How do I decide on a business idea?

You’re not choosing an answer. You’re choosing a question to go and ask.

That distinction changes everything about how the decision should be made. If the idea you pick is a final commitment, then picking well means predicting the future, and the sensible thing is to research for months. If the idea is a starting position — the first version of something the market will edit — then picking well means choosing the question that returns the most information soonest, and being genuinely willing to hear the answer.

The second one is closer to how it actually goes. In a survey of 200 US tech founders run in February 2026, 81% said their company had pivoted from its original idea. The number worth sitting with is a different one from the same survey: 42% wished they had changed direction sooner.

Nobody regrets a pivot they made early and cheaply. What they regret is the year spent defending a first draft. So the job in front of you is not to find the idea that will not need changing — it is to pick one you can start editing quickly.

1. Write down more than one

You cannot choose from a set of one, and most founders arrive holding exactly one idea and calling the decision made.

Write down three or four. Not because you are looking for the best of them — you cannot tell yet, and pretending otherwise is the trap — but because the act of putting an idea next to alternatives changes your relationship with it. One idea is an identity. Four ideas are options, and options can be edited without it feeling like a defeat.

If you genuinely only have one, that is fine. Skip the comparison and go straight to step two; just be aware you will have to work harder to stay honest, because there is nothing to hold it against.

2. Say each one as a sentence about somebody else

For each idea, write one sentence: who it is for, and what changes for them.

Not what it is. What changes. "A scheduling tool for clinics" is a product. "Clinic receptionists stop losing an afternoon a week to appointment reminders" is an idea — it names a person and a specific thing that gets better.

Two things are worth noticing about ideas that fail this test. They are usually not vague by accident: the vagueness is hiding the fact that you have not decided who the customer is. And they are almost always described from the inside — what the thing does, how it works, why the technology is clever — because that is the part you find interesting. Your buyer is not interested in the same part.

An idea that cannot be said as a sentence about somebody else is not ready to be compared against anything. It is not a bad idea; it is not yet an idea.

3. Name the problem underneath the solution

This is the step that decides how much the next six months cost you, and it takes about ten minutes.

For each idea, separate two things: the problem you believe exists, and the solution you have imagined for it. The problem is what you are actually betting on. The solution is a guess — an early, unexamined, quite likely wrong guess — about the best way to address it.

Founders reliably fall in love with the second one. It is the more enjoyable half: it is concrete, you can picture building it, you can imagine using it. The problem underneath is abstract and slightly boring. But the solution is the part the market is entitled to rewrite, and the problem is the part that has to be true.

A useful test: write the problem in one sentence without naming your product. If you cannot, then the solution is the entire idea, and every piece of feedback that contradicts it will feel like an attack rather than information.

4. Rank them by how fast you could be proved wrong

Now compare, and compare on the thing that actually matters at this stage: how quickly and cheaply each idea could be shown to be wrong.

Three questions do most of the work:

• Can you name the buyer specifically enough to reach them this week? Not "small businesses" — a group with a name, somewhere they gather, or targeting you could set up this afternoon.
• Can you describe it well enough to put on a page? If the promise still needs a conversation to make sense, the test will measure your explaining rather than the idea.
• What would it cost to find out? Some ideas can be tested for a few hundred dollars in a week. Others need a prototype, a licence, or a partner before anyone can react to them honestly.

Rank on those three, and the ordering you get will often be uncomfortable — the idea you like least frequently tests fastest. That is not an argument for building it. It is an argument for asking about it first, because you will know something real about it before you have spent anything meaningful on any of them.

5. Pick the one you would still want after it changes

Assume the idea you choose will not survive contact with the market in its current form. Assume the product narrows, the audience shifts, and the words you use change completely. Which one do you still want to be working on?

That question does more sorting than any scoring system, because it asks what you are actually attached to. If your answer is "none of them, unless it stays as I described it," you are holding solutions, and the first real piece of feedback is going to feel like a loss rather than a lead.

None of this is an argument for having no conviction. You need a strong opinion about the problem — that is the bet, and hedging it produces the bland, agreeable ideas that never find anyone. What you hold loosely is everything downstream of it.

6. Take it out as a question, not a plan

You have chosen. What you have chosen is a first draft with a specific buyer attached, and it is worth remembering that the decision you just made is the cheapest one in the whole sequence — the expensive decisions are all downstream, and they get made for you if you go quiet now.

So the next move is not building. It is a loop, and each pass costs less than the one before:

1. Check the group is real. Before spending anything, find out whether the people you named genuinely exist as a group, gather somewhere, and describe the problem in their own words. This costs nothing but time.
2. Put the promise somewhere it can be judged. One page, one promise, a price and a way to respond — built in a day, not a month.
3. Buy a small amount of attention. A few hundred dollars over a week is enough to see a large difference, which is the only kind of difference worth acting on at this stage.
4. Edit the idea on what comes back, and go again. This is the step that gets skipped, and it is the one the whole loop exists for.

Pratham Naik's account of building a project management tool is a clean example of the loop working. He spent six months building for indie hackers, guided by feature polls in that community — simple pricing, lifetime deals, dark mode, API access. His first real customer, a marketing agency owner, asked within five minutes where the client portal was, how to white-label it, and whether clients could comment without creating accounts. None of it existed.

He stopped running polls and started booking short calls with the people who would actually pay. Retention went from 12% to 41%; monthly revenue from $347 to $2,100. His own summary of what he got wrong is worth the whole story: the indie hacker community, he wrote, "is amazing for support and learning. But it's a terrible place to validate most products."

The idea was never the problem. The audience he had chosen to listen to was — and one conversation with a real buyer was enough to reset it.

When choosing isn't the hard part

When somebody is already asking you for it. The deciding is done. Go and find out what they would pay, and whether there are more people like them — that is a much better use of the next fortnight than ranking alternatives.

When every idea you have serves people you cannot reach. Reachability decides before merit does. A modest idea aimed at a group you can find beats a brilliant one aimed at a group you cannot describe, and no amount of comparing the ideas themselves will fix that.

When comparing has become a way of not starting. Endless ranking is a way of never being told no. If you have been deciding for more than a couple of weeks, the deciding is not the bottleneck — pick whichever one you could test soonest and go and get the answer.

Related questions

The first thing to check after choosing is whether the people you named genuinely cohere as a group: how to tell whether your target market is real. Then the idea needs somewhere to be judged — what to put on a website built to test an idea — and a little traffic pointed at it, which is the whole demand test end to end.

If you find someone already doing it, that is good news and it needs an angle — and the tools for each stage are worth knowing before you pay for any of them.

Draper runs this loop with you: starting from the idea you have, working out who it is for, and taking it to the market to find out what needs to change. Try Draper free →