How to validate a business idea

What validating an idea means

Validating a business idea means collecting evidence that enough people have a problem, care about it enough to act, and will pay for a fix. It's not about whether the idea is clever. It's about whether people with the problem do something when you offer to solve it.

The evidence that counts is behavior: posts people write when nobody asked them, money they already spend on workarounds, emails they leave on a page, clicks on a price. Opinions don't count, because people are polite. Ask whether your idea is good and most will say yes, since a yes costs them nothing.

You don't need a product to collect this evidence. You need a clear description of the problem, a way to reach people who have it, and a page that asks them to act.

Why validate before you build

The most common way a new product fails isn't bad code. It's that nobody wants it enough to switch from what they do today. Founders often find this out after months of building, when the launch gets a few kind words and no sign-ups.

A week of validation costs a fraction of a month of building, and it tells you which problem to build for, who has it and what they'll pay. If the answer is no, you've saved the months. If it's yes, you start building with a list of people waiting for it.

You can see the cost of skipping it in what founders write. These posts are from our ideas library: people who built or stocked up first, and then went looking for customers.

Step 1: Write the idea as a problem

Before you test anything, write down who has the problem and what it costs them. Leave the product out for now. “An app for freelancers” can't be tested. “Freelancers who wait more than 30 days to get paid and spend hours chasing invoices” can, because you can go and find those people.

A simple template helps: for [who] who [have this problem], [your idea] [does this], unlike [what they use today]. If you can't fill in the last part, find out what people use today before you go further. There's always something, even if it's a spreadsheet or putting up with it.

• Who: a group you could find online this week, not “small businesses” or “everyone”.
• Problem: something they'd describe in their own words, not a feature you want to build.
• Today: the tool, workaround or habit you'd have to replace.
• Cost: the time, money or stress the problem causes now.

Step 2: Find people already asking for a fix

People with a real problem talk about it in public. They ask for advice on Reddit, complain in forums, leave one-star reviews and post in groups for their trade. Search those places with the words someone with the problem would use: “chasing late invoices”, not “invoice automation”.

Read the posts, don't just count them. When we tested plain keyword search, 351 posts matched our keywords and only 22 of them, about 6%, actually described the problem. The rest used the same words for something else. A dozen posts that clearly describe your problem beat a thousand keyword hits.

At scale, the signal is there. In the 90 days to 19 October 2026, our tools flagged more than 28,000 Reddit posts where someone asked for a fix to a problem. When we read 1,452 of them closely for our ideas library, they grouped into 50 problems that at least three people shared.

For each post, note three things: the exact words people use, what they've already tried, and whether they mention money or time lost. Those words become your landing page's headline. The workarounds tell you what you're competing with. Here's what real posts look like, all from people with the same problem: customers who pay late.

Search Reddit for people asking about your idea, free

Step 3: Talk to five of them

Posts tell you a problem exists. Conversations tell you how much it hurts. Reply to a few posters, or reach out to people in the same community, and ask for 15 minutes about the problem, not about your idea.

The trick is to ask about their life instead of your idea, the core rule of The Mom Test, Rob Fitzpatrick's book on customer interviews. Ask about the last time it happened, what they did, what it cost and what they've tried. Facts about the past are hard to fake. Promises about the future are easy. Avoid “Would you use an app that…?” and “How much would you pay?”: both invite a polite guess.

Questions that work:

• “Tell me about the last time this happened.”
• “What did you do about it?”
• “What else have you tried? What did you like or hate about it?”
• “How much time or money does it cost you now?”
• “Who else should I talk to?”

Read the Mom Test guide

Step 4: Check the market is big enough

You don't need a precise number yet, but you should know whether there are thousands of people with this problem or dozens. A quick bottom-up estimate works better than a big industry figure: count the people or businesses you could actually reach, then multiply by what one of them would pay in a year.

For example, if there are 40,000 agencies you could reach and each would pay $600 a year, the market you can serve is about $24 million a year. That's plenty for a solid small business. If the honest count is 400, you'd need a much higher price or a wider group. These numbers are made up to show the math; use real counts for your own idea.

TAM, SAM and SOM, with a worked example

Step 5: Build a landing page with a price

Now test whether people act. Build one page that names the problem in your customers' words, says what your product does about it, shows a price and asks for one thing: an email, a pre-order or a click on a plan.

Show the price even though nothing is for sale yet. A sign-up with no price tells you people are curious. A click on a $29 plan tells you they'd consider paying. When someone clicks a plan for a product that isn't built yet, tell them straight away and offer to let them know when it's ready. That's a fake door test, and done honestly it's the fastest way to learn what people will pay for.

The page needs very little:

• A headline that names the problem, in words from the posts you read
• Two or three lines on what the product does about it
• A price, or two or three plans to choose from
• One call to action: leave an email or choose a plan
• Nothing else to click: no menu, no blog, no links away

How to run a landing page test

Step 6: Put the page in front of people with the problem

A landing page with no visitors proves nothing, and the wrong visitors prove less than nothing. The visitors that count are people who have the problem: the ones writing the posts you found, members of the communities they post in, people searching for a fix.

Reach them where they already are. Reply helpfully to recent posts and mention the page when the community's rules allow it. Share it in groups that welcome it. Message people who described the problem in detail, and refer to what they wrote. A small ad budget aimed at the same group works too, but expect ad visitors to convert worse than people who asked for a fix themselves.

Aim for at least 100 real visitors. Below that, one or two sign-ups swing the result too much to read.

Step 7: Read the numbers after 7 days

After a week, look at what visitors did, not how many came. Count only real people: filter out bots and your own visits. Then compare with the thresholds below. They're the ones Ghostlaunch's reports use, and they're starting points, not laws: a $2,000-a-year product for businesses can work with lower rates than a $5 app.

Read the sign-ups themselves too. One email from someone who wrote a detailed post about the problem is worth more than ten from a giveaway.

Step 8: Decide, then build the smallest version

A test ends in one of three decisions. Build: the numbers are strong and you know who signed up. Change: interest is there but weak, so test a different audience, problem or price. Drop: nobody acted, even after a change. Dropping an idea after a week is a win. It's months you didn't spend building the wrong thing.

If you build, build the smallest thing that solves the problem for the people who signed up: a minimum viable product. Sell it to them before you add features. Their questions and complaints are the best roadmap you'll get.

What an MVP is, and how small it should be

A 7-day plan

Here's the whole method as a week. Ghostlaunch can run days 4 to 7 for you: it builds a page made to convert, reaches people already asking for a fix, and sends a report on day 7. You can also do all of it yourself.

See what a 7-day test reports

What doesn't count as validation

Friends and family saying it's great

They want you to succeed, so they say yes to almost anything. Ask them for introductions to people with the problem instead.

Survey answers like “I'd use that”

Saying yes costs nothing. Watch what people do when it costs them an email or money.

An AI score for your idea

A model has never had your customer's problem. It can help you write the page, not tell you whether people want it.

Likes on a launch post

Likes mostly come from other founders. Count sign-ups from people who have the problem.

Competitors exist, so it must be good

Competition shows people pay for a fix. It doesn't show they'd switch to yours. Read what they complain about.