How to Validate a Business Idea

Validating a business idea means collecting evidence, before you commit serious time or money, that a specific group of people has the problem you want to solve and will act to solve it. Interest is not enough. The useful test is whether enough of them commit something real: their time, an introduction, a deposit or a payment.

I think of validation as a single decision rather than a stage with homework: is there enough evidence to keep going? Most founders I talk to did some form of validation. The ones who regret it rarely say they skipped it. They say they ran a test that was built to come back positive, then treated the result as permission.

What validation actually decides

Validation does not prove a business will work. Nothing at this stage can. What it does is reduce one specific risk: building something nobody wants badly enough to pay for. That risk is large. In a March 2026 analysis of 431 venture-backed startups that shut down since 2023, CB Insights found that poor product-market fit was cited in 43% of the 385 cases where it could identify a reason. Running out of capital was cited more often, at 70%, but the report notes that cash is usually the last cause of death rather than the root problem. These are venture-backed companies, so the numbers do not map neatly onto a small or solo business, but the pattern is familiar at every size: money runs out faster when the market was never there.

Steve Blank, who developed the customer development method that much of modern validation practice grew from, put the core idea in a 2013 Harvard Business Review article: founders begin with untested guesses, and the job is to leave the office and test them with real buyers, users and partners. His phrase for it, “get out of the building,” has outlived most of the frameworks built on top of it.

So the decision at this milestone has three possible outcomes, not two:

1. Keep going. The evidence is strong enough to justify building, selling or launching properly.
2. Change something. The problem is real but the customer, the offer, the price or the channel is wrong. You test again with the change.
3. Stop. The evidence says people do not care enough, and another round of testing would only be a way to avoid hearing it.

The evidence ladder: which signals count

Not every “yes” is the same yes. The most useful way I know to sort signals is by what they cost the person giving them. Rob Fitzpatrick’s book The Mom Test frames this as commitment: people pay in time, reputation or money, and the more they give up, the more seriously you can take what they say. Compliments, in that framing, are close to worthless because they cost nothing.

Signal	What it costs the customer	What it tells you	Main weakness
Compliments, “I’d use that”	Nothing	Very little	People are polite, especially to founders
Survey answers about future behavior	A few minutes	What people think they might do	Predictions about the future are unreliable
Email signups on a landing page	An email address	Curiosity at a stated promise	Cheap to give, easy to forget
Detailed accounts of past behavior in interviews	An hour	Whether the problem is real and how they solve it now	Says little about your price
A second meeting, a trial, feedback on a prototype	Time	Ongoing interest	Interest is still not purchase
Introductions to a colleague or decision maker	Reputation	They believe it is worth someone else’s time	Can be a polite way to pass you on
Letter of intent	Reputation, some legal and internal effort	Serious intent from a business buyer	Usually not a promise to pay
Deposit or pre-order	Money	Willingness to pay at a real price	Can be refunded, and brings legal duties
Paid pilot or first paid use	Money and time	The strongest early signal you can get	Takes longer to arrange

Two things follow from the table. First, a pile of low-cost signals does not add up to one high-cost signal. Five hundred signups do not prove that fifty people will pay. Second, the top of the ladder is not always reachable before you build. A hardware product, a regulated service or a marketplace may need a working version before anyone can pay. In that case, climb as high as the idea allows and be honest about where you stopped.

What the decision depends on

The right amount of validation is not fixed. It depends on a handful of factors, and founders get into trouble when they borrow a method built for a different situation.

How much the first version costs. If you can build a usable first version in a weekend, building it may be the cheapest test you have. If it takes six months and your savings, validate much harder first. The cost of being wrong should set the bar for evidence.

Who the buyer is. Business buyers can sign letters of intent, run paid pilots and tell you their budget. Consumers mostly cannot, so consumer tests lean on pre-orders, landing pages and early usage.

How reversible the next step is. Quitting a job, signing a lease or hiring are hard to undo. Writing code on evenings and weekends is not. The less reversible the step that follows, the more proof you need before taking it. That is why this milestone comes before many of the commitments covered in Should You Quit Your Job to Start a Business?.

How long you can afford to look. Validation has a cost too. Founders with little runway sometimes need to sell a rough version quickly, and that sale is the validation.

Whether the market already exists. If people already pay for alternatives, the question is not “is there demand” but “will they switch to you.” That is a different and often easier test: look at what they pay now and ask what it would take to move.

The common ways founders validate

Most validation methods fall into a few families. They are not rivals; many founders use two or three in sequence, from cheapest to most convincing.

