You have a spreadsheet open. Twelve columns, three scenarios, a "conservative" case that already assumes you'll beat your own sales targets by 40%. The number at the bottom is the one you need. It's the number that decides whether you hand in your notice or quietly shelve the whole thing for another year.
Here's the problem with that number: it's fiction. You built it from assumptions you never tested, and the most expensive assumption in the whole model — that people will actually pay for this — is the one you spent the least time checking. I've watched founders spend four months perfecting a financial model and four hours talking to potential buyers. The model looked beautiful. The business didn't exist.
So let's talk about how to validate a business idea before investing — real money, real time, real reputation — without dressing up guesswork as analysis.
Key Takeaways
- Validation is not research. It's evidence that someone will pay, and the only evidence that counts is a transaction or a credible commitment to one.
- Your financial model is downstream of validation, not a substitute for it. Fix the assumptions before you scale the spreadsheet.
- Compliments are not data. "That's a great idea" is a polite exit, not a purchase signal.
- Most ideas die from a lack of urgency, not a lack of interest. People agree the problem exists and still don't buy.
- Set a kill criterion before you start testing, or you'll rationalize any result into a green light.
- Cheap tests that cost a weekend (a landing page, ten conversations, a pre-sale) beat a six-month build you're afraid to abandon.
What "validating your idea" actually means
Ask five founders what validation means and you'll get five answers, most of them wrong. The most common one is "I did market research." That's a different activity with a different goal.
Research tells you the market exists. Validation tells you you can get paid inside it. Those are not the same question, and confusing them is why so many well-researched products launch to silence.
Research versus validation
Research is reading, comparing competitors, and reading the numbers. It's necessary. It's also cheap to fake — you can spend three weeks reading and feel productive while learning almost nothing about whether anyone will buy from you specifically.
Validation is behavioral. Someone does something that costs them: money, a deposit, a calendar slot, a signature, a public commitment in front of their boss. Until that happens, you have a hypothesis with a nice logo.
The three assumptions worth testing first
Every idea rests on a stack of beliefs. Most of them don't matter yet. Three do:
- Does the problem hurt enough to act on? Not "is it a problem" — plenty of things are problems nobody solves.
- Will this specific person pay this specific price? Interest and willingness to pay are different animals.
- Can you reach them at a cost below what they pay you? A validated product with broken acquisition economics is still a dead business.
Test those three before you touch anything else. Everything else is decoration.
Talking to people without getting lied to
Customer interviews are the most misused validation tool there is. Done badly, they manufacture false confidence. Done well, ten conversations can save you six months.
The failure mode is simple. You ask "Would you buy this?" and people say yes. They're not lying to you exactly. They're being polite, and you've made it easy to be polite. Hypothetical questions get hypothetical answers.
Ask about the past, not the future
Stop asking what they would do. Ask what they did. What's the last thing they tried to solve this problem with? How much did it cost? What was annoying about it? Did they switch tools, or give up?
People are unreliable narrators of their own future behavior and reasonably accurate about their past. A person who spent 200 dollars last month on a workaround has already told you the problem is real. A person who can't remember the last time it bothered them has told you something else.
The questions that actually reveal intent
- Walk me through the last time this came up.
- What did you do about it?
- What did that cost you — money, time, or both?
- Who else was involved in the decision?
- What would have to be true for you to change what you're doing today?
Notice none of these ask whether they like your idea. You're not collecting compliments. You're mapping a current behavior and looking for the crack where you could fit.
Real talk: after roughly twenty of these conversations in my own projects, a pattern becomes obvious. Either people are already paying for something to fix this, or they're not. If they're not paying for anything, they usually won't pay for you either — the problem just isn't urgent enough.
Cheap tests that produce real signal
You don't need to build anything to find out whether people want it.
