How to start a business

There is no secret first step. There is a sequence, and most of it is duller and cheaper than it looks from outside. This is that sequence — what matters at each stage, what can wait, and the parts most guides skip.

Start with a problem you can already reach

The question that decides most of what follows is not which market is biggest. It is who will reply to a message from you. An idea you can reach is worth more than a better idea you cannot, because at the start you have no audience, no budget and no reputation in the category — all you have is the distance between you and someone with the problem.

Three things you already own are the real inputs. Access: the people who will read a message from you because they know your name. Credibility: the subjects on which your opinion is taken seriously without a pitch. Repetition: something you have watched go wrong more than once, in a job, a hobby or a family business. Where those three overlap you get a business you can start on a Tuesday evening rather than one that needs a launch.

Be suspicious of ideas that only work at scale. Anything of the shape "like a big platform, but for a niche" usually needs thousands of participants before the first one gets any value, and one person cannot manufacture that. A business that helps ten people substantially can be built in weeks and grown on purpose; a business that helps ten thousand people slightly cannot be started at all.

Finally, be honest about the difference between a business and a project. A business is something a stranger pays for deliberately, more than once. If you cannot describe that stranger in a sentence — their job, what their week looks like, the moment they would go looking for a solution — you have a project. That is a fine thing to have, but do not put a company around it yet.

Test the demand before you build the thing

The cheapest version of your business is a conversation. Before you write code, order stock or design a logo, describe the idea to ten people who match your one-sentence stranger and ask what they do about the problem today. You are not listening for enthusiasm — people are polite. You are listening for effort. Somebody who has built a spreadsheet, hired a temp or paid for a tool they dislike has already told you the problem is worth money to them.

The next-cheapest test is one page and a price. Say what the thing is, who it is for and what it costs, and ask for something small in return: an email address, a deposit, a booked call. A page nobody signs up to is a far cheaper failure than a product nobody buys, and the sign-up rate is a real number you can work on rather than a feeling you can argue about.

Do not skip the price. An idea validated without a number is not validated. "Would you use this?" gets a yes from nearly everyone; "would you pay this, on the first of the month, starting now?" gets an honest answer, and the objection that comes back is usually the single most useful sentence of the whole exercise.

Give the test a deadline and a threshold, written down before you start: twenty conversations, or two weeks of a live page, and this many replies counts as a yes. Deciding what success looks like in advance is the only real defence against reading a mediocre result as encouraging.

Naming it, and taking the address

Names matter less than founders think, and caring about them in the wrong way costs weeks. A good name is short, pronounceable over a bad phone line, spellable by someone who has only ever heard it, and not already carried by a competitor in the same category. That is very nearly the whole brief. The name does not have to describe the product; it has to survive being repeated.

Two practical constraints do matter. Check that the name is not already a trademark in your industry and your country — the official register in most jurisdictions is searchable for free, and five minutes there can save a rebrand later. And get an address you own on the internet. A domain is one of the few decisions in a young business that is genuinely painful to reverse, because every link, invoice, business card and email signature ends up pointing at it.

Prefer a domain you own outright to a name inside somebody else's platform. A page on a marketplace or a social network can be suspended by someone who has never met you and owes you no explanation; a domain you registered stays yours as long as you renew it. If the exact name you wanted is gone, a shorter or slightly altered form on a domain you control beats the perfect name on an address you rent.

The admin minimum, honestly

The first question is whether you are allowed to trade as yourself. Registering a company is not automatically step one. In many places you may trade under your own name while you find out whether the business is real, and form an entity once there is income or liability worth separating; in others the registration comes first. What an entity mainly buys you is separation between the business's risks and your own, plus credibility with customers and suppliers — it is not, by itself, what makes selling legal.

The second is separating the money, and it is the one universal answer on this page: open a dedicated account for the business and run everything through it from the first transaction. Mixed personal and business money is the mistake that makes every later step harder — the tax return, the bookkeeping, the accountant, a loan application, a sale of the business — and it takes an afternoon to avoid and years to untangle.

The third is finding out what you owe, and when, before you owe it. There is a tax authority in your country with a page describing exactly what a new business in your situation has to file and by when. Read that page rather than an article about it: rules change, and the version that binds you is the official one. If a sales tax or VAT applies to what you sell, look up the registration rules early, because they often turn on a threshold you can cross without noticing.

