How to Validate a Business Idea Before You Spend a Dollar
A practical, four-step test you can run in a weekend. No spreadsheet, no funding, no building.
To validate a business idea, talk to 10-15 people who actually have the problem you're solving, ask about their current behavior instead of pitching your solution, check whether they're already paying for a workaround, and look for a recurring pattern across those conversations rather than trusting any single enthusiastic response. An idea is validated when strangers describe the problem unprompted and can point to what it currently costs them, not when friends say "I'd use that."
The 4-step validation test
- Talk to 10-15 people who have the problem, not friends or family
- Ask about past behavior ("How do you handle this today?"), not hypotheticals
- Check if they already pay for a workaround, even a bad one
- Look for a pattern across interviews, not one enthusiastic yes
Best for
- First-time founders who haven't talked to a customer yet
- Anyone about to spend money on branding, a website, or an MVP
- Ideas aimed at a specific, name-able type of customer
Not for
- Ideas that already have paying customers. You're past validation; focus on retention
- Pure R&D or deep-tech bets where the market doesn't exist yet to interview
Why "would you use this?" is the wrong question
Almost everyone says yes to a hypothetical. It costs them nothing to be encouraging, and most people genuinely want to be supportive of something you're excited about. That's why "would you use this?" and "does this sound like a good idea?" produce validation theater instead of real signal: you walk away with 10 yeses and zero evidence.
The fix, popularized as "The Mom Test," is to ask about specific past behavior instead of future hypotheticals. Not "would you pay for a tool that tracks X" but "how do you track X right now, and what's annoying about it." Behavior is hard to fake. Opinions are free.
Who to actually talk to
Skip your personal network for the core validation round. Friends and family have a relationship to protect and will soften their feedback whether they mean to or not. You want strangers who match your target customer: same role, same industry, same rough situation.
Ten to fifteen conversations is the number that shows up consistently in Lean Startup literature, and it holds up in practice. Fewer than that and one or two loud opinions can skew your read; more than that and you're usually re-hearing the same three themes without new information.
What a "yes" actually looks like
Real signal has a shape: the person describes the problem before you bring it up, they can tell you what they currently do about it (even if it's a clunky spreadsheet or a workaround), and ideally they can put a rough cost on it in time, money, or frustration. If several strangers independently describe the same pain in similar language, that's a pattern worth building against.
Weak signal also has a shape: polite enthusiasm, no specifics, and no existing workaround because the "problem" isn't actually painful enough for anyone to have solved it badly yet. That's usually a sign to keep interviewing rather than start building.
What to do with what you hear
Write answers down verbatim, not paraphrased. The exact words people use to describe their pain become your marketing copy later. After ten or so interviews, look across them for recurring themes rather than picking your favorite quote. A theme three people raised independently is worth more than the single most excited response you got.
If the interviews contradict each other, say half the room cares about speed and half cares about price, that's not a failure. It's a segmentation signal. It usually means "everyone" isn't your customer; a specific slice of that room is.
Frequently asked questions
How many customer interviews do I need before I start building?
Ten to fifteen is the commonly cited threshold: enough to see a repeated pattern, not so many that you're just re-confirming what you already know. Fewer than five and you're making a bet on a small, possibly unrepresentative sample.
What if everyone I talk to says something different?
Disagreement across interviews is data, not noise. It usually means your target customer is more than one segment, and each segment cares about something different. Narrow to the segment with the strongest, most specific pain rather than trying to solve every version at once.
Is a survey good enough for validation?
Surveys are useful for confirming a pattern you've already found in conversation, but they're a weak first step. People answer surveys quickly and generically, and you lose the follow-up question that usually surfaces the real insight. Start with conversations; use surveys to check scale once you have a hypothesis to test.