How to Validate a Startup Idea Before Building It: 7 Proven Steps

Somewhere in a drawer, on a hard drive, or in a Notion doc you've renamed four times, there's an idea you can't stop thinking about. And you've been sitting on it for months because you know one thing: the graveyard of startups is full of beautifully built products that nobody wanted.

So how do you validate a startup idea before building it? Not the vague "talk to customers" advice everyone parrots. Actual signals. Actual thresholds. Actual reasons to stop.

I'm going to give you the framework I now use after getting this wrong more than once. Not theory. The stuff that cost me time and money to learn.

Why most validation advice fails you on day one

Here's the thing nobody tells you: the problem is rarely the idea. It's that founders validate their own confidence, not the market.

I did this. In my first serious attempt at a product—a scheduling tool for small clinics—I called 22 people I'd met through a former job. I asked each one the same question: "Would you use something like this?" Nineteen said yes. I felt unstoppable.

I built it for five months. Do you know how many of those 22 installed it when I launched? Two. And both dropped off within three weeks.

What went wrong wasn't the research. It was the question. "Would you use this?" is a politeness test, not a validation test. People are wired to be encouraging. Especially when you're passionate and they can hear it in your voice.

The question that ruins everything

Rob Fitzpatrick's The Mom Test nails this, and it's still the single best piece of startup reading I've found: never ask about the future. Ask about the past.

Bad question: "Would your team use this?"

Good question: "The last time your team had this problem, walk me through exactly what happened."

One triggers polite fantasy. The other triggers memory. And memory is the only honest dataset you have access to before building anything.

The validation framework I actually use

After the clinic disaster I rebuilt my process. It has four stages, and each one has a threshold that either lets you continue or tells you to stop. This isn't a fill-in-the-blanks worksheet. It's a filter.

Stage 1: problem interviews (10-15 people, no pitching)

Your goal in these conversations is not to sell. It's to find out if the problem is urgent, frequent, and expensive enough that people already spend money or time fighting it.

What I track in every interview:

• When did this problem last happen to them (last week beats "constantly")
• What did they do about it that day
• What did that solution cost them—in euros, hours, or workarounds
• How annoyed were they on a scale of one to ten

If fewer than six of your fifteen interviews describe a recent, painful, already-paid-for problem, stop. You don't have a startup. You have a nice idea.

Stage 2: artifact tests (a landing page, a fake door, a mockup)

This is where validity gets real, because you're measuring behavior instead of opinion. A landing page costs you a weekend. It's not a substitute for a product—it's a measurement instrument.

Build one page that describes the outcome your idea delivers, not the features. Add a clear call to action: join the waitlist, book a demo, reserve a spot, pre-order. Drive a small amount of paid traffic to it—enough to get 200-300 visitors—and watch what happens.

Roughly: 1-3% converting to an email is weak. 5-8% is worth a second test. Anything above 10% on cold traffic means you've hit something worth pursuing. These numbers aren't laws; they're calibration. When I ran this for a niche B2B tool, I got 2.4% on a broad audience and 11% when I narrowed to one specific job title. Same page. The targeting was the entire difference.

Stage 3: commitment tests (ask for something costly)

Money is best. Time is second-best. A calendar invite is a real signal; a like is not.

Pre-orders are the cleanest test available to a solo founder. If you're building software, a "reserve your spot for €X, refundable any time" button filters out every polite liar in your contact list. If you're building a service or a physical product, a deposit works the same way.

The rule: if no one commits a resource before your product exists, you don't have demand. You have interest. Those are completely different things and it took me an embarrassingly long time to separate them.

Test	What it actually measures	Cost to run	Signal strength
Problem interviews	Whether the pain is real and current	Your time	Low (context only)
Landing page	Whether the message resonates	Weekend + small ad budget	Medium
Fake door / waitlist	Whether the offer attracts clicks	Days	Medium
Pre-order or deposit	Whether people will pay before it exists	Days to weeks	High
Manual concierge run	Whether you can deliver the outcome at all	One weekend	High

Stage 4: the concierge run (deliver the outcome by hand)

Before you automate anything, deliver the result manually for two or three people. Charge them for it. If you can't deliver a valuable outcome even with unlimited human effort, no amount of engineering will fix that.

I've done this twice. Once it worked and I learned which part of the deliverable mattered most (not the part I'd assumed). Once it collapsed in a week, and I killed the idea. That second one was the best week of work I did that year and I'll defend that claim to anyone.

How to validate a business idea when you're not technical

The good news: none of what I just described requires you to code. The landing page can be built with a no-code builder. The pre-order can run through a simple payment link. The concierge run is you, a spreadsheet, and a willingness to do unglamorous work.

What you can't outsource is the conversations. That's the part where founders quit, because hearing someone describe your idea as "not really a priority" hurts. Sit with the hurt. It's cheaper than building.

There's a growing category of tools that promise to validate ideas with AI—survey generators, market analysis engines, synthetic customer panels. Use them to prepare your questions, never to answer them. A model can summarize a market. It cannot tell you whether a specific operations manager at a specific company will pay €80 a month to stop doing a spreadsheet by hand every Friday. Only that manager can.

What kill criteria look like in practice

Before you start validation, write down the numbers that would make you walk away. Do it on paper, date it, and don't edit it later. The reason is simple: once you've spent three months on something, your brain will find a reason every failed test was "actually kind of positive."

A kill criteria sheet I've used looks roughly like this:

1. Fewer than 6 of 15 interviews describe a recent, costly problem → stop
2. Landing page conversion under 2% on 300 targeted visitors → stop or swap the offer
3. Fewer than 3 pre-orders in 10 days of showing the page to 100 relevant people → stop
4. Concierge run can't be delivered in under 5 hours per customer → stop or redesign
5. Nobody refers a colleague after the concierge run → stop

Note the fifth one. It's the one I added after being burned, and it's underrated. Referrals are the only signal that survives contact with the real world.

The trap of "validated startup ideas"

You'll see lists of validated startup ideas circulating online—"these 40 ideas have been validated, grab one." Treat them as inspiration, not evidence. An idea validated for someone else in a different city, with a different network, pitched to a different audience, is not validated for you.

What those lists are actually good for is the meta-signal: if an idea appears on three reputable lists, the problem it addresses is probably durable. The validation still has to be yours. The tests are cheap. Run them.

How many customer interviews are actually enough?

For a niche B2B idea, 10-15 targeted conversations typically surface the pattern. For consumer products, you often need 20-30 because the pain is more diffuse. Stop when you can predict what the next person will say before they say it—that's your signal, not a fixed number.

Can I validate with surveys instead of conversations?

Surveys are good for confirming scale after interviews have revealed the problem. They're terrible for discovery, because you can't follow up on a strange answer, and you can't read hesitation in a checkmark.

What if my idea is genuinely novel and no one does anything like it yet?

Then your validation target shifts. Instead of asking what people already pay for, ask what they currently sacrifice—time, quality, sanity—to work around the gap. Novel problems still have costs. Find them.

What I'd tell you if we were having coffee

The temptation is to treat validation as a hurdle between you and the fun part. It isn't. It's the fun part, in a different disguise—it's where you learn things you'd otherwise have to learn after spending a year and most of your savings.

I've watched founders skip validation because the idea felt obvious. I've watched founders over-validate and never ship anything. Both fail, just at different speeds. The line between them is pre-written kill criteria and a willingness to actually honor them.

The most useful question you can ask yourself right now isn't "is my idea good?" It's "what would I need to see in the next two weeks to change my mind?" If you can't answer that, you're not validating. You're just hoping with extra steps.