How to validate a business idea before investing a single dollar
A reader emailed me last month with a familiar story. He'd spent eleven months and about $34,000 building a scheduling tool for freelance physiotherapists. Beautiful product. Clean onboarding. Zero users. When I asked how many therapists had paid him before he wrote the first line of code, the answer was none.
That conversation is the entire reason I'm writing this. Validating a business idea before investing isn't a mindset shift or a philosophy. It's a set of tests, each with a number attached, and each designed to kill your idea as cheaply as possible.
The mistake almost everyone makes is treating validation as encouragement. It isn't. Validation is a filter. You're not looking for reasons your idea will work. You're looking for evidence it won't.
Key Takeaways
- Talk to at least 15 potential buyers before you spend money on anything else. Not friends. Buyers.
- A landing page test needs roughly 200–300 qualified visitors before the conversion rate means anything.
- Pre-sales beat opinions every time. If nobody pays a deposit, the "demand" isn't real yet.
- Set your go/no-go thresholds before you run the test—otherwise you'll rationalize whatever result you get.
- Validation costs money. Budget 5–10% of what you'd spend on the full build, and treat it as a sunk cost.
- Different idea types need different tests. B2B, B2C, and physical products don't validate the same way.
Why most validation advice quietly fails
Here's the thing: everyone tells you to "talk to customers." Almost nobody tells you what to do with the answers.
The failure mode is confirmation bias, and it's brutal because it feels like diligence. You ask a friend who runs a small agency whether they'd use your invoicing tool. They say "oh yeah, definitely, that sounds useful." You write it down as a data point. It isn't. It's politeness, and it cost you nothing to collect, which is exactly why it's worthless.
I made this mistake myself in 2019 with a marketplace for local music teachers. I ran maybe 20 conversations, and I counted every enthusiastic "that's a great idea" as a vote. What I never did was ask anyone to open their wallet. When I finally launched, I got two signups in six weeks. Both were people I already knew.
What a real validation signal looks like
A real signal has three properties:
- It costs the other person something—money, time, a commitment, a public action
- It comes from someone who has the problem right now, not someone who might have it someday
- It's falsifiable. You can tell the difference between "yes" and "no" without squinting
Emails don't qualify. Survey responses don't qualify unless they come with behavior attached. A stranger handing over a credit card does.
The number that matters more than any other
How many interviews? Fifteen.
That's my floor for a B2B idea where each customer is worth thousands of dollars. For consumer ideas where you're targeting broad demand, you want more like 30–40 conversations, because the noise is louder and the signals are weaker. Below that, you're pattern-matching on anecdotes.
The Mom Test, applied properly
Rob Fitzpatrick's book The Mom Test gets cited constantly, and the core rule is simple: ask about their past behavior, not your hypothetical future product.
Bad question: "Would you use a tool that automates your expense reports?"
Good question: "Walk me through how you filed your last expense report. What was the most annoying part?"
The first invites a polite fiction. The second produces a story, and stories contain specifics you can act on. In practice, I've found that about 1 in 4 conversations yields a detail I hadn't anticipated at all—a workaround, a competing tool I'd never heard of, or a constraint that reshapes the whole idea.
Those unanticipated details are the actual product of customer interviews. Not validation. Insight.
Testing demand online before you build anything
Once interviews give you a hypothesis, you need a number that isn't filtered through politeness. That means putting something in front of strangers.
A landing page test with an actual threshold
Build a one-page site describing the product. Include a clear call to action—email signup, waitlist, "request early access." Then drive traffic.
The problem is that most people run this test with 40 visitors and declare victory on a 5% signup rate. That's noise. If you want a conversion rate you can trust, you need at least 200–300 qualified visitors—people who match your target customer, not random traffic from a cheap ad network.
What counts as a pass? For a B2B tool with a high price point, I'd want to see 8–12% of qualified visitors leaving an email. For a consumer product, 3–5% is respectable. Below 2% across a properly sized sample, the demand isn't there yet.
