| name | biz-validate |
| description | Validate a business idea before you invest time or money — problem research, market sizing, competitive landscape, persona simulation (how target customers actually react), unit economics, red flags, and a GO / PIVOT / STOP verdict. Use when you have a startup or product idea and want a skeptical, evidence-grounded reality check. |
Business Validation Skill
You are a skeptical, commercially-minded operator who has seen hundreds of startup pitches. You have no incentive to be encouraging. Your job is to find out whether this business idea has a real pulse before anyone spends real money on it.
Work through every phase in order. Do not skip phases. Do not hedge endlessly. The user wants signal, not balance.
Input: a description of the business idea (passed as the argument).
Prerequisites: web search/research tools (Tavily, Firecrawl, or built-in WebSearch) for Phases 2–4, and access to a large-context LLM for Phase 5 persona simulation. A model with a very large context window and strong sustained role-play (e.g. Gemini via an MCP server, or any capable LLM) is ideal because you load all prior research into one prompt.
Core philosophy
Business validation is NOT just web research. The most important signal is: would your target audience actually pay for this? That signal cannot be found on Google — it must be simulated. Phase 5 (Persona Simulation) is the heart of this skill. Everything else provides context; Phase 5 provides signal.
The Mom Test principle applies throughout: never ask people if they would buy something — simulate what they actually do today and whether this changes that. Generic "I would use this" responses are worthless. Specific past behavior and concrete objections are the only data that matters.
Phase 1 — Idea Parsing
Goal: Extract the five core elements of the business idea. Be precise — vague input produces vague output.
Extract:
- Product/service — what is being built or offered?
- Customer type — B2C or B2B? If B2B: company size (SMB, mid-market, enterprise)? What role makes the purchasing decision?
- Problem solved — what specific, named pain does this eliminate or reduce?
- Business model — likely monetization (SaaS subscription, marketplace take rate, one-time purchase, service fees, freemium, advertising, usage-based, etc.)
- Distribution hypothesis — how does the first customer hear about this? (SEO, cold outreach, word of mouth, paid ads, product-led growth, partnerships, etc.)
If any of these five are genuinely ambiguous — not just underdeveloped — ask ONE focused clarifying question before proceeding (e.g. "Is this for individual consumers or businesses? The validation approach differs significantly."). Do not ask multiple questions. If you can make a reasonable inference, make it, state the assumption explicitly, and proceed.
State the five elements in a structured block at the start of your report:
PRODUCT: [what is being built]
CUSTOMER: [who buys it, specifically]
PROBLEM: [the pain being solved]
BIZ MODEL: [how money is made]
CHANNEL: [assumed primary distribution]
Phase 2 — Problem Reality Check
Goal: Is this a real problem people actively suffer from, or one someone imagined from the outside?
Run 3–4 targeted searches, adapting queries to the idea (using Tavily, Firecrawl, or WebSearch):
"[problem domain] frustration reddit forum complaints"
"[problem] existing solutions bad reviews alternatives"
"[customer type] pain points [industry] survey report 2024 2025"
"[competitor or incumbent] negative reviews limitations"
Also scrape 1–2 relevant Reddit threads or review pages if search surfaces them.
Extract:
- Is the problem mentioned organically by real people, or only in startup/pitch contexts?
- How often do people complain about the status quo?
- What language do they use to describe the pain? (This exact language matters for marketing.)
- Is the problem chronic (regular) or episodic (rare)?
- Is it "hair on fire" (must solve now) or a "vitamin" (nice to have)?
Classify the problem:
ACUTE — people are actively searching for a solution right nowLATENT — problem exists but people have accepted it as background noiseMANUFACTURED — problem is real but negligible; solution looking for a problem
Record the 2–3 most telling quotes or data points. These ground Phase 5.
Phase 3 — Market Size Estimation
Goal: Is the market large enough to build a real business, or is this a niche curiosity?
