Market Validation: How to Test Demand Before You Build

Most ideas that get called “validated” were never validated. They were flattered. Someone showed the idea to friends, a few advisers, maybe a survey, everyone nodded, and that nodding got written down as proof. Then the thing launched and nobody bought it.

Market validation is the process of proving that real demand exists for an idea before you build it. Not proving that people like it, or that they would like it in theory, but that the people who would actually pay behave as though they want it. That distinction is the whole game, and it is the one most guides skip. When CB Insights went through the post-mortems of more than 400 venture-backed companies that shut down since 2023, 43% blamed poor product-market fit, and two-thirds of those had never found a market at all. Running out of cash topped the list at 70%, but that is the symptom. The disease underneath it is building something people did not want.

This guide is about how to catch that before it costs you a year. It covers what market validation actually is, a simple way to grade any piece of evidence, the methods ranked by how hard they are to fake, how to talk to customers without fooling yourself, and where the whole approach still misleads you.

What market validation actually is

Market validation puts your idea in front of the people you want to sell to and measures whether there is genuine appetite for it, before you commit real money to building. The idea can be a product, a service, a feature, a whole business. The test is always the same underneath: would enough of the right people part with something to get this, and can you reach them at a cost that leaves a business standing.

It helps to separate two things that get muddled. Market research describes a market: how big it is, who is in it, where it is heading. Market validation tests one specific claim, that your idea will find paying customers inside that market. Research tells you the water is warm. Validation tells you whether this particular boat floats. You usually do a bit of research first to size the opportunity, then validation to decide whether to build.

It also helps to be honest about what validation is for. It is not a research exercise you tick off to feel prepared. It exists to change a decision: build it, change it, or drop it. If a test cannot move what you do next, it is theatre, and you are paying for reassurance. That framing runs through everything below.

The two questions worth separating are the problem and the solution. Is the problem real, frequent and painful enough that people are already doing something about it? And will they choose your answer over what they use now? Most people leap straight to the second question, demo their solution, and get told it is lovely. They never checked whether the problem was worth solving in the first place. Validate the problem first. A brilliant solution to a problem nobody has is still a dead business.

Grade the evidence by what it costs to give

Here is the single idea that fixes most bad validation. Not all evidence is equal. Grade every signal by what it costs the person to give it. The more it costs them, the more it means.

A compliment costs nothing. “That’s a great idea, you should build it” is worth exactly what it took to say, which is nothing, and it usually means the person wants you to feel good and to end the conversation. An opinion about the future costs almost nothing either. “Yeah, I’d definitely buy that” is a prediction people are terrible at and feel no obligation to honour. Move up the ladder and the currency changes. Giving you their time for a proper conversation costs something. Handing over an email address costs a little more. Joining a waiting list, referring a colleague, putting down a deposit, pre-ordering, actually paying: each of these costs the person more, and each is a stronger signal than everything below it.

The practical rule that falls out of this: chase the most expensive signal you can afford to ask for. When a founder tells me their idea is validated, my first question is always the same. What did anyone actually give up to tell you that? If the answer is “they said they liked it”, nothing has been validated yet. A Kansas Small Business Development Center guide puts the same point well: real validation only happens when a prospect sacrifices something of value, whether that is time, data, or money.

Validate the problem before you validate the solution

The cheapest way to waste six months is to fall in love with a solution and skip straight to building it. Steve Blank’s rule for this is blunt and correct: there are no facts inside your building, so get outside. Everything you believe about the problem is a guess until someone who has the problem confirms it.

Problem validation asks whether the pain is real. Is it frequent? Is it expensive or annoying enough that people already spend time or money trying to fix it? What are they using today, however badly, to cope? The presence of an ugly workaround is one of the best signals there is, because it proves the problem is worth effort. Nobody rigs up a spreadsheet, a WhatsApp group and three browser tabs to solve a problem they do not care about.

Demand validation comes second, and asks whether they will choose your answer. This is where you test the actual thing: the offer, roughly the price, the positioning. And this is where you climb the evidence ladder from talk towards commitment.

There is one honest caveat to hold here. Customers are excellent at describing problems they have and poor at reacting to solutions they have never seen. Ask people what they want and, as the line attributed to Henry Ford goes, they will ask for a faster horse. This does not mean you should ignore them. It means you validate the problem by listening, and validate a genuinely novel solution by getting it in front of them in some testable form rather than by asking them to imagine it. More on where this bites in the limits section below.

The methods, ranked by signal strength

There are more validation methods than anyone needs. What matters is not the menu, it is knowing which signal each one actually gives you and how each one lies to you. Here are the ones worth your time, weakest to strongest.

Customer interviews. The foundation. Done well, they tell you whether the problem is real and how people cope today. Done badly, they produce a wall of polite encouragement. The failure mode is asking about your idea instead of their life. Interviews validate the problem far better than the solution.

