How to validate a business idea in six steps

By the FounderFold teamPublished

To validate a business idea is to find evidence that a defined group of people has a specific problem and will pay to solve it. You look for that evidence before spending serious money or months. In practice it takes six steps, in this order. First write the problem down and name who pays. Then map what people use today, look for demand signals that already exist and run one cheap experiment with a number set in advance. Last, write down when you will stop.

The order matters because each step costs less than the next one. A week of conversations is cheaper than a landing page with ads, and a landing page is cheaper than a product nobody asked for.

1. Write the problem in one sentence

Start with one sentence: who has the problem, and what you offer them. "Small accounting firms lose hours every month chasing clients for receipts; we collect them over WhatsApp" can be tested. "An app for freelancers" cannot, because you don't know whom to call.

Then talk to people who have the problem. Ask about their work, not about your idea. Ask about the last time it happened and what it cost them. Then ask how they dealt with it. Rob Fitzpatrick's The Mom Test is a short book on exactly this: how to avoid biased feedback and how to tell whether someone is really going to buy (momtestbook.com, accessed November 19, 2026).

Our starting point is five conversations in seven days. If at least three people describe the problem without you suggesting it, and tell you how they handle it today, carry on.

2. Name who pays, and roughly how much

The person with the problem and the person who pays are often different. In a company, the user may be an analyst while the budget sits with a manager two levels up. Write down who signs, what they pay today for anything similar and what you would charge.

If nobody in your conversations has ever paid to ease the problem, write that down too. It may hurt less than it seemed.

3. List the competitors and what people use today

Your real competitor is often a spreadsheet or an email thread, and sometimes it is doing nothing at all. List each alternative with its URL and its public price. FounderFold's reference cycle for software built by one person, the stages and gates we start from for that type of business, asks for exactly that before any experiment is designed: the competitors and their reference prices, each with its URL.

Finding no competitor is not good news by default. Either you found a gap, or few people care enough to look for a fix. The conversations from step 1 tell you which.

4. Look for demand signals that already exist

Before you build anything, look for evidence that people already search for a fix or pay for one. Complaints about the lack of one count too. Free sources go a long way. Google Trends shows whether interest in a term is growing. One-star reviews of a competitor list what its customers miss. Industry forums and communities show the questions people actually ask.

Write each signal down with its link and the date you saw it. When you cannot link to it, record it as not verified, as you would any evidence.

The Test your idea tool does a first pass for you. FounderFold's AI asks a few questions, searches the web for who already offers something similar and what evidence there is of demand, and shows each finding with its source, plus what it could not verify. It is free and needs no sign-up. The sample report shows the whole output on an invented idea.

5. Run one cheap experiment with a number

An experiment validates something only if you wrote down the result that counts as success before you started. Pick the riskiest assumption, usually whether anyone will pay, and design the smallest test that answers it. A page with the price and a button to reserve or pay in advance works. So does a pre-sale to the people you interviewed, or a service delivered by hand to the first three customers.

Write four things before launch: what you will measure, the threshold, the deadline and who decides. For example: at least 10 sign-ups or replies in 30 days from one channel, and at least one person who pays in advance; the founder decides on day 30.

The best-known case is Dropbox. Before anyone could try the product, Drew Houston published a three-minute demo video, and visitors could only join a waiting list. He later said that list went from 5,000 to 75,000 people overnight (TechCrunch, Eric Ries, October 19, 2011, accessed November 19, 2026). A waiting list measures interest. A payment says more, which is why our starting threshold asks for at least one.

6. Write down when you will stop

Kill criteria are the results that make you drop the idea, written before the experiment starts so that a weak month cannot buy itself another one. Each one names a metric and a threshold, plus a deadline and the person who decides.

FounderFold's reference cycle for software built by one person starts from these three:

Kill criterionStarting threshold
Few demand signals (sign-ups, replies or pre-sales)Fewer than 10 in 30 days
Nobody agrees to pay or pre-orderZero when the experiment closes
Validation costs too muchMore than €150 or more than 3 days of work

These numbers are our starting point, not industry data: we have found no public benchmark for them. Adjust them to your price and your channel. A B2B service sold at several thousand euros a year can justify a test with far fewer signals than a consumer app. What matters is that you set the number before you start.

The kill criteria generator turns them into rules you can check and export in the browser, without sign-up.

When is an idea validated?

When the experiment reached its threshold in time and no kill criterion was triggered. That earns the next, slightly larger test. It does not prove the business will work: each stage after it needs its own number.

If you are weighing several ideas, score them first with the idea scorecard, which counts a dimension as zero when there is no source behind it, and validate only the strongest.

Our view: one experiment at a time per idea, with its criterion written first, teaches more than five tests run in parallel and read afterwards. When the results come in, the temptation is to move the threshold. A threshold written down with a date is much harder to move.

Where FounderFold fits

FounderFold is the agentic OS for building new businesses; for founders it is in early access by invitation. It keeps this process in writing for each idea. Each idea has its own record, with its hypotheses, the experiment and the criterion written down before it starts. Each experiment ends in a decision to carry on or stop, and the reason stays in the history. Discarded ideas stay in the portfolio with their reason, so you never research the same idea twice without knowing it. See how it works for founders.

Questions people ask

How long does it take to validate a business idea?

With the starting thresholds above, about a month for a first answer: a week of conversations and up to 30 days of experiment. The work itself should fit in three days and €150, spread over those weeks. If it doesn't, cut the test down until it does.

Can I validate a business idea without building the product?

Yes. Conversations, a page with a price and a pre-sale all test demand before you write any code. Dropbox's waiting list grew from a video, not from the software.

How many customer interviews do I need?

We know of no number with evidence behind it that holds for every market. Our starting point is five in a week, and at least three that describe the problem unprompted. If the answers point in different directions, talk to more people before you design the experiment.

What counts as a demand signal?

Any sign-up or reply from someone who has the problem counts, recorded with its date. A reservation or a payment counts more. Compliments and "I would use it" do not count. A commitment of time or money does.

Is an AI idea validator enough?

No. An AI searches faster than you and can show what you have not checked yet. It cannot talk to your customers for you or pay for your product. Use it to decide what to check first, then run the experiment with people.

Sources

All accessed on 2026-11-19.

  • Rob Fitzpatrick, The Mom Test, book website: https://www.momtestbook.com/
  • Eric Ries, "How DropBox Started As A Minimal Viable Product", TechCrunch, 2011-10-19: https://techcrunch.com/2011/10/19/dropbox-minimal-viable-product/ (quotes Drew Houston on the waiting list going from 5,000 to 75,000 people)
  • Google Trends: https://trends.google.com/ (free tool, cited as a source of signals, not for any figure)
  • FounderFold reference cycle "Software or SaaS built by one person", version 1.3 (2026-10-24): kill criteria kv1 and kv2; kill criterion kv4; milestone v2. Kill criteria kv1 and kv4 are marked as not verified against sector data; kv2 (no payment) is a rule, not a benchmark.

Start with the stage you are in.

Early access is by invitation. Tell us what you are discovering, building or growing, and we will set up your organization.