Most startup validation advice is useless.
Not because the advice is wrong in principle — "talk to customers" is correct. The problem is that the advice tells you what to do without telling you why most people who follow it still end up building the wrong thing.
This guide is not about the mechanics of validation. It's about the logic of it. Why validation works when it works, and why it fails when founders go through the motions without engaging with the substance.
What Validation Is Actually For
Founders often think of validation as a box to tick before building. Talk to 20 people. Get positive responses. Proceed with confidence.
This misunderstands what validation is for.
Validation is not about proving that your idea is good. It is about identifying which of your assumptions are wrong before you've committed capital, time, and reputation to them.
Every startup idea is a bundle of assumptions. Some are assumptions about the problem: who has it, how severely they experience it, how they currently address it. Some are assumptions about the solution: whether your approach actually works, whether people will pay for it, whether they'll keep paying for it. Some are assumptions about the business: what it costs to acquire customers, what margins look like, how the company grows.
Validation done correctly forces you to surface these assumptions explicitly and test the ones that are most likely to be wrong and most dangerous if they are.
The Validation Trap
The reason most startup validation fails to prevent bad decisions is the validation trap.
The validation trap occurs when founders seek information that confirms their hypothesis rather than information that would disconfirm it.
It looks like this:
- Pitching the idea to people who are likely to be supportive (friends, family, fellow founders)
- Asking leading questions that prime positive responses ("Wouldn't it be useful if you could…?")
- Interpreting "that's interesting" as "I would pay for that"
- Counting expressions of enthusiasm without testing willingness to act
- Accepting "we'd definitely consider using that" as validation of demand
The result is a founder who has done twenty customer conversations and collected twenty positive responses — and is no closer to knowing whether they're building something people will actually pay for.
The Assumption Map
Before any customer conversation, write down every significant assumption your business requires to be true.
Then rank them on two dimensions:
Impact: How wrong is the business if this assumption is false?
Confidence: How certain are you, right now, that this assumption is correct?
High-impact, low-confidence assumptions are your validation priorities. These are the beliefs that your business is most dependent on and least certain about. If they're wrong, the business either doesn't work or requires fundamental redesign. They deserve the most rigorous testing before you commit to building.
High-impact, high-confidence assumptions still need testing — confidence is not the same as correctness. But you can invest less validation effort here initially.
Low-impact assumptions are lower priority. Being wrong about them is survivable.
This exercise takes less than an hour and immediately reveals what you actually need to find out — rather than spending your validation time collecting enthusiasm from people unlikely to use your product.
What Good Customer Conversations Look Like
The goal of a customer conversation is to understand reality, not to sell your idea.
This requires a specific posture: curiosity without agenda. You are trying to understand how the world works for the person in front of you, not convince them that your solution is right.
Questions that work:
- "Tell me about the last time you experienced [the problem]. What happened?"
- "What do you currently do about [the problem]? How much does that cost?"
- "What have you tried before? What didn't work about it?"
- "How much time does [the problem] cost you in a week?"
- "Who in your organisation cares most about this? Who makes the decision to address it?"
Questions that don't work:
- "Would you use a product that [describes your solution]?"
- "Do you think [the problem] is a significant issue for you?"
- "How much would you pay for something that solved this?"
The distinction is between understanding current behaviour and projecting hypothetical future behaviour. People are unreliable predictors of what they will do. They are accurate reporters of what they have done.
If someone tells you they currently pay $2,000/month for an inadequate solution to the problem you're solving, that is real information. If someone tells you they would pay $100/month for your solution, that is a hypothesis that requires actual testing.
The Only Validation That Counts
Verbal validation is necessary but not sufficient.
The only reliable test of whether someone will pay for something is asking them to pay for it.
