How to Validate a Startup Idea: A Practical, Evidence-First Guide

To validate a startup idea, find out whether a specific group of people has a painful problem, is already trying to fix it, and would pay you. Do it in this order: write the problem down, talk to 10 to 15 people who have it, ask what they have done about it, check who else is selling a fix, and then ask for money or a firm commitment before you build. Treat each step as a way to kill the idea cheaply. If it survives, you have evidence. If it doesn't, you saved months.

Why most validation fails

Most founders don't skip validation. They do a version of it that can't return a "no." They describe the idea to friends, get polite nods, and count the nods as demand. Y Combinator's talk on talking to users, as summarized by eChai, makes the same point: the interview is there to extract data, not to sell, and hypothetical questions like "would you use this?" should give way to what people have actually done about the problem.

The fix is to decide in advance what would change your mind. That is also how we built our scoring methodology: five criteria, weighted, then discounted when the evidence is thin.

The six steps

1. State the problem, the buyer and the current workaround

One sentence each. "Small home-care agencies lose hours every week arranging shift swaps by text" is testable. "Scheduling is broken" is not. If you can't name who writes the check, stop here and narrow it. Ideas like shift swaps for home care or a quote builder for pressure washing are written in that shape: one buyer, one job.

2. Interview 10 to 15 people who actually have the problem

Skip friends and skip people who "might be interested." Find people in the role. Ask about the last time the problem came up, what it cost them, and what they tried. Don't pitch. If they have never looked for a fix, that tells you something about how much it hurts.

3. Look for proof of pain, not interest

Pain shows up as behavior: a spreadsheet someone maintains by hand, a part-time hire made to handle it, a tool they pay for and complain about, a cost they can put a number on. Opinions are cheap. We weight pain highest in our rubric (30%) for this reason.

4. Check why now

Something should have changed recently that makes this idea possible or newly urgent: a regulation, a platform shift, a cost drop, a behavior change. YC's Jared Friedman lists this as one of the key questions in his idea evaluation framework (the link is a reader's notes on the talk). If nothing changed, ask why nobody has already built it. Sometimes there's a good answer. Sometimes the answer is that it's a tarpit.

5. Map the competition and the money

Search for who sells to your buyer today, what they charge, and what reviewers complain about. Competition is mostly good news, since it shows people pay. Be honest about where incumbents win: distribution, trust, integrations, years of data. Then work out roughly how many buyers exist and whether the price you have in mind covers a real business. Our daily ideas list real comparables for this reason; an idea with no comparable usually means no one has shown the money exists.

6. Ask for a commitment

This is the step that separates validation from research. Offer a paid pilot, a deposit, a signed letter of intent, or a preorder at a real price. A smaller ask also counts if it costs the person something: an intro to their boss, access to their data, an hour of their team's time. "Sounds great, send me a link" counts for nothing.

Scoring what you find

Once you have notes, put them on one scale so you aren't swayed by the last call you had. Y Combinator's partner Jared Friedman suggests scoring an idea on four criteria and taking the average, according to the YC Startup Library transcript. We use a different set, with unequal weights:

The weighted total is then multiplied by an evidence factor between 0.70 and 1.00. An idea backed only by a founder's hunch can lose up to 30% of its score. One with interviews, comparables and a paid commitment keeps all of it. Pain counts three times as much as ease, on purpose: easy-to-build ideas for people who don't care are the most common way to waste a year. Full details are on the methodology page.

You can see the rubric applied to one idea a day on today's idea, and compare scored ideas on the leaderboard. The Lab is the companion to this guide if you want to work through your own idea against the same criteria.

Common mistakes

• Asking about the future. "Would you pay for this?" gets a kind answer. "What did you do the last time this happened?" gets a true one.
• Interviewing the wrong people. A fleet manager and a driver have different problems. Talk to whoever owns the budget and whoever feels the pain.
• Treating a waitlist as demand. Email signups cost nothing. Treat them as weak evidence at best.
• Reading competition as a stop sign. Existing sellers usually prove the market. A gap in their coverage of a niche, such as trucking or restaurants, can be the opening.
• Ignoring tarpit ideas. Search whether others tried the same thing and what went wrong before you spend a month on it.
• Letting a score replace talking to people. A rubric organizes evidence; it can't create it. Tools that score ideas, including the ones on our comparison pages, are only as good as the inputs.

How long should validation take?

Two to four weeks of part-time work is enough to get a clear signal for most small-business ideas. The YC advice, as the Startup Library transcript puts it, is to land somewhere between trying the first idea that comes to mind and waiting for a perfect one. Treat your first idea as a starting point, and be ready to change it after the first five calls.