Sep 21, 2026
By Victor Teran
How to Validate a Startup Idea Without Building It
Most startup idea validation is theatre. Founders ask people whether they would use a thing, everyone is encouraging, and encouragement costs nothing to give. Four questions produce real information, and none of them is about your idea.
Build
Zero to One
Research
The problem with asking people about your idea is that they are answering a different question than the one you asked. You want to know whether they will change their behaviour. They hear a request for support from someone who is clearly excited, and they give it, because they are not monsters.
So the encouraging conversations pile up, the idea survives, and the first real test arrives eighteen months later when a stranger has to decide with their own money and no relationship to you.
Everything below is designed to get that test earlier and cheaper. The stakes are set by the base rate: 4.6% of newly launched subscription apps reach $10,000 in monthly recurring revenue within two years.
Key takeaways
Ask about what people already do, not about what they would do. Past behaviour is evidence, intention is politeness.
The strongest signal is discovering they have already built a bad workaround. That means the problem is real and already has a budget.
If nobody currently spends money, time or effort on this problem, you are not competing with a rival, you are competing with indifference, which is harder.
The output of validation is not a yes or a no. It is a sharper definition of who you are for.
Why does asking "would you use this?" fail?
Because the honest answer to a hypothetical is always yes, and it commits nobody to anything.
There is no cost to agreeing. There is a social cost to telling a visibly invested founder that their idea is unremarkable. So the data you collect is a measurement of politeness with your idea's name attached.
Worse, it is directionally consistent. Everyone is encouraging, so you conclude the market is broadly positive, when what you have measured is that people are broadly kind.
Replace the hypothetical with history. Not would you use a tool that does X, but what did you do the last time X happened? The second question has a factual answer, and the answer is either a story with specifics in it or an admission that X does not really happen.
What are the four questions that actually work?
They are all about the past, and none of them mentions your product.
1. When did this last happen to you? If they cannot place it, the problem is not frequent enough to build a business on. Vague frequency is a red flag; people remember real pain with dates attached. 2. What did you do about it? This is the important one. You are looking for a workaround. A spreadsheet, a manual process, a freelancer, a Zapier chain held together with hope. 3. What did that cost you? In money, hours, or errors. This is your value ceiling. A problem that costs someone twenty minutes a quarter cannot support a subscription however elegant your fix is. 4. Why didn't you solve it properly? The answer tells you what you are actually competing with, which is often not a product. It is budget, priority, or the fact that the workaround is good enough.
Notice that a person can answer all four without ever hearing what you are building. That is the point. The moment you describe your solution, you have converted a witness into a supporter.
FREE GUIDE
The 4 product leaks costing you growth
A short audit guide for founders. Find the four places your product leaks revenue, and what to fix first.
What does a strong signal look like?
Someone has already built a bad version of your product for themselves.
This is the finding worth reorganising a plan around. A hacked-together spreadsheet, a manual weekly routine, a paid freelancer doing something a script should do. It means the problem was painful enough to spend real effort on, without anyone selling them anything.
What you hear | What it means |
|---|---|
"We built a spreadsheet for that" | Strong. Real problem, existing effort, quantifiable cost |
"We pay someone to do it manually" | Strongest. There is already a budget line to redirect |
"We tried a tool but stopped" | Good. Find out precisely why they stopped |
"That would be useful" | Nothing. Politeness with no behaviour behind it |
"We just live with it" | Weak. Real problem, no urgency, hard first sale |
The bottom two rows are where most validation ends up, and they are frequently misread as encouragement. "That would be useful" is the sound of someone helping you feel good.
Can an AI validator do this for you?
No, because it has no access to the only evidence that counts: what people already did.
Two of the tools ranking for this question will score your idea in about two minutes. What they are scoring is your description of the idea, which is your own reasoning handed back to you with a number attached. That is a market-sized guess, not validation.
The distinction is simple. A model can tell you the category looks crowded or the wording is vague, both of which are useful before you write an outreach email. It cannot tell you that the last time this problem hit, a real operations lead built a spreadsheet and has been maintaining it every Friday since. That fact is the validation, and it only exists in a conversation.
Use the tools to sharpen the questions. Then go and ask ten people.
How many conversations is enough?
Enough that you stop hearing new things, which is usually fewer than you fear and more than you have done.
The signal is repetition. Y Combinator's own account of finding product-market fit describes the same thing arriving as a pattern rather than a verdict (YC, How to Find Product-Market Fit). When the fourth person in a row describes the same workaround in the same words, you have found a pattern rather than a person. That is typically somewhere between eight and fifteen conversations inside one narrow segment, and it collapses quickly if you keep changing who you talk to.
