How to Validate a Business Idea

Most failed businesses did not fail because the product was bad. They failed because nobody wanted it at the price it cost to make. The expensive part is finding that out after the website is built, the inventory is ordered and the first year of runway is gone.

Validation is the work you do before that. Done properly it takes two to four weeks and costs a few hundred dollars. Done badly it takes an afternoon, feels great, and tells you nothing.

What validation is, and what it is not

Validation is evidence that specific people will pay a specific price for a specific thing. All three words matter.

Here is the test. For evidence to count, the other person has to give up something real: money, a written commitment, or an hour of their working day. If saying yes to you costs them nothing, their yes is worth nothing.

That single rule eliminates most of what people call validation:

• A survey is not validation. Surveys measure what people will click while waiting for a coffee. "Would you use an app that does X?" gets a near-universal yes, because agreeing is free and imagining a better life is pleasant.
• A focus group is not validation. People perform in groups. The loudest opinion becomes the room's opinion within about four minutes.
• Search volume is not validation. It tells you people have the problem, not that they will pay you to solve it rather than keep solving it for free.
• Your own certainty is definitely not validation. You have thought about this for months. Of course it seems obvious.

What does count: someone pays you, signs something, leaves a deposit, or spends an hour of their working day helping you get it right because they want it to exist.

Validation is also not a yes/no verdict on the whole idea. Most ideas are not wrong, they are wrongly aimed. The usual output is "this works, but not for the customer I assumed, at a price higher than I planned, solving the smaller half of the problem."

Liking it and buying it are different behaviours

The most dangerous sentence in early-stage business is "everyone I talked to loved it."

People are generous with enthusiasm because it costs nothing and makes the conversation pleasant. They are stingy with money because it costs money. The gap between those two behaviours is where most first businesses die. Watch for it in how people respond:

So, bluntly: enthusiasm from friends and family is worth nothing as evidence. Not "worth a little". Nothing. Your sister is not evaluating your business, she is being your sister. Your friend at the pub agrees because disagreeing would be rude. Your former colleague is being supportive because you were good to work with.

They are not lying. They genuinely think they mean it. But none of them are in the buying situation — with a budget, an alternative, and a reason to say no — which is the only situation that predicts revenue. If your evidence base is people who would be uncomfortable telling you the truth, you have no evidence base. The same goes for anyone who would like to be paid by the business you are describing.

How to find fifteen people to talk to when you know nobody

Almost everyone gets stuck here, decides research is impractical, and skips straight to building. Here are the methods that actually produce conversations, roughly in order of how well they work.

1. Go where the customer physically is. Selling to plumbers? Stand at the trade counter of a supply house at 7am. Restaurant owners? Go at 3pm, after the lunch rush. Ten minutes asked politely in person works far more often than people expect, because you are clearly not selling anything yet.

2. Ask for introductions, but ask specifically. "Do you know anyone who runs a dental practice?" gets you nothing. "You mentioned your cousin manages a dental practice — could you text her and ask if she would give me fifteen minutes?" gets you a conversation. Specific asks get acted on; vague ones get sympathy.

3. Online communities where your customer already complains. Subreddits, trade Facebook groups, industry forums. Do not post your idea. Read three months of posts, find the people describing the problem in their own words, and message them: "I saw your post about X. I am researching this, not selling anything. Would you spend fifteen minutes telling me how you handle it?"

4. LinkedIn, if you sell to businesses. Search the exact job title, filter by region, send a short note. No pitch, no calendar link in the first message. Fifty messages tends to produce a handful of conversations.

5. Paid recruiting, when your time is worth more than the money. Twenty-five to a hundred dollars per interview as a gift card fills a schedule fast. You are buying their time, not their opinion — but it does bias toward people who want fifty dollars.

Fifteen conversations with the same type of buyer is the target. Not fifteen assorted humans — fifteen people who share a role, a situation and a budget. If you are hearing wildly different things, that is usually not a complicated market, it is you talking to four different markets. And by the tenth conversation you should be able to predict what the person is about to say. When you stop being surprised, you have the pattern.

What to actually ask

The purpose of a customer conversation is to learn what the person has already done about the problem. Past behaviour is data. Future intention is fiction.

So do not describe your idea at the start — describe it at the end, if at all. As soon as you pitch, the conversation stops being research and becomes politeness. Open with something like: "I am trying to understand how people handle [problem]. I am not selling anything. Can you walk me through the last time it came up?"