Search and marketplace demand. Free and fast. If people already type the problem into Google, ask about it in forums, or buy imperfect alternatives on Amazon and Etsy, demand exists in some form. The failure mode is confusing traffic with willingness to pay for your version specifically. Existing demand is necessary, not sufficient.

Competitor and substitute analysis. Counterintuitively, competitors are good news, because they prove someone is paying. The real substitute is often not another product but “doing nothing” or “a spreadsheet”. The failure mode is dismissing an incumbent because your idea is better, without asking whether better is enough to make anyone switch.

Landing page and smoke test. Build a page that describes the offer as if it exists, drive a little traffic to it, and measure how many people click “buy” or hand over an email. The classic example is Dropbox. When Drew Houston could not show a finished product, he posted a short demo video to a community of early adopters, and the beta waiting list jumped overnight from 5,000 to 75,000 people. Eric Ries later made it the textbook minimum viable product. The failure mode is celebrating email sign-ups as if they were sales. An email is a weak signal. It is a start.

Pre-sales and crowdfunding. Now we are on the expensive rungs. Asking people to pay before the thing exists is the strongest pre-build signal you can get, because money is the one currency people do not spend to be polite. A pre-order, a paid deposit, or a funded crowdfunding campaign is real demand. The failure mode is almost none, which is why this is the method to reach for whenever your idea allows it.

Concept testing with purchase intent. When you cannot pre-sell yet, a structured concept test measures how a defined audience responds to the idea and how many say they would buy, read against a benchmark rather than in isolation. It is more rigorous than a casual survey and faster than a pre-sale. The catch is that stated intent overstates real buying, so the number is directional. If a concept test is the right fit, the tooling varies a lot by budget and rigour, and I have compared the main options in the guide to concept testing platforms. Where the open question is specifically price, a structured method like Gabor-Granger tells you more than asking “what would you pay”.

Concierge and Wizard of Oz tests. Deliver the outcome manually before you build the machine. Do the work by hand for your first few customers, or put a human behind what looks like an automated product. You get real paying users and real evidence of what the product must actually do, without writing the software first. The failure mode is running it so long that you build a business you cannot scale, but as a validation step it is one of the most honest going.

How to talk to customers without fooling yourself

Interviews are where most validation quietly fails, because people lie to you, and they lie out of kindness. Rob Fitzpatrick’s The Mom Test is the fix, and the title is the whole idea: ask questions so specific and so grounded in the person’s real life that even your mum could not give you a false positive.

Three rules carry most of the weight. Talk about their life, not your idea, because the moment you pitch, they switch from informant to supporter and the truth stops. Ask about specific things that happened in the past, not opinions about the future, because “how do you currently do this and what did it cost you last time” is answerable and “would you use a tool that does this” is a guess dressed as data. And treat compliments as a warning, not a win. When someone praises the idea, you have learned nothing and probably steered them into it.

The tell that an interview went well is that you heard about problems, workarounds, money already spent, and moments of real frustration. The tell that it went badly is that you feel great and everyone loved it. If you leave every conversation encouraged, you are pitching, not learning.

Set your pass mark before you run the test

The most common way to fool yourself is to run a test with no threshold and then interpret whatever comes back as encouraging. Set the bar first. Before you run anything, write down what result would make you build, what result would make you stop, and what would send you back to change the offer. A number decided in advance is a decision rule. A number interpreted afterwards is a horoscope.

The bar has to account for one stubborn fact: people overstate what they will do. Intentions are a famously unreliable guide to behaviour, and Sheeran and Webb’s review of the intention-behaviour gap documents just how many people who fully intend to act never actually do. In practice this means you discount stated intent hard. If your survey says 40% would “definitely buy”, the real figure that converts is a fraction of that. This is exactly why the expensive signals matter: a pre-order does not need discounting, because the person already acted.

So write the rule in the currency of the strongest signal you can collect. Not “people seemed keen” but “at least 25 of the 200 people who see the landing page leave a deposit”. Then the test can actually tell you no, which is the only thing that makes a yes worth anything.

A worked example: validating demand before building

Numbers below are illustrative, to show the shape of a real validation, not figures from a specific case.

Say you want to launch a refillable, concentrated cleaning brand: you ship a small pod, the customer adds water at home, less plastic, lower shipping weight. Lovely story. Is it a business?

You start with the problem, not the pod. You interview 18 people who buy premium cleaning products, and you ask about their last purchase, not your idea. You learn that most of them do not lie awake about plastic, but a clear subset are annoyed at paying to ship what is mostly water, and several already decant and dilute products themselves. That workaround is your signal. The problem is real for a segment, and it is narrower than you hoped.

Next you check demand cheaply. Search interest for refillable cleaning products is rising, there are two established competitors doing well, and the reviews of both complain about the same thing: the refills are fiddly and run out too fast. Competitors are paying customers who have told you where the gap is.

Now you set a pass mark before spending. You decide that a landing page describing the product must convert at least 8% of paid traffic into email sign-ups, and, more importantly, that a follow-up offer of a founding-customer bundle at a real price must convert at least 3% of those sign-ups into pre-orders. You have chosen an expensive signal on purpose.

