How to validate a business idea in 24 hours (the AI method)

What "validation" actually means

Most founders use the word validation to mean "a few people said they liked my idea." That is not validation. That is politeness. Validation is the moment a stranger does something costly because of your idea. They book a call. They pay a deposit. They drop their work email into a fake-door page after reading three sentences. Costly behaviour is the signal. Verbal enthusiasm is noise.

Rob Fitzpatrick wrote The Mom Test about this exact gap. The whole book is one argument: people lie to you about future intent because they are being kind. They tell the truth about past behaviour because it already happened. So the question is never "would you use a tool that did X?" The question is "walk me through the last time you tried to do X. What did you use? What broke? What did you spend?" Past behaviour is the cheapest, fastest signal you can get.

For a deeper take on where ideas come from in the first place, the startup ideas pillar walks through the five frameworks that actually produce ideas worth validating. This post assumes you already have one. If you do not, start there, then come back.

The 24-hour validation framework

The framework has four blocks. Each block has a single goal. If a block fails, you stop and either reshape the idea or kill it. You do not push through. Pushing through is how solo founders end up four months deep on something nobody wanted.

1. Hour 1-4Pain test. Find the pain in writing, in the wild, before you talk to anyone.
2. Hour 5-12Buyer-urgency test. Five Mom-Test calls with real prospects. Past behaviour, not future intent.
3. Hour 13-20Smallest signal. A fake-door page or a smoke test. Cost: about $20 in ads.
4. Hour 21-24Score across 5 dimensions. Buyer Urgency, Market Proof, Solo Executability, Revenue Speed, Defensibility. Decide.

None of this is original. Eric Ries called it Lean Validation. YC partners call it "build something people want, then check." What is new is the speed. With AI research agents, the synthesis work that used to take a week (reading 200 forum threads, mapping competitors, drafting a positioning sketch) now takes about an hour. That compression is what makes 24 hours realistic.

Hour 1-4: The pain test

Before you talk to a single person, find the pain written down by strangers. If nobody has complained about this problem in public, you are probably inventing it. Real pain shows up as Reddit threads, Indie Hackers comments, X replies, support forum posts, and one-star reviews of the closest existing tool.

Spend the first two hours searching. Use the buyer's own language, not yours. If your idea is "a tool that helps freelance designers track invoices," do not search "invoice tracking." Search what they actually type: "chasing late client payments," "client ghosted me after invoice," "designer cash flow nightmare." Save 15 to 20 quotes in a doc. You will reuse the exact phrases on your fake-door page in hour 13.

Spend hours 3 and 4 mapping who already gets paid to solve a worse version of this problem. Existing competitors are good news. They prove a budget exists. The Reddit research method walks through this in detail. The output of hour 4 is a one-page brief: the buyer, the pain in their words, the budget signal, and three competing tools they already pay for or hack around.

If you cannot find any of this in four hours of searching, that is your signal. Real pain leaves a trail. No trail means no buyer.

Hour 5-12: The buyer-urgency test

Now five conversations, twenty minutes each, spread across eight hours. You can run all five on a Friday afternoon if you book ahead. DM five people from the threads you saved in hour 1. Use this exact opening: "Saw your post about X. Not pitching anything. Trying to understand the problem better. Got 15 minutes this week?" Most will say yes. People love being asked about their problems by someone who actually read what they wrote.

On every call, ask three Mom Test questions, in this order. One: "Walk me through the last time you ran into this. What were you trying to do?" Two: "What did you use to try to solve it? What did that cost you in time or money?" Three: "Have you ever paid for anything to fix this? What was it? Why did you stop?" Listen. Do not pitch. Do not say "what if there was a tool that." The moment you describe your idea, the conversation contaminates and the data is dead.

After five calls, score each prospect on urgency. Did they describe the pain happening this week, this month, or vaguely "sometimes"? This-week pain is gold. This-month pain is workable. Sometimes-pain means low Buyer Urgency, which is the first of the five scoring dimensions and the one that sinks the most projects. If three of five say this-week or this-month, keep going. If four of five say sometimes, the niche is too broad or the pain is too small.

Hour 13-20: Build the smallest signal

Now you have a hypothesis with verbal evidence. Time to make a stranger pay a tiny cost. Two options work in eight hours. A fake-door test (a landing page that takes emails or deposits for a product that does not exist yet) or a smoke test (a manual offer where you do the work by hand the first time and charge for it).

