Idea Validation
Overview
Kill bad ideas fast, confirm good ones cheaply. Walk through every phase in order. Each phase has a kill-check — if the idea fails, document why and stop before wasting further time.
Step 1: Problem Definition
Everything starts here. A vague problem = a vague business.
Answer these four questions precisely:
- Who — The exact person. Not "small businesses." Something like "freelance graphic designers juggling 3-8 client projects at once."
- What — The specific painful moment. "They spend 4+ hours/week manually exporting deliverables and coordinating revision feedback via email chains."
- Why it hurts — The real cost: time lost, revenue lost, stress, missed deadlines, damaged relationships. Quantify where possible.
- What they do now — Their current workaround. This IS your real competition — not just competitor apps, but the status quo itself.
Kill check: Cannot answer all four concretely → problem is not well-defined. Do more discovery first.
Step 2: Demand Signal Gathering
Prove real people care. Do not rely on assumptions or polite friends.
Check 3+ of these signal sources:
| Signal | Where | Positive Signal |
|---|---|---|
| Search volume | Google Trends, Ubersuggest free | Stable or growing volume on core problem keywords |
| Forum pain | Reddit, HN, Slack/Discord | Threads with 10+ comments describing this exact pain |
| Existing tool gaps | G2, App Store reviews | Tools solving adjacent problems with reviews citing the gap you'd fill |
| Job postings | LinkedIn, Indeed | Roles that exist only because this problem is expensive to solve manually |
| Social venting | Twitter/X search, LinkedIn | People publicly complaining about this unprompted |
Kill check: Fewer than 3 positive signals → problem may not be painful enough. Pivot or kill.
Step 3: Solution Fit Check
Pressure-test whether your proposed solution actually solves the problem well enough to build a business on.
- 10x rule: Is your solution 10x better (not 10%) than the current workaround in speed, cost, ease, or quality? Marginally better won't make people switch.
- Workflow change audit: Map exactly what the user must change in their current routine. High friction = low adoption.
- Solo-build feasibility: Can a working MVP be built by one person in weeks-to-a-few-months? If it needs a 10-person engineering team, that's a different company.
- Unfair advantage: Why you specifically? Skills, industry access, data, network, credibility — something competitors can't easily replicate.
Kill check: Fail the 10x rule or have zero unfair advantage → move on.
Step 4: Customer Discovery (Talk to Humans)
10-15 conversations with real potential customers. Non-negotiable. No amount of desk research replaces this.
Finding people:
- Post in 2-3 relevant communities asking for 15-min feedback chats.
- DM matching personas on LinkedIn or Twitter.
- Offer a small incentive if needed (gift card, free future access).
Conversation flow (15-20 min):
1. CONTEXT (2 min)
"Tell me about your work. Walk me through a typical week
around [problem area]."
→ Listen only. Do NOT pitch or explain your idea.
2. PAIN EXPLORATION (5 min)
"What's the most frustrating part of [workflow]?"
"How often does that come up?"
"What have you already tried to fix it?"
"What's still missing?"
→ Ask "why" and "what happens when" repeatedly. Dig.
3. OUTCOME PROBING (3 min)
"If a tool could [describe the outcome, not your feature set],
how would that change things?"
"What would that be worth — in time, money, stress?"
→ Let them quantify. Never suggest a number first.
4. WILLINGNESS TO PAY (2 min)
"Would you pay for something like that? What range feels fair?"
→ Discount stated numbers by ~50% when planning.
5. WARM REFERRAL (1 min)
"Know anyone else who hits this same wall? Mind if I reach out?"
Track across all conversations:
- % who described the problem without prompting (pain signal)
- % currently spending real money or time on workarounds (revenue signal)
- % with clear willingness to pay (demand signal)
- Features multiple people independently requested (product signal)
- Assumptions you held that conversations proved wrong (kill features early)
Kill check: Fewer than 60% confirm pain + willingness to pay → rework or kill.
Step 5: Riskiest Assumption Test (RAT)
One assumption, if wrong, kills everything. Find it. Test it in under 2 weeks and $200.
Common riskiest assumptions and cheap tests:
| Assumption | Minimum Test | Pass Threshold |
|---|---|---|
| People will pay | Landing page + "pre-order" button. Drive 200 targeted visitors. | 3-5% convert to payment/deposit |
| CAC is affordable | Run $100 targeted ad campaign. Measure cost-per-lead. | CPL < 20% of planned customer lifetime value |
| People will change workflow | Offer a manual version of the service to 5 people for free. See if they actually use it. | 3+ out of 5 use it consistently |
| I can build it solo | Build the single hardest technical feature as a prototype first. | Working prototype in ≤ 2 weeks |
Process:
- State the assumption.
- Define what "confirmed" looks like (a concrete, measurable number).
- Design the cheapest possible test.
- Set a hard deadline (max 2 weeks).
- Run it. Record results with zero spin.
Kill check: RAT fails → idea in current form is not viable. Pivot the solution, the customer, or the model — or kill entirely.
Step 6: Go / No-Go Scorecard
| Dimension | Score (1-5) | Weight |
|---|---|---|
| Problem severity & clarity | __ | 20% |
| Demand evidence gathered | __ | 15% |
| Customer discovery confirmation | __ | 20% |
| Solution fit (10x + feasibility) | __ | 15% |
| RAT result | __ | 20% |
| Your unfair advantage | __ | 10% |
Weighted score = Σ (score × weight)
- 4.0–5.0 → GO. Move to MVP planning + business model canvas.
- 3.0–3.9 → CONDITIONAL GO. Resolve the weakest dimension with one more test round first.
- < 3.0 → NO-GO. Kill or fundamentally pivot. Write down lessons.
Idea Validation Mistakes to Avoid
- Not documenting findings in one place — insights scattered across notes, emails, and chats become useless. Keep one running "Idea Validation" doc from day one.
- Letting validation drag on indefinitely. Time-box the entire process to 2-3 weeks max. Validation paralysis is as fatal as skipping it entirely.
- Testing one idea in isolation. Run 2-3 ideas in parallel when possible — the comparison sharpens judgment and reveals which assumptions you're making about all of them.
- Confusing polite interest with real demand. Friends and colleagues will say nice things. Pay-or-commit signals from strangers are the only ones that count.
- Skipping the kill checks. Each phase has a kill check for a reason. Ignoring a failed check and pressing forward wastes weeks or months on a dead-end idea.