That last point is where most founders lose the thread. Ten conversations across ten different kinds of company produces ten unique answers and no pattern. Ten conversations with the same role at the same stage in the same vertical produces a definition.
Narrow deliberately. You are not limiting your market, you are choosing where to be undeniably right first. That narrow definition is also what makes an activation event definable later, because you will know whose behaviour you are measuring.
What is the output supposed to be?
A sharper sentence about who you are for, not a verdict on your idea.
Validation rarely returns a clean yes or no. It returns something more useful: the discovery that your idea is compelling to a narrower group than you assumed, for a slightly different reason than you assumed.
The sentence to walk out with: for [specific person] with [specific recurring problem], we [do the thing] so they get [outcome] instead of [current workaround]. Every clause should be traceable to something a real person told you, and the last one should name the workaround you actually heard.
That sentence then decides what you build first, which we have written about separately in what to build first. It also decides what you do not build, which is the part that saves the runway.
It is worth being honest that this only narrows the risk rather than removing it. Dalton Caldwell's account of Socialcam is the standing reminder that enormous demand signals can sit on top of no retention at all (YC, The Real Product-Market Fit), and no amount of pre-build research substitutes for a cohort that comes back.
And if you cannot fill in the last clause, that is the finding. A problem with no current workaround is usually a problem nobody is trying to solve, and the reason is rarely that everyone else missed it.
WHAT NEXT
Want this fixed in your product, not just explained?
Sep 21, 2026
By Victor Teran
How to Validate a Startup Idea Without Building It
Most startup idea validation is theatre. Founders ask people whether they would use a thing, everyone is encouraging, and encouragement costs nothing to give. Four questions produce real information, and none of them is about your idea.
Build
Zero to One
Research
The problem with asking people about your idea is that they are answering a different question than the one you asked. You want to know whether they will change their behaviour. They hear a request for support from someone who is clearly excited, and they give it, because they are not monsters.
So the encouraging conversations pile up, the idea survives, and the first real test arrives eighteen months later when a stranger has to decide with their own money and no relationship to you.
Everything below is designed to get that test earlier and cheaper. The stakes are set by the base rate: 4.6% of newly launched subscription apps reach $10,000 in monthly recurring revenue within two years.
Key takeaways
Ask about what people already do, not about what they would do. Past behaviour is evidence, intention is politeness.
The strongest signal is discovering they have already built a bad workaround. That means the problem is real and already has a budget.
If nobody currently spends money, time or effort on this problem, you are not competing with a rival, you are competing with indifference, which is harder.
The output of validation is not a yes or a no. It is a sharper definition of who you are for.
Why does asking "would you use this?" fail?
Because the honest answer to a hypothetical is always yes, and it commits nobody to anything.
There is no cost to agreeing. There is a social cost to telling a visibly invested founder that their idea is unremarkable. So the data you collect is a measurement of politeness with your idea's name attached.
Worse, it is directionally consistent. Everyone is encouraging, so you conclude the market is broadly positive, when what you have measured is that people are broadly kind.
Replace the hypothetical with history. Not would you use a tool that does X, but what did you do the last time X happened? The second question has a factual answer, and the answer is either a story with specifics in it or an admission that X does not really happen.
What are the four questions that actually work?
They are all about the past, and none of them mentions your product.
1. When did this last happen to you? If they cannot place it, the problem is not frequent enough to build a business on. Vague frequency is a red flag; people remember real pain with dates attached. 2. What did you do about it? This is the important one. You are looking for a workaround. A spreadsheet, a manual process, a freelancer, a Zapier chain held together with hope. 3. What did that cost you? In money, hours, or errors. This is your value ceiling. A problem that costs someone twenty minutes a quarter cannot support a subscription however elegant your fix is. 4. Why didn't you solve it properly? The answer tells you what you are actually competing with, which is often not a product. It is budget, priority, or the fact that the workaround is good enough.
Notice that a person can answer all four without ever hearing what you are building. That is the point. The moment you describe your solution, you have converted a witness into a supporter.
FREE GUIDE
The 4 product leaks costing you growth
A short audit guide for founders. Find the four places your product leaks revenue, and what to fix first.
What does a strong signal look like?
Someone has already built a bad version of your product for themselves.
This is the finding worth reorganising a plan around. A hacked-together spreadsheet, a manual weekly routine, a paid freelancer doing something a script should do. It means the problem was painful enough to spend real effort on, without anyone selling them anything.