The landing page test
One page. A clear promise, a price, and a button. Drive a small amount of traffic to it — paid ads, a post in a community where your buyers actually gather, an email to a list you already have. What you're measuring is not clicks. Clicks are cheap and meaningless. What you're measuring is how many people who saw the price still tried to give you money.
If the button leads to a checkout and nobody gets past the price, you've learned something valuable for the cost of an afternoon. If you can't even get traffic to the page, that's your first real problem, and you found it before writing a line of code.
Pre-sales and deposits
A deposit is the closest thing to proof you can get without shipping. It's awkward to ask. Do it anyway. "I'm building this, it ships in eight weeks, here's the price, would you put down a deposit to lock in early access?"
The people who say yes are your first customers. The people who say "sounds great, keep me posted" have just told you, politely, that this isn't a priority. Both answers are useful. Only one of them pays rent.
The concierge version
Deliver the outcome manually, by hand, for one or two customers, before you automate anything. This is the test almost nobody runs because it's unglamorous and doesn't scale. It also tells you whether the outcome you're promising is actually achievable and whether people value it enough to pay for the manual version. If they won't pay for it delivered by hand, they won't pay for it delivered by software.
How this differs from just building an MVP and seeing what happens
Building first and validating later feels faster. It isn't. You've inverted the order, and now every piece of feedback arrives after you've already spent the money, which makes it much harder to hear and much harder to act on.
| Approach | What it costs upfront | What you learn | How hard to walk away |
|---|---|---|---|
| Landing page test | A weekend and a small ad budget | Whether the promise and price attract anyone | Easy — nothing to abandon |
| Pre-sale deposits | Two weeks of awkward conversations | Whether people will commit money before the thing exists | Moderate — you can refund and stop |
| Concierge delivery | A few weeks of manual work | Whether the outcome is real and valued | Moderate — no code to throw away |
| Full MVP build | Months of development and design | Whether people use the product you chose to build | Hard — sunk cost fights back |
Notice the pattern in the last column. The more you build before validating, the more your own investment argues against the truth. That's the real cost of skipping validation, and it doesn't show up in the spreadsheet.
What AI can and can't do for your validation
There's a lot of noise right now about using AI to validate an idea. Some of it is genuinely useful. Most of it is a way to feel productive without talking to a single human being, which is the opposite of validation.
AI is good at the research layer. It can map competitors, summarize a pile of forum threads, draft interview questions, stress-test your pricing logic, and help you pressure-test the assumptions in your model. That's real work, and it's faster than doing it by hand.
AI cannot tell you whether your specific buyer will pay your specific price. It has no access to that person's budget, their boss, or their last purchase. If a tool generates a confident-sounding market verdict, that verdict is a synthesis of things other people wrote, not evidence about your situation. Use it to prepare for the conversation with a real customer. Don't let it replace that conversation.
The trap is subtle. An AI that confidently confirms your idea feels like validation. It isn't. It's a mirror with a vocabulary.
Deciding before you find out
Here's the part most people skip, and it's the part that actually protects your money. Before you run a single test, write down what result would make you stop.
Pick a number. "If fewer than two out of twenty people put down a deposit, I don't build this." "If I can't get the cost to reach a customer below a third of what they pay, I don't scale this." Write it down, date it, and put it somewhere you'll see it in three months.
The reason is simple. Once you've spent money and time, you'll want the evidence to say yes, and you'll find a way to read any result as a maybe. A kill criterion you set in advance is the only version of you that can be trusted to judge the outcome fairly — the version who hadn't yet fallen in love with the idea.
And the honest truth about validation is that most of the time it tells you no, or it tells you not yet, or it tells you the problem is real but your customers are different from the ones you assumed. Those are wins. They cost a weekend instead of a year. The validation isn't there to bless your idea. It's there to tell you the truth about it cheaply, while the truth is still cheap to hear.
The number at the bottom of the spreadsheet was never the answer. The answer is whether a stranger, unprompted, handed you money. Go find out which one is true, and do it before the model gets any prettier.
