is built around the middle one, because it is the one founders most often assume rather than test.

Key takeaways

Start with a problem, not a product or a company name.
Test before you build: conversations, then a commitment.
Build the smallest version that delivers the outcome, even by hand.
Sell it yourself. The first sales teach you the business.
Spend as little as possible until the evidence says scale.

What are the steps to start a startup?

Six steps, each with its own evidence. Move to the next only when the current one has produced what it should.

Six steps from idea to first sale
Step	What you do	Evidence to move on
1. Pick a problem	Choose one you understand, with a buyer you can name	A one-sentence buyer problem
2. Talk to buyers	Customer discovery conversations	Repeated, unprompted descriptions of the problem
3. Get a commitment	Offer a pilot, pre-order or deposit	Someone agrees to pay or commit
4. Build the smallest version	Deliver the outcome, by hand if needed	The first customers get the result
5. Sell it yourself	Founder-led sales to the next customers	A repeatable way deals close
6. Decide what to scale	Double down on what worked	A channel and motion you can grow

How do you pick a problem to start a startup around?

Choose a problem you have seen up close, where the people who have it already spend money or time on a workaround, and that you can reach this month.

Graham's advice from a later essay: "The way to get startup ideas is not to try to think of startup ideas. It's to look for problems, preferably problems you have yourself" (Paul Graham). Our guide to finding a business idea turns that into a method.

How do you test a startup idea before building?

Talk to the people who have the problem about the last time it happened, then ask the most interested for a commitment such as a paid pilot or a pre-order.

Customer discovery and the business idea validation checklist cover both steps. A free Quick Validate gives a first read on the idea and names its weakest point, which is a good place to aim the first conversations.

What should you build first?

The smallest thing that delivers the outcome your first committed customers paid for. Often that is partly manual.

Graham's "Do Things That Don't Scale" makes the case for doing early work by hand: "The most common unscalable thing founders have to do at the start is to recruit users manually" (Paul Graham). The same goes for delivery. Doing the work yourself shows which parts are worth automating. Our guide to the minimum viable product covers how to scope it.

Who should sell a new startup's product?

The founder. The first sales are where you learn who buys, why and what stops them, and that learning has to reach the product.

See founder-led sales for a six-step playbook, from building a buyer list to converting a paid pilot.

How much money do you need to start a startup?

Less than most founders assume, if you test before building and deliver the first version by hand. Spend on what the evidence says works, not on what a startup is supposed to have.

Graham's third requirement, "to spend as little money as possible", is partly about survival and partly about learning: the less you spend before you know what works, the more attempts you can afford. When you do need money, customers are the cheapest source. Our guide on how to get money to start a business covers the options.

Where does NELL fit when you start a startup?

At each step: research and scoring to choose and test the idea, a go-to-market plan and build brief to launch it, and operators to help execute.

Quick Validate is free. DeepValidate ($49 for your first two reports) adds sourced research on market, competitors and risks. Launchpad ($149) adds a 21-section go-to-market strategy. Pro Builder adds NELL operators working with you on execution. Compare them on the pricing page.

Frequently asked questions

What is the first step to starting a startup?

Pick a specific problem you understand and a buyer who has it. Everything else, including the product and the company, follows from that.

Do I need to incorporate before starting a startup?

Not to begin testing an idea. You will need a legal entity before taking investment, signing some contracts or hiring. Check the rules in your country.

Do I need a co-founder to start a startup?

Not necessarily, but Graham lists starting with good people as one of three requirements. A co-founder with complementary skills can share the work and the judgement.

How long does it take to start a startup?

Testing an idea with buyers can take weeks. Reaching a first sale depends on the market; the evidence-first steps aim to get there before you spend heavily on building.

Can I start a startup while working a full-time job?

Many founders test ideas part-time. Conversations, commitments and a hand-delivered first version fit around a job better than building a full product does. Check your employment contract first.

Start where you are

Start with the problem. A first read takes minutes.

Sources

This guide is general information, not legal or financial advice. Company formation, employment and tax rules vary by country.