Validation guides
Business Idea Validation: An Evidence Checklist
Validation is not a feeling that the idea is good. It is five pieces of evidence you can point to. Here is the checklist, what counts as proof for each item, and what only looks like proof.
Business idea validation means collecting evidence that a specific buyer has a problem, already spends time or money on it, can be reached, and will commit something before you build. Compliments, survey likes and waitlist sign-ups are not evidence on their own. A named buyer who pays, pre-orders or signs a letter of intent is.
The stakes are real. CB Insights studied 431 venture-backed companies that shut down from 2023 onward and found that 43% failed on poor product-market fit, the second most common reason after running out of capital (CB Insights, 2026). Running out of money is usually how a company dies. Building something the market did not need is often why.
Key takeaways
- Validate the buyer before the product. The first question is who pays and why now, not whether your features are good.
- Opinions are not evidence. Past behaviour and commitments (time, reputation, money) are.
- Use a five-item checklist: problem, current spend, reachable buyers, a commitment, and a price that covers the cost of selling.
- Write your kill rule before you start, so a weak result ends the idea instead of being explained away.
- A free first read takes minutes. Run the idea through Quick Validate before you spend on interviews or code.
What is business idea validation?
It is the work of turning the assumptions an idea depends on into evidence, before the expensive part starts. The expensive part is usually building, hiring or quitting a job.
Every idea rests on a short list of beliefs: that a particular person has the problem, that it costs them enough to act, that you can reach them, and that they will pay a price that works. Validation tests those beliefs in order of how badly the idea breaks if they are wrong.
Steve Blank put the rule behind this most plainly in his Customer Development Manifesto: "There are no facts inside your building, so get outside" (Steve Blank). A spreadsheet, a pitch deck or a model's opinion can organise your assumptions. None of them can confirm one.
Validation is also not the same as research. Research tells you a market exists. Validation tells you whether your offer, to this buyer, at this price, gets a yes.
What evidence should you collect before you build?
Five things, in this order: a real problem, current spending on it, buyers you can reach, a commitment, and a price that covers the cost of selling. If any one is missing, the idea is not validated yet.
| Checklist item | What counts as evidence | What does not count |
|---|---|---|
| 1. A specific problem | Five or more people in the same role describe the same problem, unprompted, from recent experience | People agreeing it "sounds like a problem" when you describe it |
| 2. Current spend | They already pay for a workaround, a tool, a contractor or staff time, and can say roughly how much | "I would pay for that" |
| 3. Reachable buyers | You can name where 50 more of them gather, and you got replies from a cold message | A large total market figure with no route to the first buyer |
| 4. A commitment | A pre-order, deposit, paid pilot, letter of intent, or an introduction to the budget owner | Likes, follows, survey responses, waitlist sign-ups on their own |
| 5. A workable price | The price they accepted covers what it costs you to find, sell to and serve them | A price chosen by copying a competitor's page |
Work the list top to bottom. There is no point testing a price with someone who does not have the problem.
Two rules make the checklist honest. First, each item needs evidence from the same buyer. Enthusiasm from students and budget from procurement teams do not add up to one validated customer. Second, write down before you start what result would stop you, for example: "If fewer than 3 of 10 target buyers describe this problem unprompted, I drop or reshape the idea."
Why don't compliments and surveys count as validation?
Because they cost the person nothing. People are kind about ideas, especially to your face, and predictions about their own future behaviour are unreliable.
Rob Fitzpatrick's The Mom Test is built on this point, subtitled "how to talk to customers and learn if your business is a good idea when everyone is lying to you" (The Mom Test). As summarised by Sachin Rekhi, its three rules are:
- Talk about their life instead of your idea.
- Ask about specifics in the past instead of opinions about the future.
- Validate by asking for commitments of time, reputation or money.
A commitment is useful because the person can say no. Fitzpatrick's line, as Rekhi quotes it: "it's not a real sales lead until you've given them such a concrete chance to reject you."
This is why a survey where 80% say they "would use" something tells you very little, and why a waitlist is a starting point rather than a result. Neither asks anyone to give anything up.
How do you test demand before building anything?
Put a real offer in front of real buyers and measure what they do, not what they say. A landing page with a price, a concierge version delivered by hand, or a paid pilot all work.