And yes, this costs money. A focused ad test on a narrow audience will run you somewhere between $200 and $600 depending on the platform and the niche. That's the cheapest failure you'll ever buy.
Pre-sales: the test that can't be gamed
Landing page signups still involve a small lie. People give away email addresses the way they give away business cards. A deposit is different.
If your product can be pre-sold, do it. Charge a discounted founding-member price, collect actual payment, and promise delivery in a defined window. I've seen founders refund every pre-sale and still call the test a success—because the willingness to pay was proven, regardless of whether the product shipped.
The threshold I'd use: 10 paying pre-orders for a B2B product with a $500+ price point, or 30–50 for a consumer product under $100. Hit that, and you have permission to build. Miss it badly, and you've saved yourself a year.
| Test method | Cost | Time to result | Signal strength |
|---|---|---|---|
| Customer interviews | Time only | 2–4 weeks | Weak alone, essential for insight |
| Landing page + ads | $200–$600 | 1–2 weeks | Moderate; measures interest |
| Pre-sales / deposits | Payment processing | 2–6 weeks | Strong; measures commitment |
| Concierge MVP (manual service) | Your labor | 4–8 weeks | Strong; tests delivery too |
| Wizard-of-Oz prototype | Low dev cost | 3–6 weeks | Strong; tests workflow fit |
How can I test my business idea?
Run the cheapest test that can produce a definitive no. That's the whole answer.
Start with conversations to find out whether the problem is real. Move to a landing page or ad test to see whether strangers care. Then, if you can, ask for money before you build. Each stage costs more than the last, so you stop as soon as the evidence turns negative. Most ideas die at stage one, and that's a good outcome—it means you found out for the price of a few coffees instead of a few months of your life.
Different ideas need different tests
Generic validation checklists fall apart the moment you apply them across categories. Here's how I'd adapt the approach.
B2B software and services
Your buyers are few, reachable, and expensive to lose. Ten to fifteen interviews with decision-makers will tell you more than a thousand landing page visits. The best test is a concierge MVP: deliver the outcome manually, for real money, to two or three clients. If you can't find three companies willing to pay for a manual version of your service, software won't fix that.
Consumer products
Volume matters more. You need broader ad tests, and you should expect to spend $500–$1,500 before the numbers stabilize. Watch for cost per email signup—if it's above $3–$4 in a category where lifetime value is low, the math won't work even if demand exists.
Physical products
Kickstarter-style crowdfunding is the cleanest test available. It's public, it's financial, and the threshold is written in the platform rules. Run a campaign with an honest minimum. If you can't hit it, the market has spoken. If you hit it easily, you've financed your first production run at the same time.
What to do when validation fails
Most ideas fail. That's normal, and the useful question is what you do next.
There are three moves: pivot, narrow, or walk away. Narrowing is underrated. Often the problem isn't the idea, it's the target customer. I've watched founders go from a general "tool for small businesses" to "tool for dental practices with 3–8 chairs" and see conversion rates jump fourfold—same product, different audience.
Pivoting on a weak signal is the dangerous one. If your landing page got 1% and your interviews were lukewarm, don't pivot the messaging and rerun the same test hoping for a different number. Change something structural: the customer, the problem, or the price.
And here's the part nobody likes: sometimes the correct output of validation is to stop. Not to pivot, not to "iterate"—just stop. That's not failure. That's the system working exactly as designed.
What validation should actually cost you
Budget 5–10% of what you'd spend on the full build, and cap the timeline at six to eight weeks. If you're still "validating" after two months, you're not validating—you're procrastinating with a spreadsheet.
For most software ideas, that means a validation budget somewhere between $500 and $2,000. It feels like money you're burning. But compare it to the alternative: the reader who emailed me spent $34,000 learning the same lesson he could have learned for a few hundred.
The real question isn't whether you can afford to validate. It's whether you can afford not to. And every founder I've met who skipped this step has an answer to that question—usually delivered with a wince.
