validate（validate 文档）

Idea Validation

The #1 reason startups fail is "no market need." Validation isn't about asking people if they'd use something — it's about observing whether they'll pay, sign up, or take action. This skill helps you test demand before writing a single line of code.

Core Principles

• Ideas are free. Validated demand is valuable. Never skip validation because you're excited.
• "Would you use this?" is a useless question. "Will you pay $X right now?" is the only one that matters.
• The goal of validation is to fail fast and cheap — not to confirm what you already believe.
• You don't need to build anything to validate. Landing pages, waitlists, and conversations come first.
• Validation is not a one-time event. You re-validate at every stage: idea, MVP, pricing, features.

Pressure-Test Your Idea

Before running experiments, pressure-test the idea itself. These six questions expose fatal flaws fast — answer them honestly, not optimistically.

Which Questions to Answer

The Six Questions

Q1 — Demand Reality: What evidence do you have — beyond your own experience — that someone else actually wants this? Not "I think people need it." What have you seen, heard, or measured?

Q2 — Status Quo: What are people in your field doing right now to handle this — even badly? What does that workaround cost them in time, money, or errors?

Q3 — Desperate Specificity: Name one specific person who needs this most. Not "dentists" — which dentist, at which practice, with what problem? If you can't name someone, you haven't found your customer yet.

Q4 — Narrowest Wedge: What's the smallest version of this someone would pay for this week — not after you build the platform? One screen, one workflow, one outcome.

Q5 — Observation: Have you watched a colleague struggle with this task without helping them? What surprised you about how they actually do it vs. how you assumed?

Q6 — Future-Fit: How does your industry change in 3 years, and does that make this tool more essential or less?

Interest is not demand. Waitlist signups are not demand. Someone would be genuinely upset if it disappeared — that's demand.

"Everyone in my field needs this" means you haven't found anyone specific yet. The more universal you think the need is, the less validated it actually is.

The status quo is your real competitor — not the other startup. It's the spreadsheet-and-email workaround people already live with. You have to be dramatically better than "good enough."

Validation Levels

Level 1: Problem Validation (Do People Have This Problem?)

Before you validate your solution, validate that the problem exists and is painful enough to pay for.

Where to look for evidence:

Tell AI:

Research the problem of [describe the problem].
Find evidence that people are actively looking for solutions:
- Search volume for related terms
- Reddit/forum threads where people discuss this pain
- Competitors that exist (even partial solutions)
- How much people currently pay to solve this (or workarounds they use)
Summarize: Is this a real, painful, frequent problem?

Level 2: Solution Validation (Will People Want YOUR Solution?)

The Mom Test — Never ask leading questions. Instead:

Conversation template:

1. "What's the hardest part about [area]?"
2. "Tell me about the last time that happened."
3. "How did you deal with it?"
4. "What didn't work about that solution?"
5. "If you could wave a magic wand, what would change?"
6. "How much time/money does this cost you today?"

Talk to 10-15 potential customers. If 8+ describe the same pain with intensity, you have signal.

Level 3: Willingness to Pay (Will They Open Their Wallets?)

The strongest validation signals, ranked:

For Domain Experts: Your Network Is Your Validation Lab

If you're a [profession] building for other [professionals], you already have what most founders spend months trying to get: direct access to target customers.

• Skip the cold outreach. Message 10 peers you actually know: "Hey, how do you handle [pain]? I'm thinking about building something."
• You've already had 1,000 customer conversations. Mine your memory: what do colleagues complain about at conferences, in group chats, over lunch?
• Your professional associations are focus groups. Post in the group: "Quick question — how long does [task] take you?" Count the replies.
• Validate in days, not weeks. You don't need to "find" your market. You're standing in it.

See translate skill for identifying which pain is worth building for, and niche-advantage for leveraging your professional network.

Smoke Tests (Validate Without Building)

Landing Page Test

Create a landing page describing your product. Drive traffic. Measure signups.

Tell AI:

Create a landing page for [product idea] that:
- Clearly describes the problem and solution
- Has a CTA: "Join the waitlist" or "Get early access"
- Collects email addresses
- Optionally asks 1-2 qualifying questions (role, company size)
Target: 100 visitors, measure signup rate.

Benchmarks:

• < 5% signup rate → Weak interest. Rethink positioning or audience.
• 5-15% signup rate → Moderate interest. Worth exploring further.
• 15%+ signup rate → Strong signal. Build an MVP.

Fake Door Test

Add a button or link for a feature that doesn't exist yet. Measure clicks.

1. Create a CTA for the feature: "Try [Feature Name]"
2. When clicked, show: "This feature is coming soon!
   Sign up to be the first to know."
3. Collect email.
4. Measure click rate.

Pre-Sale Test

Offer the product at a discount before it exists. If people pay, you have validation.

"[Product] launches in [timeframe]. Get 50% off as a founding member.
$X/month (normally $Y/month). Cancel anytime."

If 10+ strangers pay, build it. If 0 pay, pivot.

Go / No-Go Decision Framework

After running validation experiments, score your idea:

Validation Scorecard:
                                          Score (1-5)
Problem frequency (daily=5, yearly=1):    ___
Problem intensity (hair on fire=5):       ___
Willingness to pay (pre-paid=5):          ___
Market size (>$1B TAM=5):                 ___
Your unique advantage (deep=5):           ___
Current solutions (none/bad=5):           ___
                                   Total: ___/30

25-30: Strong go. Build the MVP.
18-24: Promising. Run one more validation experiment.
12-17: Weak. Pivot the angle or audience.
<12:   No go. Find a different problem.

Where to Find People to Validate With

Tell AI:

Help me find 5 specific online communities where [target audience]
hangs out and discusses [problem area]. For each, give me:
- The community name and link
- How active it is
- Rules about self-promotion
- A non-spammy way to start conversations about [problem]

Validation Timeline

Week 1: Problem research + 5 customer conversations
Week 2: 5 more conversations + landing page live
Week 3: Drive traffic to landing page (100+ visitors)
Week 4: Analyze results, make go/no-go decision

Total cost: $0-200
Total time: 10-15 hours

Common Mistakes

Success Looks Like

• Clear evidence of demand before writing a single line of code
• 10+ customer conversations documented with recurring pain points
• Landing page with measurable signup rate
• Go/no-go decision backed by data, not gut feeling
• Confidence that you're building something people will pay for

Related Skills

• translate — Turn your professional expertise into a product spec (start here if you're a domain expert)
• niche-advantage — Leverage your industry network and credibility for distribution
• customer-research — Go deeper with interviews and personas
• market-research — Size the market and analyze competitors
• landing-page — Build the landing page for your smoke test
• plan — Write the spec once you've validated the idea
• pricing — Validate willingness to pay alongside demand