What you hear | What it means |
|---|---|
"We built a spreadsheet for that" | Strong. Real problem, existing effort, quantifiable cost |
"We pay someone to do it manually" | Strongest. There is already a budget line to redirect |
"We tried a tool but stopped" | Good. Find out precisely why they stopped |
"That would be useful" | Nothing. Politeness with no behaviour behind it |
"We just live with it" | Weak. Real problem, no urgency, hard first sale |
The bottom two rows are where most validation ends up, and they are frequently misread as encouragement. "That would be useful" is the sound of someone helping you feel good.
Can an AI validator do this for you?
No, because it has no access to the only evidence that counts: what people already did.
Two of the tools ranking for this question will score your idea in about two minutes. What they are scoring is your description of the idea, which is your own reasoning handed back to you with a number attached. That is a market-sized guess, not validation.
The distinction is simple. A model can tell you the category looks crowded or the wording is vague, both of which are useful before you write an outreach email. It cannot tell you that the last time this problem hit, a real operations lead built a spreadsheet and has been maintaining it every Friday since. That fact is the validation, and it only exists in a conversation.
Use the tools to sharpen the questions. Then go and ask ten people.
How many conversations is enough?
Enough that you stop hearing new things, which is usually fewer than you fear and more than you have done.
The signal is repetition. Y Combinator's own account of finding product-market fit describes the same thing arriving as a pattern rather than a verdict (YC, How to Find Product-Market Fit). When the fourth person in a row describes the same workaround in the same words, you have found a pattern rather than a person. That is typically somewhere between eight and fifteen conversations inside one narrow segment, and it collapses quickly if you keep changing who you talk to.
That last point is where most founders lose the thread. Ten conversations across ten different kinds of company produces ten unique answers and no pattern. Ten conversations with the same role at the same stage in the same vertical produces a definition.
Narrow deliberately. You are not limiting your market, you are choosing where to be undeniably right first. That narrow definition is also what makes an activation event definable later, because you will know whose behaviour you are measuring.
What is the output supposed to be?
A sharper sentence about who you are for, not a verdict on your idea.
Validation rarely returns a clean yes or no. It returns something more useful: the discovery that your idea is compelling to a narrower group than you assumed, for a slightly different reason than you assumed.
The sentence to walk out with: for [specific person] with [specific recurring problem], we [do the thing] so they get [outcome] instead of [current workaround]. Every clause should be traceable to something a real person told you, and the last one should name the workaround you actually heard.
That sentence then decides what you build first, which we have written about separately in what to build first. It also decides what you do not build, which is the part that saves the runway.
It is worth being honest that this only narrows the risk rather than removing it. Dalton Caldwell's account of Socialcam is the standing reminder that enormous demand signals can sit on top of no retention at all (YC, The Real Product-Market Fit), and no amount of pre-build research substitutes for a cohort that comes back.
And if you cannot fill in the last clause, that is the finding. A problem with no current workaround is usually a problem nobody is trying to solve, and the reason is rarely that everyone else missed it.
WHAT NEXT
Want this fixed in your product, not just explained?
Sep 21, 2026
By Victor Teran
How to Validate a Startup Idea Without Building It
Most startup idea validation is theatre. Founders ask people whether they would use a thing, everyone is encouraging, and encouragement costs nothing to give. Four questions produce real information, and none of them is about your idea.
Build
Zero to One
Research
The problem with asking people about your idea is that they are answering a different question than the one you asked. You want to know whether they will change their behaviour. They hear a request for support from someone who is clearly excited, and they give it, because they are not monsters.
So the encouraging conversations pile up, the idea survives, and the first real test arrives eighteen months later when a stranger has to decide with their own money and no relationship to you.
Everything below is designed to get that test earlier and cheaper. The stakes are set by the base rate: 4.6% of newly launched subscription apps reach $10,000 in monthly recurring revenue within two years.
Key takeaways
Ask about what people already do, not about what they would do. Past behaviour is evidence, intention is politeness.
The strongest signal is discovering they have already built a bad workaround. That means the problem is real and already has a budget.
If nobody currently spends money, time or effort on this problem, you are not competing with a rival, you are competing with indifference, which is harder.
The output of validation is not a yes or a no. It is a sharper definition of who you are for.
Why does asking "would you use this?" fail?
Because the honest answer to a hypothetical is always yes, and it commits nobody to anything.
There is no cost to agreeing. There is a social cost to telling a visibly invested founder that their idea is unremarkable. So the data you collect is a measurement of politeness with your idea's name attached.