Method	Best for	What it can prove	Typical trap
Problem interviews	Any idea, as a first step	The problem exists and how people handle it today	Pitching instead of listening
Landing page or smoke test	Consumer and self-serve products	People respond to the promise	Treating signups as sales
Concierge or manual service	Services, software that can be faked by hand	People will use and pay for the outcome	Never automating, or automating too soon
Minimum viable product	Software, simple physical products	People use the real thing and come back	Building far more than needed
Pre-sale or deposit	Products with a clear delivery date	People will pay before it exists	Promising a date you cannot meet
Paid pilot	Business buyers	A company will pay for a defined result	Pilots that never convert to contracts

There are well-documented cases for most of these. Food on the Table, a meal-planning company in Austin, began with a single customer, a parent whose menus the founders planned by hand at a weekly meeting in a grocery store parking lot, and automated only after they had learned what worked, according to an interview Eric Ries gave Inc. in 2012. In the same interview Ries described how Groupon, after spending its first funding on a different product, ran a single deal for a pizza place in its own building by hand. Dropbox tested demand with a narrated demo video before the product was ready for the public; Drew Houston said the beta waiting list went from 5,000 to 75,000 people overnight, as Ries recounted in TechCrunch. Each of these tested one thing cheaply rather than everything at once.

Where founders usually get it wrong

The mistakes that come up in conversations with founders are not exotic. They are small habits that make a test unable to fail.

• Asking people who love you. Friends, family and former colleagues want you to succeed. Their encouragement is real and useless as evidence.
• Pitching instead of asking. The moment you describe your solution, people start reacting to you instead of describing their life. Ask how they handled the problem the last time it came up and what it cost them.
• Counting the wrong thing. Signups, likes and survey percentages feel like data. Unless someone gave up time or money, they are mostly noise.
• Validating the problem but not the price. Many founders confirm that a problem is painful and never test whether anyone will pay to solve it. Painful and paid are different things.
• Moving the goalposts. Twenty interviews were the plan. The first twenty were lukewarm, so the target became fifty, then “the right segment.” If the bar moves after you see the results, it was never a bar.
• One enthusiastic customer standing in for a market. A single big yes is encouraging. It is also a sample of one.
• Endless validation. The opposite failure. Some founders keep testing because each test postpones the risk of building and selling. At some point the only test left is the real thing.

Set the kill line before you start

The single habit that most improves validation is cheap: before running a test, write down what result would make you keep going, change something or stop. “If fewer than five of the next twenty people I speak to agree to a paid pilot, I change the customer segment.” “If the pre-sale page converts fewer than one in fifty visitors who reach it, I stop.” The numbers are yours to set, and they should come from your economics: what a customer is worth, what it costs to reach one, how long you can wait.

A written line protects you from the most human part of validation, which is wanting the answer to be yes. It also makes the result useful even when it is bad. A test that clearly fails gives you a reason to change course. A test with no bar gives you a mood.

The two tools this milestone leans on

Two articles in this section go deeper into the methods founders use most often at this point.

How to Build an MVP covers where the term came from, the main types (concierge, Wizard of Oz, landing page and single feature), the documented cases behind them, and the hardest question: how small is small enough to learn something real without embarrassing yourself into silence.

How to Pre-Sell Before You Build covers pre-orders, deposits, letters of intent and paid pilots, the strongest signals you can get before a product exists. It also covers the obligations that come with taking money early, including the US Federal Trade Commission’s rule on shipping times and refunds, and how to be honest with buyers when things slip.

Other topics planned for this milestone include testing an idea in a week, recognizing when to kill an idea, and the validation mistakes founders regret.

How to know you are ready for the next milestone

You are ready to move on to finding customers in earnest when most of these are true:

• You can describe the person who has the problem specifically enough that you could find ten more of them this week.
• Several of those people have given you something that cost them: a payment, a deposit, a signed pilot, a serious introduction.
• You know roughly what they pay today, in money or time, to deal with the problem.
• You have a first version, manual or automated, that delivers the result, even if it is rough.
• You set a bar before your last test, and the result cleared it.

If only the first item is true, you have a hypothesis, not validation. That is a fine place to be, as long as you call it that. If your idea is still loose, it may be worth going back to How to Evaluate a Business Idea before spending more on tests.

Where this sits on the route

Validation is milestone 4 and the start of the Launch phase. Before it comes Should You Quit Your Job to Start a Business?, the last decision in the Decide phase, and many founders find it easier to make that call once they have real evidence from this one. After it comes How to Get Customers for a New Business, where the question changes from “does anyone want this?” to “how do I reach the people who do, again and again?”