How to test a business idea before you build it

A few times a year I sit on the jury at a startup demo day. Founders pitch, a panel of investors listens, and then the questions start. I have watched this enough times to know what is coming, because the investors almost always ask the same three things.

Have you talked to your customers, and what did you learn? Have you put this in front of real users? What signal do you have from the market, beyond your own belief that it is a good idea?

And a lot of founders freeze. Not because the idea is weak, but because nobody told them that talking to customers was the job from day one. They spent months building, so they have conviction and a nice demo, but no evidence. You can watch it land on the panel in real time.

That gap is why I built Product Build Day, and why testing sits inside it instead of being bolted on at the end. It is also why I am writing this. The tests that answer those investor questions are cheap, fast, and you can run them yourself this week.

One thing before we start. You do not need to hire an expensive user researcher, and you do not need a research agency that charges five figures. Not at the beginning, and honestly not ever as your only way of learning. You can do all of this yourself, and you should, because the founder who talks to customers directly hears things a report never captures. Keep doing it when you grow to enterprise scale too, which I hope you do. The skill does not expire.

Here is the stake, in one number.

CB Insights (a company that tracks startups and venture funding) read more than a hundred startup post-mortems and found the top reason they died, at 42%, was no market need. They built something nobody wanted. When building an app took months and real money, that was one expensive mistake. Now that a working app is a weekend and a few coffees, it is a cheap mistake, and an easy one to make often. So test the idea before you pour your life into it.

One rule sits over everything below.

Ask what they actually did, not what they think of your thing. "Tell me about the last time you dealt with this" beats "would you use this?" every time. The second you pitch, they get kind, and kind is useless. That is the whole idea behind Rob Fitzpatrick's The Mom Test. Your idea is the last thing you reveal, not the first.

Seven ways to get real signal follow, roughly in the order you would use them. None needs a line of code.

Before you test anything, work out what to test first. Fill a one-page Lean Canvas, Ash Maurya's startup version of the business model canvas, then circle your riskiest assumption. For most ideas it is the same one: does anyone actually have this problem badly enough to pay to fix it? Point your first test straight at that.

This is half a day with a sheet of paper, and it stops you from spending a month testing the wrong thing. Do it when you start, and again whenever a result forces a rethink. Skip the fancy tools. A napkin works. As Maurya puts it, love the problem, not your solution.

The cheapest test there is costs nothing but your time. Find five to fifteen people who have the problem and ask about the last time they hit it: what they did, what they tried, what it cost them. Do not pitch. Watch for the three answers that mean nothing: a compliment, a vague "I would definitely buy that", and a feature request you did not ask for.

The trick is in the questions. A small change turns a useless answer into a useful one.

Ten honest conversations teach you more than a month of building, and they sharpen everything after them: your copy, your pricing, your pitch. Use this first, always. It is strongest in B2B (businesses selling to other businesses), where you can name the exact person. The one time to pair it with something else is a product people cannot picture, where you may need a quick demo alongside the talk.

People leave a trail, so mine it before you spend a cent. Read the Reddit threads and Facebook groups where your audience gathers. Pull the one-star and five-star reviews of the nearest competitor: the one-stars are the gaps to attack, the five-stars are what people will not give up. Check Google Trends to see whether demand is already there and which way it is heading. This is Steve Blank's "get out of the building" without leaving your desk.

It is close to free and you can do it tonight. Use it to find the customer's own words and to size a market. The limit is a brand-new category nobody searches for or discusses yet, where there is no trail to read.

Talk is cheap, so make people act. Build one page with a sharp promise and a single button, and measure who signs up. Joel Gascoigne did exactly this for Buffer in 2010: a plain page that described the product as if it existed, a "plans and pricing" button, and emails collected as proof. He got a paying customer within days of building the real thing.

A signup is a real action, not an opinion, and it costs almost nothing. Use it to test demand and your message for anything digital. If a signup would not really mean intent to pay, add a price step, which brings us to the sharper version.

Advertise the thing before it exists and see who tries to walk through. Put up a button, a plan, or an ad for the feature. When someone clicks "buy" or "get it", show an honest "coming soon, join early access" and capture the intent. This is a named technique from Alberto Savoia, who coined pretotyping at Google, and it is one of the cleanest demand reads you can get.

Use it to test appetite for one feature or a whole concept before building it. The rule is to treat the clickers with grace: an apology, a waitlist, or an alternative, never a dead end. Skip it if repeated fake doors would burn trust, or if you need to know why people clicked, in which case go interview them.

The strongest signal short of a product is money or a signature. Take a pre-order or a deposit from a consumer, or a signed letter of intent from a business. This is the heart of Steve Blank's customer development: stop collecting opinions and ask people to commit.

It separates a nice-to-have from a must-have in one question, and it can even fund the build. Use it in B2B, high-ticket, or hardware, where a buyer can commit to something real. Go carefully with cheap impulse products, or where taking money before you deliver raises trust or legal problems. And remember: "I would definitely buy that" is a lie until a deposit backs it.

You can deliver the whole thing manually first, in the open, to real paying customers. This is the concierge MVP, where MVP means minimum viable product, the simplest version that tests the idea. Every step is done by hand, with the customer knowing a human is behind it. Manuel Rosso ran Food on the Table this way, building grocery lists by hand and charging about $10 a week, long before any software existed.

You learn what actually matters in real time, from people who paid. Use it for service-heavy or complex products where deep learning beats scale. Skip it when the value only shows up at scale, like a marketplace, or when your own charm would fake the result. Charge from day one anyway. Free hides the truth. A bill reveals it.

Where this leaves you

You do not need all seven. You need the one that hits your riskiest assumption, then the next. For most ideas that is two moves: talk to ten people the Mom Test way to check the problem is real, then stand up a landing page with a price to check people will act on it. One de-risks the problem, the other de-risks the demand. Together they answer most of what an investor will ask you, and neither needs a build.

Then read the companion guide on testing after you build, because a prototype you can throw together in a day makes some of these tests far stronger. The full method, with the worksheets, is free at productbuildday.com/playbook. Run one, then tell me how it went. You ran it, you get the credit.

Frequently asked questions