Buffer is the classic example. Joel Gascoigne put up a two-page site describing a product that did not exist yet, then added a pricing page between the two to see which plan visitors clicked. In his words, "The extra step tests the pricing (by detecting which plan they click on) and also tests further the demand for the product." He had his first paying customer within 4 days of launching a product he described as "rough around the edges" (Buffer).
Three tests, from cheapest to strongest:
- An offer page with a price. Measures interest at a price, not just interest. Follow up personally with everyone who clicks.
- A concierge version. Deliver the outcome by hand, with spreadsheets, calls and your own time, to three to five buyers. You learn what they actually need, and whether they come back.
- A paid pilot or pre-order. The strongest evidence short of a repeat sale. If nobody will pay for a small, manual version, a polished one rarely changes that.
Paul Graham's advice to founders is the same idea from the other side: the most common unscalable thing to do at the start is to recruit users manually (Paul Graham, "Do Things That Don't Scale").
How do you decide whether to build, reshape or drop the idea?
Compare the evidence to the kill rule you wrote before you started. Build when all five checklist items have evidence from the same buyer. Reshape when the problem is real but the buyer, channel or price is wrong. Drop it when the problem itself does not hold up.
| What you found | What it usually means | Next step |
|---|---|---|
| Problem confirmed, commitment received, price works | Validated for this buyer | Build the smallest version that delivers the outcome |
| Problem confirmed, nobody will commit | Wrong buyer, wrong moment, or the pain is too small | Find who owns the budget, or test a sharper offer |
| Problem confirmed, price does not cover selling cost | The business model does not work yet | Change the buyer, the channel or the delivery model |
| Problem not described unprompted | The problem is yours, not theirs | Drop or reshape before spending more |
Our guide to kill criteria for a business idea covers the deal-breakers that should end an idea even when the rest of the evidence looks good. If you are building software, how to validate a SaaS idea walks through the same process for a subscription product.
Where does NELL fit in validating a business idea?
NELL does the desk research and scoring, so your interviews and tests start from the riskiest assumption rather than a blank page. It does not replace talking to buyers.
Quick Validate is free and unlimited. It gives a viability score, a go, refine or pivot verdict, and next moves, without live web research. Use it to decide whether an idea is worth a week of interviews.
DeepValidate adds live web research: real competitors, sourced market sizing, demand signals, and a score across 11 dimensions in 5 pillars, from "Is the problem real?" to "Can we monetize?". Every claim links to its source, so you can check it. You can see what a full report contains in the sample report.
Neither one can give you a commitment from a buyer. That part of the checklist is still yours to do.
Frequently asked questions
What is the fastest way to validate a business idea?
Ask ten people who fit your target buyer about the last time they had the problem, what they did about it and what it cost them. Then offer the most interested ones a small paid version. Neither step needs a product.
How many customer interviews do you need to validate an idea?
There is no fixed number. Keep going until new conversations with people in the same role stop surfacing new problems. If every conversation still describes a different problem, your target buyer is probably too broad.
Is a waitlist enough to validate a business idea?
No. A waitlist shows interest, but joining one costs nothing. Treat it as a list of people to talk to, then ask them for a commitment such as a pre-order, a deposit or a paid pilot.
Can AI validate a business idea for you?
AI can do the desk research, find competitors, estimate a market and point out the riskiest assumptions. It cannot tell you whether a specific buyer will pay you. Use it to decide what to test, then test it with real buyers.
What if my idea already has competitors?
Competitors are often evidence that people pay to solve the problem. The question is whether you can reach a buyer they serve badly, at a price that works, and why that buyer would switch.
Start where you are
A checklist tells you what to look for. A first read on your own idea tells you where to start.
Sources
- CB Insights, Why Startups Fail: Top 9 Reasons (March 2026)
- Steve Blank, Nail the Customer Development Manifesto to the Wall
- Rob Fitzpatrick, The Mom Test
- Sachin Rekhi, A Primer on Talking to Customers From The Mom Test
- Buffer, Idea to Paying Customers in 7 Weeks: How We Did It
- Paul Graham, Do Things That Don't Scale
This guide describes a general validation process, not a guarantee of results. Interview counts and thresholds are starting points; adjust them to your market.