How to build a startup from scratch

I get asked this in some form every week, usually by someone who already has an idea and wants to know what the first step is. The answer is unglamorous, and it is almost never the step people expect. Nothing in the first month involves a company, a logo, a deck, or a developer.

Start with a problem you can describe in one sentence, said by a specific person you can name. Not a market, not a trend. If the sentence needs the words digital transformation or platform to make sense, you do not have it yet. The test is whether you can call ten people who have this problem this week. If you cannot find the ten, that is the finding, and it cost you nothing.

Then talk to them without pitching. Ask what they do today instead, how long it takes, what it costs them, and what they tried before. People are unreliable about the future and quite reliable about the past. The question that separates a real problem from a polite one is whether they are already spending money, time, or a spreadsheet on a workaround. A workaround is proof of demand. Enthusiasm is not.

Only then build something, and build the smallest thing that produces the outcome. Not the product you imagine, the outcome the person wants. It can be a form, a spreadsheet, a script, a manual service you personally perform while pretending it is automated. I build these myself without being a developer, and the point is never the artifact. The point is that a prototype ends debates a document would extend.

Charge for it earlier than feels comfortable. The first invoice teaches you more than the first hundred users of a free tool, because paying reorders someone's priorities and reveals what they actually value. In a small market this matters twice as much: local venture rounds in Southeast Europe averaged under 4 million euro across 109 deals in 2025, so revenue is the main thing you control that moves your position.

Distribution is the part almost everyone underestimates. Decide before you build how the first fifty customers will hear about you, and make it a channel you can operate yourself: direct outreach, one community you are genuinely part of, one partner who already has the audience. If your plan is that people will find it, you do not have a plan, you have a hope.

On the team: bring in a co founder when there is work neither of you can avoid, not to feel less alone. Give it a trial project with a deadline first. Split equity with a vesting schedule, write it down while you still like each other, and never let the person doing the least work hold the most of it. Most of the founder disasters I have watched were not about money, they were about undocumented expectations.

Now the parts you can safely ignore for the first ninety days. Company registration beyond what your first invoice requires. Branding past a name you are not embarrassed to say. Pitch decks, unless a specific person asked for one with a date attached. Accelerator applications before you have anything to show, because a program compresses time for a company already moving and does very little for one that is not. I co created an accelerator and I still tell people this.

Where people here lose the year, in the order I see it. Building for six months without a single conversation with a buyer. Building only for the home market, which is too small to support the company they are describing. Waiting for a grant or a program to start rather than starting and using the program as leverage. And treating the idea as the asset, when the asset is the speed at which you learn.

A workable first ninety days looks like this. Weeks one to three: twenty conversations, no pitching, written notes. Weeks four to six: the crudest working version, put in front of five of those people. Weeks seven to nine: the first paid customer, at whatever price makes them say yes. Weeks ten to thirteen: repeat the same sale four more times without changing the pitch every time. If you can do that last part, you have a business to build. If you cannot, you have learned the most expensive thing at the cheapest possible price.

That is the sequence: a named person's problem, evidence of a workaround, the smallest thing that delivers the outcome, money, a channel you own, then a team. Every step after that is scale, and scale is a different essay.

If you are in the middle of this and stuck at a specific step, write to me and tell me which one. The answer is usually different for step two than for step five, and generic advice is what makes it feel harder than it is.