Questions that produce useful answers

1. Walk me through the last time this happened. What did you do?
2. How often does that happen — weekly, monthly, twice a year?
3. What did you do about it? What are you using now?
4. How did you find that solution? What were you searching for?
5. What does it cost you today — in money, hours, or missed work?
6. What have you already tried that did not work? Why did you stop?
7. Who else is involved when you decide to buy something like this?
8. What did you spend on it last year?
9. What would have to be true for you to change what you are doing now?
10. Is this in your top three problems this quarter, or somewhere below that?
11. Who is the worst at handling this, in your industry?
12. If I built something for this in the next three months, would you want to see it early — and would you be willing to put money down to be first?

Question 5 gives you the price ceiling. Question 8 is the most useful in the list, because a real number spent last year is proof of budget. Question 10 is the one people skip, and it explains why a genuinely useful product still fails to sell: it solves problem number nine.

Questions that produce useless answers

• "Would you use this?" — Yes. Everyone says yes.
• "Do you think this is a good idea?" — You are asking them to be a critic, and they will be a kind one.
• "How much would you pay for this?" — People are terrible at this and answer with whatever feels polite.
• "Would you pay $49 a month for this?" — Anchoring your own price and asking permission. Ask what they pay now instead.
• "What features would you want?" — You have just hired an untrained product manager who does not live with the consequences.
• "Does this make sense?" — Of course it does. You explained it.
• Anything beginning "Wouldn't it be great if…"

The pattern: any question that asks someone to imagine the future gets an imagined answer. Any question about what they actually did gets a fact.

Take notes in their words, not yours. The exact phrasing people use for their own problem becomes the copy on your website later, and you will not remember it accurately by evening.

How to read the signal

After fifteen conversations, sort what you heard by what it cost the person to say.

Strong signals

• They tried to buy it during the conversation.
• They have already paid for a bad alternative, and can tell you exactly how much.
• They have built a spreadsheet or a manual workaround to cope. Duct tape is the clearest evidence of real pain.
• They introduce you to someone else unprompted, or ask when it will be ready twice.
• They describe the problem with a number attached: hours per week, dollars per month, jobs lost.

Weak signals

• "That's clever."
• They agree with every problem you name, including contradictory ones.
• They talk about the industry in general rather than their own last month.
• The enthusiasm is about the technology rather than the outcome.

Warning signs

• Everyone loves it and nobody has ever spent money on the problem.
• The person who is enthusiastic is not the person who signs.
• You have to explain the problem before they recognise it. If you are teaching people they have a problem, you are funding market education, which is the most expensive marketing there is.

One useful discipline: before you start, write down what result would make you stop. "If fewer than three of fifteen have paid for a solution before, I stop." Written in advance, that number is analysis. Written afterwards, it is negotiation with yourself, and you will lose.

Pre-selling: the only signal that settles the argument

Everything above is preparation for this. The strongest evidence available is someone giving you money before the thing exists, and it comes in these forms, lightest first:

• A deposit. Fifty to five hundred dollars, refundable, to reserve a place, an early rate or a delivery slot.
• A pre-order at full price, with a stated delivery date and a refund policy in writing.
• A paid pilot. Common in business-to-business. They pay a reduced fee for the first version, knowing it is early, in exchange for shaping it.
• A letter of intent with a number in it. Weaker than cash, but on long sales cycles it may be the only thing available inside your timeframe.
• Doing the work manually for a paying customer. Before you build software, deliver the outcome by hand for three customers who pay. Slow, unscalable, and the most complete validation available — you learn the real cost of delivery at the same time.

Three rules for taking money before you can deliver: say plainly that it does not exist yet and give a date; refund anyone who asks, immediately and without argument; keep the money aside until you deliver, because it is not revenue yet.

Founders resist this step harder than any other, for one reason: a pre-sell attempt can fail clearly, and a survey cannot. That is exactly why it is worth doing. Ten polite noes to a real offer beat a hundred yeses to a hypothetical one, because the noes come with reasons, and the reasons are the product spec.

If you cannot get anyone to pay a deposit, that is the answer. Not a sign you need better marketing — the answer.

Smoke tests and landing pages, done honestly

A smoke test is a page that describes the offer as though it exists and measures how many people try to buy. Fast, cheap, and easy to do dishonestly. The honest version:

• The page describes the real offer at the real price, not a vague "join the waitlist for something exciting."
• The button says what it does. "Reserve a place", not "Buy now" leading to a fake checkout you never intended to honour.
• If you take payment details, you either deliver or refund on a stated date.
• Everyone who signs up hears from you within a week, even if the answer is "not enough people were interested, so I am not building it."

Taking real card details for a product you have no intention of building is fraud, not a tactic. The honest version tells you nearly as much anyway.

