Validate a business idea by naming the one assumption that would sink it if it were wrong, then running the cheapest test that could prove it wrong within a few weeks. Compliments, surveys and sign-ups are weak evidence. People changing what they do, or paying you, is strong evidence. Decide before you start what result would change your mind.
On an Indie Hackers thread about validation, one co-founder described shops that told him they wanted his product and then didn’t do anything about it for nine months. Some of them never did. Another founder ran Google, Facebook and Reddit ads and got click-throughs but no sign-ups (Indie Hackers).
Both of them had evidence, of a kind that can’t tell you whether a business will work. That gap is expensive: when CB Insights studied 101 failed startups, 42% of them had failed because there was no market need for what they built (DigitalOcean, citing CB Insights).
Why most validation advice doesn’t help
Search for this and you’ll find lists of eight or twelve steps: research the market, study competitors, run a survey, build a landing page, check Google Trends. Each step has its uses, and you can do every one of them, feel thoroughly validated, and still have no idea whether anyone will pay.
Most of those steps measure interest, which people give away freely. They’ll say yes to almost anything that doesn’t cost them anything today.
A useful validation process picks the one belief your idea can’t survive without, tests it with something that costs the customer a little, and decides in advance what answer would make you stop.
What counts as evidence?
Some evidence is much stronger than the rest. A rough ladder, weakest first:
| Rung | What you have | Example | How much it tells you |
|---|---|---|---|
| 1 | What people say they’d do | “I’d definitely use that” | Very little |
| 2 | What people have already done | “Last month I spent two days on this and paid a freelancer to fix it” | A lot about the problem |
| 3 | What people commit | A second call, an introduction to their boss, a signed letter of intent | Real interest |
| 4 | What people pay | A deposit, a pre-order, a paid first month | Demand |
Most founders stop at rung 1 because it feels good. The work of validation is climbing to rung 3 or 4 with as little building as possible.
Rung 2 is the one people skip, and it’s the most useful conversation you can have. Ask about the last time the problem happened: when it was, what it cost, what they did about it. People are poor at predicting what they’ll do and quite good at remembering what they did.
Start with the assumption that would sink you
Every idea rests on a stack of beliefs. Three of them sit under almost every business:
- A specific kind of person has this problem badly enough to act.
- They’ll pay you, at a price that works, to solve it.
- You can reach them without spending a fortune.
Write down every belief your idea depends on. Then give each one two scores from 1 to 5: how much it would cost you if it turned out to be wrong, and how confident you honestly are that it’s right. Be strict with the second score: “two people seemed keen” earns a 2 at most.
The belief with a high cost and low confidence is your riskiest assumption. Test that one first. It’s usually the one you’ve been avoiding, because it’s the one that could end the idea.
Run the cheapest test that could prove you wrong
Once you know what you’re testing, pick the smallest test that could give you a clear no. Some good ones:
- Conversations about the last time. Ten conversations with people who match your first customer, asking about what already happened, never about what they would do.
- A one-page offer with a price. Who it’s for, what they get, what it costs, how to start. Send it to people from those conversations and ask for a decision.
- A deposit or pre-order. Even a small amount shows you who’s serious.
- The manual version. Deliver the result by hand, with spreadsheets and email, before you build anything. If nobody will pay for the manual version, software rarely changes that.
The test should take days or a few weeks. If it needs six months of building first, it’s too big. Shrink the offer until you can put it in front of a real person this month.
Decide what would change your mind, before you start
This step is the easiest to skip, and it’s the one that keeps the test honest. Before you run the test, write one sentence: “If this happens, I’ll stop or change the idea.”
For example: “If fewer than two of the ten owners I speak to have spent money on this in the last year, I’ll narrow the customer or stop.”
Without that sentence, every result looks like encouragement. Three sign-ups become “early traction”. A polite no becomes “they’re not ready yet”. Writing the line down first means the test can actually tell you something.
When the results come in, you have four honest options: go ahead, narrow to a smaller customer, change the offer, or pause. Any of the four is a good outcome. Carrying on for months without deciding is the expensive one.
What the usual advice gets wrong
“Talk to 10 people.” Ten is a reasonable number. Who you talk to and what you ask matter more: ten friends telling you it’s a great idea are worth less than three strangers describing what the problem cost them last month.
“Run a survey.” A survey collects rung 1 evidence at scale. It can help you size something you’ve already heard in conversations, but it can’t tell you whether anyone will pay.
“Look at the competition.” Competitors prove a market exists. Whether this customer would switch to you is a separate question, and it’s the one you need answered.
“Just build it and see.” Sometimes that’s right, when building takes a weekend and the only question is whether people will use it. When building takes months, it’s the most expensive test you could choose.
The idea: part-time HR for creative agencies with 10 to 30 staff.
The assumption that would sink it: agency owners will pay around $2,000 a month for help they currently do themselves. Cost if wrong: 5. Honest confidence: 2, because the only evidence is two owners who “seemed keen”.
The cheapest test: ask five owners what they spent on HR last year, in money and in their own time. Then offer one of them a specific paid first month.
What would change your mind: if four of the five say they’d rather use a template and a lawyer when they need one, you narrow to agencies that just lost a hire badly, or you stop.
The idea: a booking and reminder app for independent dog groomers.
The assumption that would sink it: no-shows cost groomers enough that they’ll pay a monthly fee to cut them. Cost if wrong: 5. Honest confidence: 3, because groomers complain about no-shows online, but complaining is free.
The cheapest test: ask eight groomers how many no-shows they had last month and what they do about them now. Then offer to run their reminders by hand for two weeks for a small fee.
What would change your mind: if most of them already use a free reminder tool and lose fewer than two bookings a month, the pain is too small to sell. Look for the groomers who lose more, or move on.
Presold about $5000 worth of revenue before committing.
How long should it take?
Probably less than you’re giving it. You can decide whether an idea is worth testing in a couple of days, by writing down what you know, what you’re guessing, and the one question that matters most. The test itself should give you an answer within a few weeks.
That’s the logic behind Humble’s 30 days: a short run of small daily jobs, where the first week is spent on this exact work, before anything gets built.
Questions
How many people do I need to talk to?
Keep going until the answers start repeating, which for one tightly defined customer is often somewhere between five and ten conversations. If the answers are all over the place after ten, your customer is too broad. Narrow it and start again.
Is a waitlist proof that people want my product?
It's proof of interest. Joining a waitlist costs a click. Ask some of the people on it for something that costs more, like a call, an introduction or a deposit, and see how many say yes.
Can I validate an idea without building anything?
Usually, yes. Describe the offer on one page with a price, show it to people who match your first customer, and ask for a commitment. For many ideas you can also deliver the first version by hand before you build any software.
Should I use a survey?
Surveys tell you what people say they would do, which is the weakest kind of evidence. They're useful for sizing something you've already seen in conversations. They can't tell you whether anyone will pay.
How long should validation take?
Days to decide whether an idea is worth testing, and a few weeks to get a real answer. If you've been validating for three months, you're probably avoiding the test that would give you a straight answer.
What if the test says no?
A no is a result. You have four options: go ahead anyway with a clear reason, narrow to a smaller customer, change the offer, or pause. A no in week three is much cheaper than a no after six months of building.
Is your idea worth 30 days? Find out free.
Five questions about your idea and a straight answer. It's week zero of Humble's 30 days, and it costs nothing.
Take the free first step