For the fake-door route, build a one-page site. Headline using the exact phrase from your hour-1 quote doc. Three bullets. One CTA. A waitlist form or a $19 deposit Stripe link. Run $20 of ads to a tight audience. If 100 visitors give you 5 emails or one deposit, you have signal. If 100 visitors give you nothing, the headline or the offer is wrong. Iterate the headline once, run another $20, and decide.

For the smoke-test route, message ten people from your hour-5 calls and offer to do the work manually for a flat fee. "I will do this for you, by hand, this weekend, for $99. Want in?" If two say yes and pay, you have validated faster than any landing page. This is the move Pieter Levels uses on almost every project. The Pieter Levels framework covers the playbook in full.

Either route, the goal is the same. Get a stranger to do something costly. Email, deposit, payment. Pick whichever is fastest for your idea and ship it before hour 20.

Hour 21-24: Score it across 5 dimensions

By hour 21 you have a research brief, five call notes, and a smoke-test or fake-door result. Now write the verdict. Score the idea 0 to 10 on each of these five commercial dimensions. These are the same dimensions IdeaTwister uses on every variation it generates.

• Buyer Urgency. Did three of five prospects describe this-week or this-month pain? 8+. Vague pain? 4 or below.
• Market Proof. Are people already paying competitors for a worse version? 8+. No alternatives in the market? Lower, because no budget exists yet.
• Solo Executability. Can one person ship the first useful version in two to four weeks? 8+. Needs a team or compliance work? 4 or below.
• Revenue Speed. Can you charge in week one without a 90-day enterprise sales cycle? 8+. Long sales cycle? 4 or below.
• Defensibility. Will you have a real moat in 12 months (network, data, brand, distribution)? 6+. None? Doesn't kill the idea, but plan for thin margins.

Composite score above 35 out of 50 with no single dimension below 5? Build it. Score below 30 or any single dimension below 4? Walk away or reshape. The full scoring rubric shows exactly how each dimension is graded inside the engine.

If you want the scoring done for you against 50 mutated versions of your seed, that is exactly what IdeaTwister does in about an hour. 15 specialised agents apply 15+ mutation lenses, score every variation across the five dimensions above, and return a ranked HTML report. Top 5 picks include 30-day validation plans you can run instead of inventing your own.

Bad validation vs good validation

Most founders think they validated when they actually surveyed their friends. Here is the difference, side by side.

Common pitfalls (and how to dodge them)

Pitfall one: pitching during validation calls.The moment you describe your idea, the call becomes a sales conversation. People give you the answer they think you want. Solve this by promising upfront: "I am not pitching anything, I am trying to understand the problem." Then keep that promise.

Pitfall two: validating with the wrong buyer. If you are building a tool for ops managers and your five calls are with founders, you have not validated anything for ops managers. Pick the exact title and stage of the buyer in hour 1 and only talk to that profile in hour 5.

Pitfall three: confusing interest with urgency."Yes that sounds useful" is interest. "I had this problem yesterday and spent an hour on it" is urgency. Only urgency converts to revenue. The Buyer Urgency dimension exists because most ideas die on this exact failure.

Pitfall four: skipping the smoke test. Calls alone are not enough. People say generous things on calls. Until a stranger pays you something, you have hypothesis, not proof. Always run the fake-door or the manual offer. Always.

Pitfall five: validating one version forever. Most ideas need a mutation, not a kill. The right pricing, the right niche, the right channel. The SaaS-idea guide covers the most common mutation patterns. Better still, hand your seed to IdeaTwister and let 15+ mutation lenses propose the variations for you instead of guessing.

When to walk away

Walking away is the hardest part. You spent 24 hours on this. Sunk-cost is loud. Walk away anyway when any of these are true.

• You cannot find the pain written down anywhere by strangers, even with the buyer's own search terms.
• Four of five prospects describe the pain as "sometimes" or "not really."
• A $40 ad spend produced zero emails on the fake-door page after two iterations of the headline.
• Nobody has ever paid anyone for a worse version of this. No competitors means no budget.
• The composite score is below 30 or any single dimension is below 4.

Walking away is not failure. It is the entire point of validation. The whole reason you ran a 24-hour test was so you could kill bad ideas in a day instead of a quarter. A walk-away is a successful test. You saved yourself three months and a few thousand dollars.

What you do not do is "just build it anyway." That instinct is what kept your last project on life support for too long. If the data says no, the data says no. Pick the next seed and start again. For a steady pipeline of seeds worth testing, look at the YC Requests for Startups list and the Paul Graham frontier method for finding under-served frontiers.

Frequently asked questions