Most startup ideas do not fail because the product was bad. They fail because nobody validated that a reachable group of buyers had the problem, cared enough to pay, and could be found. The build was fine; the market was a guess.
Validation is how you stop guessing. Done right, it is not a six-month research project and it is not a poll of your friends. It is a short, structured hunt for evidence, and it ends in one of two good outcomes: you build with conviction, or you kill the idea in two weeks instead of two years. Both are wins.
What validation actually means
Strip the jargon and market validation is five questions:
- Is the pain real? Do people describe this problem in their own words, unprompted, in public?
- Who exactly has it? Not "SMBs". A nameable buyer: role, company size, industry, situation.
- What do they pay today? Money already moving is the strongest demand signal there is.
- Why would they switch? What is broken about the current answer, in their words?
- Can you reach them? A real segment you cannot contact is the same as no segment.
Everything below is just a systematic way to answer those five.
Step 1: Find the pain in the wild
Before you interview anyone, read what buyers already wrote when nobody was selling to them. Review sites (G2, Capterra) are a goldmine: the one-star and three-star reviews of the incumbent tools in your space are literally your target market describing what they wish existed. Reddit and niche communities carry the same signal in rawer form. Support forums, changelogs, and public roadmaps show what incumbents are being pushed to fix.
You are looking for repetition. One person complaining is an anecdote. Thirty people describing the same broken workflow in the same words is a market.
Step 2: Name the buyer, bottom-up
Top-down sizing ("the market is $5B, we need 1%") impresses nobody and informs nothing. Do it bottom-up instead: define the ideal customer profile precisely enough that you could build a list. Role, company size, industry, tooling, trigger event. Then actually build the start of that list. If you cannot find fifty real companies that fit, the segment is not real yet, no matter what the TAM slide says. The full sizing math is in our TAM, SAM, SOM guide.
Step 3: Read the competition as evidence
Founders treat competitors as bad news. Flip it: paid, imperfect competitors are proof the problem is worth money. What you are reading for is the gap. Study their pricing pages (what do buyers pay, and for what tiers), their reviews (what do paying customers still hate), and their positioning (who are they ignoring). An empty market usually means no demand; a crowded market with visible dissatisfaction is an invitation. The full method is in our competitor analysis framework.
Step 4: Test willingness to pay
Interviews come now, after the desk research, so you arrive with a hypothesis instead of a blank page. Ten conversations with people who match the ICP. Do not pitch; ask about the last time they hit the problem, what they did, what it cost them, and what they pay today for anything adjacent. The strongest signals are unprompted: they describe the pain before you name it, they volunteer a number, they ask when they can try it. Polite enthusiasm ("sounds cool, I would try that") is not a signal; it is how conversations end.
Step 5: Score it and decide
Force a verdict. Rate the idea one to ten on each of the five questions, with a sentence of evidence per score. Below thirty total, kill it or reshape it; the evidence told you something. Above it, you now have more than a validated idea: your ICP, your competitor read, and your positioning are the first campaign brief, not just a research artifact. That is the core loop behind a go-to-market engine: the validation work feeds the acquisition work directly.
The manual cost, and the shortcut
Done by hand, this playbook takes about two focused weeks: one for desk research across review sites, traffic tools, and pricing pages, and one for interviews. The desk-research half is exactly what AI market validation tools now compress into hours: demand sizing, competitor teardowns, ICP discovery, and a scored verdict with sources attached. Cafiyn Lens is our version of that, built so the output is decision-grade rather than a wall of unsourced text. The interviews stay yours; no tool should do those for you.
However you run it, run it before you build. Two weeks of validation, or a few hours of it, is the cheapest insurance a founder can buy.