How to Validate a Business Idea

Almost everyone who sets out to validate a business idea collects the wrong evidence, and does it enthusiastically. They run a survey. They post in a founder group. They describe the idea to twenty people and count how many nodded. Then they build, and the people who nodded do not turn up.

The problem is not effort. It is that these methods all measure the same thing — how agreeable people are — and agreeableness is free. A better approach is to rank every piece of evidence by one question: what did this cost the person who gave it to me?

Cheap signals are abundant and mean almost nothing. Expensive signals are rare and mean almost everything. Most of validation is the discipline of refusing to be encouraged by the first kind.

The signal ladder

Here is the same idea receiving evidence at six different prices.

That last row surprises people. Someone who has already hacked together a spreadsheet, hired a contractor, or is misusing a competitor's tool to do the job has paid for the problem before you ever appeared. They did not do it to be encouraging. That is the strongest evidence available, and it can be found before you build anything.

Everything below is about climbing this ladder as fast as possible.

Step 1: Write the assumption down, specifically

You cannot test "will my idea work." You can test a sentence of this shape:

[A specific kind of person] has [a specific problem] often enough that they currently [a specific costly workaround], and would pay [an amount] for a better answer.

Filling that in is uncomfortable, which is the point. "Small businesses struggle with bookkeeping" cannot be proven wrong. "Freelance designers in the UK billing under £80k spend two evenings a month reconciling invoices by hand and would pay £15/month to stop" can be — and every clause in it is separately checkable.

Vague assumptions survive every test. That is not a feature.

Step 2: Look for the problem before you pitch the solution

Before you describe your idea to anyone, go and find out whether the problem shows up in places where nobody is performing for you.

Search for it in support forums for adjacent tools, in product reviews — the three-star ones, which are the honest ones — in subreddit threads where someone describes their setup, and in job listings. Job ads are the most under-used source in validation: a company paying salary for a role whose description is your feature list is a company already spending money on your problem.

You are looking for complaints nobody was asked to make, and workarounds nobody was paid to build. This is problem validation, and it costs you a few hours.

If you cannot find the problem anywhere, that is a result. It is a much cheaper result now than in a year.

Step 3: Talk to people, but ask about the past

Interviews are useful for language and context, and dangerous for demand. The danger is entirely in the questions.

Ask "tell me about the last time this happened" and "what did you do about it" and "what did that cost you." These request facts that already exist.

Do not ask "would you use this" or "would you pay for this." Those invite the other person to be kind, and most people are. A yes costs them nothing, which is exactly why it is worth nothing. The reliable tell that you asked the wrong question is that the answer felt good.

The goal of this step is not a verdict. It is to sharpen the sentence from step 1. See customer discovery for the full method.

Step 4: Make people take a costly action

This is where most validation stops too early, and it is the only step that produces evidence from the top half of the ladder.

Put up a landing page that describes the product as though it exists, with one clear action, and drive real traffic to it. Then count. This is a fake door test, and the details that determine whether it means anything are these:

The traffic must resemble the market. Sending it to your own network measures how much your friends like you. It has to reach people who match the sentence in step 1 and have never heard of you.

The action must cost something. An email is the minimum. A booked call is better. A card is best. Pageviews are not evidence.

Decide the threshold before you look. Write down the number that would count as a yes before you run it, or you will find a way to read whatever you get as encouraging. Everyone does this. Writing it down first is the only reliable defence.

Give it enough volume to mean anything. Four signups from forty visitors and four from four thousand are opposite results.

Step 5: Decide, and be willing to hear no

A test that cannot come back negative was not a test. If the number falls below the line you set in step 4, the honest reading is that this specific idea, aimed at this specific audience, with this specific pitch, did not find demand.

That is three variables, and it is worth being precise about which one failed. Most ideas that come back negative are not dead ideas — they are correct ideas aimed at the wrong segment, or right segments approached with the wrong message. The next test should change one thing, not start over.

An inconclusive result is also a result: usually it means the sample was too small or the traffic was wrong, and the fix is a better-run test rather than a new idea.

The mistakes that account for most failures

Asking people to imagine the future. Nobody can predict their own behaviour, and everybody thinks they can.

Testing on people who know you. Warm audiences produce the highest numbers and the least information.

Moving the goalposts after the fact. Related to not setting them beforehand.

Confusing interest with demand. Interest is free. Demand costs the person something.

Validating the solution and skipping the problem. A well-built landing page can pull signups for a problem that turns out to be shallow. That is market validation without problem validation, and it fails later and more expensively.

Stopping at the first yes. One enthusiastic buyer is an anecdote. The question is whether there are a thousand of them and whether you can reach them repeatedly.

How long this should take

Steps 1 to 3 are a few days of unglamorous reading and half a dozen conversations. Step 4 is one to two weeks — long enough for outreach to land and replies to come back, short enough that you have not started building in the meantime.

If validating an idea is taking two months, the test is too elaborate. The purpose is to be wrong quickly and cheaply, not to be thorough.

Where theLab fits

theLab automates step 4, which is the step most founders skip because it is the one that requires infrastructure. You describe the idea; it extracts the assumption, builds the landing page, finds real people matching your ICP, runs the outreach, and returns a verdict based on signups and preorders — signals from the upper half of the ladder rather than the free end.

The first experiment, including the market report, is free. What it will not do is tell you your idea is good because you would like it to be. A meaningful share of experiments come back Not Validated, and that is the outcome the whole exercise exists to be able to produce.

For the mechanics of the demand test itself, see how to test demand for a product.

theLab builds the landing page, finds real people who match your audience, runs the outreach, and returns a verdict from real signups. First experiment free, no card.