The fourth is writing down what you promise. Terms of sale, a refund policy and a privacy notice are not corporate theatre; they are the document you will be glad to have during your first difficult customer, and they are the reason a payment processor will keep working with you. If you collect personal data, find out which privacy rules apply to your customers — they usually follow the customer's location, not yours.

Getting online: an address, an inbox, a way to be paid

You need three things and they are smaller than they sound: somewhere people can read about you, somewhere they can reach you, and somewhere they can pay you.

The site does not have to be impressive; it has to be specific. A stranger arriving on it should learn within one screen what you do, who it is for, what it costs and how to begin. Most first sites fail on the last two: the price hidden behind a call, and no obvious next action. Four honest pages — what it is, who is behind it, what it costs, how to get in touch — is a complete site, and a complete site beats a beautiful one every time.

The inbox should be at your own domain. An address at the domain you just registered says the business exists; a free consumer address on a quote or an invoice quietly says the opposite, and fixing it takes minutes.

Payment is the piece people delay longest and regret most. Set it up before you feel ready. A payment account in your own name, connected to your own bank, means the person who decides to buy can buy at the moment they decide — and a business becomes real at its first transaction, not at its launch. Keep the account in your own name rather than as a balance inside somebody else's platform, so that the money, the customer records and the payment history are yours.

If that block sounds like a week you do not have, it is the week we built First Employee to remove. You leave one sentence about the business; it researches the market, chooses a name, and puts a live multi-page site with its own inbox online in about three minutes. Building it is free, and every step it took is written down as a receipt you can read and export. See exactly what it does.

The first ten customers

Most guides stop at "launch", which is where the actual difficulty begins. The first ten customers do not arrive from an audience, because you do not have one. They come from people you can name.

So write the list. Actual individuals who plausibly have the problem — former colleagues, people in a trade group, the owner of a business two streets away, someone who complained about exactly this in public last month. Message them one at a time, about their situation rather than about your launch, and ask a question you genuinely want answered. A note that could only have been written to that person gets replies that a broadcast never will.

Charge the first one. A free pilot teaches you almost nothing, because nobody guards a free thing or complains about it honestly. A paying customer, even at an embarrassing price, tells you what the product must do and gives you a sentence to use on the next person. Then ask the ones who said no what they are doing instead — those answers are a description of your real competition, which is usually a spreadsheet or a habit rather than a company.

Serve the early ones conspicuously well, further than is sustainable. It is the only period where you can afford it, and it buys the two things that compound: the specific words customers use about the problem, and referrals. Ask for the referral directly, right after somebody tells you they are happy. Almost nobody does, which is why it works.

Then pick one route to new customers and give it a fair test before adding another. Search, a partner who already serves your customer, a community you are actually part of, direct outreach, or paid ads — each of these takes weeks to read properly. Running five at once produces five unreadable results and a founder who is tired.

Staying alive, week after week

Starting is a weekend. Week twenty is the hard part. Businesses at this stage rarely die of competition; they die of neglect, because the work that keeps them alive is unglamorous, never urgent, and always the first thing dropped when a customer needs something today.

That work is roughly the same everywhere: publish something so there is a reason to find you, follow up on the enquiries you already have, keep the site current with what you actually sell, read the few numbers that matter, and answer your email. Each item is small. Together they are a second job, and you already have the first one.

This is the part of the sequence we built a product for, and the only part where we would claim to be the answer. First Employee is an AI employee that works a shift on your business every night: it picks up where the last shift ended, does the week-to-week work, and leaves one page in the morning saying what shipped, what it learned and the single question it needs answered. Anything irreversible — sending mail to a real person, posting publicly, spending money — waits for you. Every action it takes writes a receipt with the tool, the timing and the artefact, so the work is auditable rather than asserted.

The build is free. Nightly shifts come with a plan, and the plan is the whole bill on the self-managed side: we take 0% of your revenue and 0% of your ad spend. If you would rather we ran the business side for you, the managed plan takes 15% of what the company earns. On either one your customers pay an account in your own name, so we never hold your money. How a shift works · What a plan costs

Questions people ask at this stage