Most business ideas die for one reason: the founder built something nobody was waiting for. CB Insights' long-running post-mortem analysis of failed startups puts "no market need" at or near the top of every edition — ahead of running out of cash, ahead of team problems. The fix isn't working harder on the product. It's refusing to build until you have evidence of demand. This guide is a seven-day, zero-budget process for gathering that evidence.
The Rules of the Week
- No building. No code, no product, no logo, no company registration. Only evidence-gathering.
- No spending. Every tool used this week has a free tier. Paid ads come later, if ever.
- Written predictions. Before each test, write down what result would count as a pass. Deciding after you see the data is how founders fool themselves.
- Talk to strangers. Friends and family validate you, not the idea. They don't count.
Day 1: Write the Riskiest Assumption Down
Every idea rests on a stack of assumptions: the problem exists, it's painful enough to pay for, the buyer can be reached affordably, your solution fits how they already work. You don't need to test all of them — you need to find the one that kills the business if it's false, and test that first.
A useful format: "We believe [specific type of person] struggles with [specific problem] badly enough to pay [rough price] for [category of solution]." Every bracket must be concrete. "Small businesses need better marketing" is untestable; "Dutch restaurant owners lose bookings because they can't answer the phone during service, and would pay €99/month to stop losing them" is testable by Friday.
Then write your pass/fail line: for example, "at least 5 of 10 owners rank this in their top three problems, and at least 2 ask about price unprompted."
Day 2: Find 30 People Who Have the Problem
Validation dies without prospects, so day two is pure list-building. You're looking for 30 named people or businesses you can actually reach — enough that ten conversations is realistic after no-shows and non-replies.
- Where they already complain: Reddit communities, Facebook groups, industry forums, review sites of competing products (1-star and 2-star reviews are validation gold).
- Where they're listed: Google Maps for local businesses, LinkedIn for roles, directories and member lists for niches.
- Where they gather: Discord servers, Slack communities, local meetups.
For each prospect note where you found them and one line of evidence that they have the problem. That line becomes your opener — "I saw your post about X" outperforms any template.
Days 3–4: Ten Problem Interviews
The interview's purpose is to learn how the problem behaves in the wild — not to pitch. The moment you pitch, people switch from honesty to politeness. A workable script, adapted from the approach in Rob Fitzpatrick's The Mom Test:
- "Walk me through the last time you dealt with [problem]." (Past behaviour, not future intention.)
- "What have you already tried?" (If they've tried nothing, the pain is probably low.)
- "What does this cost you — in time, money, or stress?" (Quantifies the pain.)
- "If this disappeared tomorrow, what would change for you?" (Reveals whether it matters.)
Signals that count as a pass: they've paid for or hacked together a workaround; they talk for ten minutes without prompting; they ask "when can I try this?" unprompted; they offer to introduce you to someone else with the problem. Signals that don't: compliments, "sounds interesting," and hypothetical enthusiasm.
Day 5: Build a One-Page Offer
Now — and only now — you describe a solution, shaped by the language from your interviews. Use the prospects' own words for the problem; it's the difference between a page that reads like marketing and a page that reads like mind-reading.
A free landing page (Carrd's free tier, Google Sites, or a Notion page) needs only four blocks: the problem in the customer's words, the outcome you deliver, a price or price range, and one call to action. The call to action should cost the visitor something: "book a 15-minute call" or "reserve your spot with a refundable €25 deposit" — not "join the waitlist."
Day 6: Drive 100 Real Visitors
Without spending, you have three channels that can produce a hundred visitors in a day: the communities you mapped on day two (share the page as a genuine ask for feedback where rules allow), your interviewees ("this is what I took from our conversation — did I get it right?"), and direct outreach to the remaining twenty prospects on your list.
A hundred targeted visitors is a small sample, but it's enough to distinguish "nobody clicked" from "one in ten booked a call." You're not measuring statistical significance; you're checking whether the strongest possible audience — people hand-picked for having the problem — takes a costly action.
Day 7: Score It Honestly
Return to the pass/fail lines you wrote on day one and score the week like an examiner, not a fan:
- Strong signal: pre-payments or deposits, multiple unprompted "when can I buy," intros to other buyers. Proceed — and consider pre-selling before you build.
- Mixed signal: real pain confirmed but no willingness to pay your price, or enthusiasm from a different segment than expected. Pivot one variable (segment, price, or promise) and re-run days five to seven next week.
- No signal: couldn't fill ten interviews, or interviews revealed mild annoyance rather than pain. Kill it. A week is a cheap price for that answer — most founders pay a year.
The Concierge Escalation
If day seven gives you a strong signal, the next validation rung is delivering the outcome manually to two or three customers before writing any code or ordering any stock — the "concierge" test. A meal-planning app becomes a weekly WhatsApp message; an analytics platform becomes a hand-built PDF report; a marketplace becomes you matching both sides by email. Manual delivery is slow and unscalable on purpose: it forces contact with every rough edge of the real problem while your only cost is time, and paying concierge customers become your spec sheet, your testimonials, and your first revenue. Once the manual version is repeatedly paid for, the follow-up questions become operational — formalising the business, putting up a proper site, and eventually deciding whether outside funding fits.
Frequently Asked Questions
How many customer interviews are enough to validate an idea?
Ten to fifteen genuine conversations with people who have the problem is usually enough to see a pattern. If eight out of ten describe the problem the same way and three ask when they can buy, you have signal. If you cannot find ten people willing to talk about the problem at all, that is also an answer.
Is a landing page with sign-ups real validation?
Email sign-ups are weak validation because they cost the visitor nothing. Stronger signals, in order: a scheduled follow-up call, a signed letter of intent, a refundable deposit, and an actual pre-payment. Aim for at least one signal that costs the prospect something — time, reputation, or money.
What if my idea requires a product before anyone can judge it?
Almost no idea truly does. You can sell the outcome manually first: consultants sell reports before software, marketplaces begin as spreadsheets and phone calls, and hardware founders pre-sell with renders. If you genuinely cannot deliver any version of the outcome manually, validate the demand for the outcome itself with the concierge test described above.
Getting Started
The whole framework compresses to one sentence: replace opinions with costs. Before you register a company, buy a domain, or write a line of code, spend one week collecting signals that were expensive for someone else to give you. Ideas that survive that week are worth building; ideas that don't just saved you a year.
Sources & References
- CB Insights – The Top Reasons Startups Fail
- Rob Fitzpatrick – The Mom Test (customer interview methodology)
- Steve Blank – Customer Development essays
- U.S. Small Business Administration – Market Research and Competitive Analysis