Worse, it is directionally consistent. Everyone is encouraging, so you conclude the market is broadly positive, when what you have measured is that people are broadly kind.
Replace the hypothetical with history. Not would you use a tool that does X, but what did you do the last time X happened? The second question has a factual answer, and the answer is either a story with specifics in it or an admission that X does not really happen.
What are the four questions that actually work?
They are all about the past, and none of them mentions your product.
1. When did this last happen to you? If they cannot place it, the problem is not frequent enough to build a business on. Vague frequency is a red flag; people remember real pain with dates attached. 2. What did you do about it? This is the important one. You are looking for a workaround. A spreadsheet, a manual process, a freelancer, a Zapier chain held together with hope. 3. What did that cost you? In money, hours, or errors. This is your value ceiling. A problem that costs someone twenty minutes a quarter cannot support a subscription however elegant your fix is. 4. Why didn't you solve it properly? The answer tells you what you are actually competing with, which is often not a product. It is budget, priority, or the fact that the workaround is good enough.
Notice that a person can answer all four without ever hearing what you are building. That is the point. The moment you describe your solution, you have converted a witness into a supporter.
FREE GUIDE
The 4 product leaks costing you growth
A short audit guide for founders. Find the four places your product leaks revenue, and what to fix first.
What does a strong signal look like?
Someone has already built a bad version of your product for themselves.
This is the finding worth reorganising a plan around. A hacked-together spreadsheet, a manual weekly routine, a paid freelancer doing something a script should do. It means the problem was painful enough to spend real effort on, without anyone selling them anything.
What you hear | What it means |
|---|---|
"We built a spreadsheet for that" | Strong. Real problem, existing effort, quantifiable cost |
"We pay someone to do it manually" | Strongest. There is already a budget line to redirect |
"We tried a tool but stopped" | Good. Find out precisely why they stopped |
"That would be useful" | Nothing. Politeness with no behaviour behind it |
"We just live with it" | Weak. Real problem, no urgency, hard first sale |
The bottom two rows are where most validation ends up, and they are frequently misread as encouragement. "That would be useful" is the sound of someone helping you feel good.
Can an AI validator do this for you?
No, because it has no access to the only evidence that counts: what people already did.
Two of the tools ranking for this question will score your idea in about two minutes. What they are scoring is your description of the idea, which is your own reasoning handed back to you with a number attached. That is a market-sized guess, not validation.
The distinction is simple. A model can tell you the category looks crowded or the wording is vague, both of which are useful before you write an outreach email. It cannot tell you that the last time this problem hit, a real operations lead built a spreadsheet and has been maintaining it every Friday since. That fact is the validation, and it only exists in a conversation.
Use the tools to sharpen the questions. Then go and ask ten people.
How many conversations is enough?
Enough that you stop hearing new things, which is usually fewer than you fear and more than you have done.
The signal is repetition. Y Combinator's own account of finding product-market fit describes the same thing arriving as a pattern rather than a verdict (YC, How to Find Product-Market Fit). When the fourth person in a row describes the same workaround in the same words, you have found a pattern rather than a person. That is typically somewhere between eight and fifteen conversations inside one narrow segment, and it collapses quickly if you keep changing who you talk to.
That last point is where most founders lose the thread. Ten conversations across ten different kinds of company produces ten unique answers and no pattern. Ten conversations with the same role at the same stage in the same vertical produces a definition.
Narrow deliberately. You are not limiting your market, you are choosing where to be undeniably right first. That narrow definition is also what makes an activation event definable later, because you will know whose behaviour you are measuring.
What is the output supposed to be?
A sharper sentence about who you are for, not a verdict on your idea.
Validation rarely returns a clean yes or no. It returns something more useful: the discovery that your idea is compelling to a narrower group than you assumed, for a slightly different reason than you assumed.
The sentence to walk out with: for [specific person] with [specific recurring problem], we [do the thing] so they get [outcome] instead of [current workaround]. Every clause should be traceable to something a real person told you, and the last one should name the workaround you actually heard.
That sentence then decides what you build first, which we have written about separately in what to build first. It also decides what you do not build, which is the part that saves the runway.
It is worth being honest that this only narrows the risk rather than removing it. Dalton Caldwell's account of Socialcam is the standing reminder that enormous demand signals can sit on top of no retention at all (YC, The Real Product-Market Fit), and no amount of pre-build research substitutes for a cohort that comes back.
And if you cannot fill in the last clause, that is the finding. A problem with no current workaround is usually a problem nobody is trying to solve, and the reason is rarely that everyone else missed it.
WHAT NEXT