This can take different forms depending on what you're building:
- Pre-orders or letters of intent with a specific price: Not warm praise, but a written commitment to purchase at a stated price
- A paid pilot: Even a small payment creates accountability. People who pay attention differently
- A waiting list with a payment deposit: Tests whether the commitment survives moving from conversation to transaction
- A manual service delivered before the software exists: Validates willingness to pay for the outcome without waiting for the technology to exist
The common element is that the customer has to do something uncomfortable — commit time, sign something, or move money. If they're not willing to do that, the validation has not occurred.
This is not cynicism. It is an accurate understanding of how human behaviour works. The gap between enthusiasm and action is where most startup ideas fail.
What Invalidation Looks Like
Knowing what bad validation looks like is as important as knowing what good validation looks like.
Your idea is not (yet) validated if:
- Everyone says it's a great idea but nobody will pre-order, pilot, or sign a letter of intent
- The people who are most enthusiastic are not the people who would actually pay
- You can only find customers by pitching to your existing network rather than through channels that will eventually scale
- Customers are interested in the problem but want to solve it in a fundamentally different way to your proposed solution
- The pain is real but intermittent — people experience it occasionally but not frequently enough to justify switching costs
- The market exists but is currently addressed by a solution customers are satisfied with
None of these invalidate the idea permanently. They identify the specific assumptions that need reworking before the idea becomes viable.
Common Validation Mistakes
Validating the problem, not the solution
The problem may be real and the solution may still be wrong. Customers who agree that something is broken do not necessarily agree with your proposed fix. Test both.
Sampling from the wrong population
Founders naturally talk to people they have access to — which often means people in their own industry, network, or demographic. If your customer is a 55-year-old franchise operator, talking to 25-year-old tech founders will not validate your assumptions about that customer's behaviour or willingness to pay.
Confusing market research with validation
Reading industry reports, studying competitors, and calculating total addressable market is useful context. It is not validation. Validation is about your specific business model and your specific solution with your specific customers.
Moving the goalpost
After a round of conversations that don't produce strong signal, it's tempting to revise the customer definition to match the people who did respond positively. This can be legitimate — customer discovery sometimes reveals a better initial segment. But it can also be a mechanism for avoiding the conclusion that the idea doesn't work. Be honest about which is happening.
What Validated Looks Like
A startup idea is genuinely validated when:
- You have identified a specific, reachable population of people who experience the problem
- You have confirmed that they currently address the problem in a way they are not satisfied with
- Some of them have committed — with money, time, or a written agreement — to using your solution
- The economics of acquiring those customers and delivering your solution are workable
- You have a credible mechanism for reaching more of them
This is a high bar. Most startup ideas, honestly assessed, do not clear it. That is the point. Clearing a lower bar is how founders spend a year building something before discovering the bar wasn't high enough.
Frequently Asked Questions
How many customer conversations do I need before I've validated my idea?
There's no magic number, but the goal is to reach the point where conversations stop producing new information — where patterns repeat and surprises stop. For most B2B ideas, this is somewhere between 20 and 40 conversations. Consumer ideas often require larger sample sizes due to higher behavioural variability.
What if my idea is so new that customers can't articulate the problem?
If customers can't articulate the problem, they probably don't feel it acutely enough to pay to solve it. The "visionary product" framing — where you create a need customers didn't know they had — is real but rare, and it requires a very different go-to-market approach. Most founders who believe they're in this category are actually in the category of "I haven't found the right customers yet."
Should I validate before building an MVP?
Yes, to the extent possible. The MVP is itself a validation tool, but it's an expensive one. The more validation you can do with conversations, mockups, and manual processes before writing code, the cheaper it is to discover you need to change direction.
What's the difference between validation and product-market fit?
Validation is a pre-build assessment. Product-market fit is a post-launch measurement. You validate to decide whether to build. You measure product-market fit to determine whether what you built is working. Many companies discover product-market fit problems that good pre-build validation would have surfaced earlier.
My early customers love the product — does that mean I'm validated?
Early customers are a necessary but not sufficient signal. You need to understand why they love it — whether those reasons generalise to a broader market, whether they're paying you at a sustainable price, and whether they're staying. Love from a small, unrepresentative sample is fragile validation.