Budget about $300 to $600 of ad spend to a specific audience over a week or two, sending three hundred to a thousand targeted visitors. Below a hundred visitors the numbers mean nothing.

Conversion rates vary too much by industry for a universal benchmark, but the shape of the result is usually clear without one. A cold page converting in the low single digits to a real commitment is a live signal. A page converting near zero after several hundred well-targeted visitors is telling you something — though before concluding the idea is dead, check that the traffic really was your buyer and that the page uses the words your interviewees used.

The trap: a smoke test measures your page and your traffic as much as your idea. It is a good second test after conversations, and a poor first one, because a low number gives you no explanation. Interviews tell you why. Smoke tests tell you how many.

Competitor analysis that tells you something useful

Most competitor research is a grid of feature ticks nobody acts on. Skip it. Ask four questions instead.

1. What do their customers complain about? Read one-star and three-star reviews, support forums, cancellation threads. Three-star reviews are the goldmine: people who wanted it to work and will explain exactly where it did not. That is the gap you can build in.

2. What do they charge at each tier? This is your pricing reality. If three established competitors are at $79/month, your $400/month needs a specific reason, and "better service" is not one.

3. How do they get customers? Search ads, trade shows, referral, cold calling, retail shelf. That is the channel your market responds to, and it comes with a cost you will also pay. If everyone in the category sells through expensive outside reps, you cannot plan to win with a website.

4. Who left, and where did they go? Search "[competitor] alternative" and "switching from [competitor]". The answers show you the real switching triggers.

And the most misread result in the whole exercise: no competitors is bad news, not good. It usually means people have tried and the market did not pay. Spend an hour finding the companies that tried this five years ago and are now gone, and work out what killed them. Competitors are proof of a budget. Your job is to be different in a way a specific customer cares about — not to be alone.

The methods, compared

Run them in that order and stop as soon as something works. If two people pre-pay in week one, you do not need the smoke test. The point is not to complete the list, it is to reach a decision you can defend.

Deciding: go, adjust, or stop

Write the criteria down before you start, then judge honestly against them.

Go — build the first version — when most of this is true:

• Ten or more of fifteen interviewees described the problem without prompting.
• At least a third of them are already paying for something to handle it, however badly.
• You have three or more paid commitments, deposits, or signed pilots.
• The price you need is inside the range they already spend.
• You can name the first ten specific people or companies you will sell to.

Adjust — change one thing and test again — when the problem is clearly real but demand is not landing. Usually one of four things is wrong: the customer (right problem, wrong segment), the price (right thing, wrong number), the scope (the whole thing when they wanted a piece), or the urgency (real, but ranked eighth). Change one variable, run five more conversations. In our experience this is the most common outcome by a wide margin, and it is not failure.

Stop when nobody has ever paid to solve this, the enthusiasm evaporates the moment money enters the conversation, acquisition obviously costs more than anyone will pay, or you cannot find fifteen people to talk to. That last one is decisive on its own — if you cannot reach fifteen prospects for free while researching, reaching thousands while selling will not be easier.

Stopping is a legitimate outcome. Killing an idea in week three costs a few hundred dollars and some pride. Killing it in month fourteen costs the savings, the relationships and the year. If go is the answer, the work turns into getting the first paying customer.

What validation cannot tell you

It cannot tell you whether you will execute well, whether the market will still be there in three years, or what a funded competitor does next spring. It does not replace the judgement call you make with incomplete information. It moves that call from "I have a feeling" to "here is what fifteen buyers said and what four of them paid" — the difference between a calculated risk and a guess, and worth two to four weeks of anyone's life.

Then the question becomes what to build first — the smallest thing that delivers the outcome, covered in turning the idea into an MVP. If the problem was real but the money was not, revisit how the business is meant to make money before you revisit the idea. And if you would rather not run this alone, pressure-testing the idea is the first thing we do in a blueprint engagement.

The short version

• Validation is evidence that specific people will pay a specific price. If saying yes costs them nothing, it is not evidence.
• Friends and family enthusiasm is worth zero, as is any opinion from someone who would find it awkward to tell you no.
• Talk to fifteen people who share a role and a budget. Ask what they already did about the problem, never what they would do about your idea.
• The strongest signal is money before the product exists: a deposit, a pre-order, a paid pilot, or delivering it by hand for three paying customers.
• Smoke tests measure how many; interviews explain why. Run interviews first, and never take money you do not intend to honour.
• Write your stop criteria before you begin. Most ideas need adjusting rather than killing — but decide against a number you set in advance, not a feeling you have afterwards.