You run it. The page pulls 11% into sign-ups, which is encouraging but weak on its own. The pre-order offer converts 4% of them. That second number is the one that matters, because those people paid. You have not proven a company. You have proven enough real demand in a defined segment to justify building a first batch, and you have learned that the message that landed was “stop paying to ship water”, not “save the planet”. That is what a validation is meant to produce: a decision, and a sharper idea than you started with.

Where market validation misleads you

Validation reduces risk. It does not remove it, and pretending otherwise is its own kind of failure. It is also worth resisting the line, common in the guides selling you a framework, that startups which validate simply succeed more often. Some of that is real and some is selection: the founder disciplined enough to validate honestly is usually the same one who prices, builds and sells with more care, so validation quietly takes credit for the whole package. Treat it as one contributor to better odds, not a lever you pull for a guaranteed lift. Five traps are worth naming.

False positives. Enthusiasm is cheap and abundant, especially in surveys and especially from people who like you. A test that only ever collects opinions will tell you almost everything is a good idea. The fix is the whole point of this piece: climb to signals that cost something.

False negatives. The faster-horse problem is real. Genuinely novel ideas can test badly because people cannot picture using something that does not exist in their world yet. This is not a licence to ignore the market and trust your gut. It is a reason to validate the problem rigorously and to get a novel solution in front of people in some concrete, testable form rather than asking them to imagine it.

Sample bias. Friends, family, your existing audience and people who already love your brand are not the market. They are the most flattering slice of it. If your validation only reaches people who are inclined to say yes, you have measured your reach, not your demand.

Validating the problem but not the business. People can genuinely want something that still will not make money. If it costs more to acquire and serve a customer than they are worth, you have validated a hobby. Real validation eventually has to touch price and rough unit economics, not just desire. Marc Andreessen’s point that product-market fit is “the only thing that matters” is right about direction, but a great market still has to be reachable at a price that works.

Validating demand but not the channel. This is the twin of the last one and the trap that kills more “validated” brands than a weak idea ever does. You can prove people want the thing and still have no repeatable, affordable way to reach them. An idea that only sells when you hand-sell it, or only converts on paid traffic that costs more than the customer is worth, has validated appetite and not a business. Somewhere in the process you have to test not just whether people will buy, but whether you can find and acquire them at a price that leaves a margin. Demand you cannot reach profitably is a trap dressed as a win.

A faster demand signal

The honest constraint on everything above is time. Recruiting interviewees, building a landing page, running a pre-sale: each takes days or weeks, and most founders have several ideas competing for that time. This is the gap TestFeed is built for. You can put a concept, a claim, a name, or an in-context price in front of your target audience and get back a purchase-intent score, the reasons behind it in shoppers’ own words, and a clear next move, in days rather than weeks.

Treat it as what it is: a fast, directional signal to decide which ideas deserve a real pre-sale and which to cut before you spend. It is a pre-launch read, not a market forecast or a guaranteed sales number, and it works best as the first filter in front of the expensive, high-commitment tests, not as a replacement for them. The strongest validation is still someone handing over money. This gets you to the ideas worth asking that of, faster.

Frequently asked questions

What is market validation?

Market validation is the process of gathering evidence that real demand exists for an idea before you build it. Instead of asking whether people like the idea, you test whether the people who would actually pay for it behave as though they want it, by talking to them, watching what they already do, and asking them to commit something of value such as time, an email address, or money.

How do you validate a business idea with no money?

You do not need a budget to validate a business idea. Interview 15 to 20 people in your target market about the problem, using past behaviour rather than opinions on your solution. Check whether people are already searching for and paying for alternatives. Then run a small commitment test, such as a landing page that collects email sign-ups or a pre-order, so you measure action rather than enthusiasm. All of this is free or close to it.

What is the difference between market validation and market research?

Market research describes a market: its size, its segments, and its trends. Market validation tests one specific claim, that your idea will find paying customers in that market. Research tells you the water is warm; validation tells you whether this particular boat floats. You usually do light research first to size the opportunity, then validation to decide whether to build.

How many customer interviews do you need to validate an idea?

There is no magic number, but most founders start to hear the same patterns after 15 to 20 focused conversations with the right people. The quality of who you talk to matters far more than the count. Twenty real buyers of the problem will teach you more than a hundred people who are being polite. Stop when new interviews stop surprising you.

Can you validate a business idea with a survey?

A survey is useful for sizing a problem and spotting patterns, but weak as final proof, because stated intent consistently overstates real buying. People who tick ‘definitely would buy’ often never do. Treat survey results as a directional signal, discount the enthusiasm, and confirm demand with a commitment test where people put time or money on the line.

Where to start

Pick your riskiest assumption, the one thing that, if it is wrong, sinks the idea. Usually it is “people will pay for this”. Then design the cheapest test that asks someone to give up something real to prove it, and write down the number that means yes before you run it. If you only do one thing this week, do that. Everything else in market validation is detail on top of it.