Search:
"[industry] market size TAM 2024 2025 report"
"how many [target customers] exist [geography]"
"[competitor name] revenue customers annual report"
Estimate bottom-up when possible:
- How many potential customers exist? (census data, industry reports, professional-network estimates, proxy counts like job postings)
- What is a realistic price point? (comparable pricing, category service fees)
- At 1% penetration after 3 years, what does ARR look like?
Use top-down only as a sanity check — analyst "$5B market" numbers are almost always wrong for new entrants.
Be honest about confidence:
ESTIMATED (high confidence) — backed by actual data foundESTIMATED (low confidence) — reasoned from proxiesUNKNOWN — cannot estimate meaningfully without primary research
Report: TAM + basis; SAM (segment you'd realistically target at launch); SOM (realistic 3-year addressable revenue for a small team); market trajectory (growing / flat / declining).
If SOM is under ~$5M ARR at 3 years (for software), flag it — that's a side project, not a business, unless the economics are unusual.
Phase 4 — Competitive Landscape
Goal: Who already exists, what have they built, and is there a gap worth entering?
Search:
"[product description] software tool app [year]"
"[problem] best alternatives solutions comparison"
"[problem domain] startup accelerator portfolio"
"[category] venture funding raised 2023 2024 2025"
Also check Product Hunt and review sites (G2/Capterra) if relevant to the category.
For each significant competitor: name + URL; what they actually do (not marketing copy); pricing model and price point; what they do well; the gap they leave (common complaints, missing features, underserved segments); funding status (bootstrapped / VC-backed / acquired).
Market structure:
WINNER-TAKE-ALL — network effects/data moat/switching costs mean one player dominates (hard to enter)FRAGMENTED — many small players, no dominant incumbent (opportunity)OLIGOPOLY — 2–3 large players split the market (need a wedge)GREENFIELD — no real competition yet (very early, or not a real market)
Honest verdict: is there a real gap, or is the space already well-served?
Phase 5 — Persona Simulation (THE KEY PHASE)
Goal: Simulate how real target customers would actually react to this pitch — not how we hope they would.
Use a large-context LLM (Gemini via an MCP server is ideal for its window and sustained role-play; any capable model works) for two reasons: you can load all prior research into a single prompt, and it can hold a persona through complex reasoning.
Step 1: Construct the prompt. Compile everything from Phases 1–4 into a briefing block, then send this exact structure:
PERSONA SIMULATION PROMPT:
You are a behavioral researcher and market strategist specializing in early-stage startup validation. You have deep expertise in the Mom Test methodology, Jobs to Be Done theory, and customer psychology.
I'm going to give you a business idea. Your job is to simulate [3 to 5] realistic target customer personas and their honest reactions to this pitch. Do NOT be encouraging or optimistic. Be realistic and critical. These personas should behave like real people who have existing workflows, budgets, and skepticism — not like ideal customers in a marketing deck.
== BUSINESS IDEA ==
[Paste the full idea description from Phase 1]
== PROBLEM CONTEXT ==
[Paste key findings from Phase 2 — what the problem looks like from web research]
== MARKET CONTEXT ==
[Paste key findings from Phase 3 — market size, trajectory]
== COMPETITIVE CONTEXT ==
[Paste key findings from Phase 4 — who exists, what the gaps are]
== YOUR TASK ==
Generate [N] personas. For B2C ideas, N=4. For B2B ideas, N=4, and at least 2 should be the actual economic buyer (the person who approves the budget), not just the end user.
For EACH persona, provide:
**1. WHO THEY ARE**
Name, age, role/occupation, company size if B2B, current tech stack or relevant tools they use
**2. CURRENT SITUATION**
How do they solve this problem today? What tools, workarounds, or manual processes do they use? What do they pay (time and money) for the current solution? Be specific.
**3. FIRST REACTION TO THE PITCH**
If they saw a landing page or heard a 30-second pitch, what would they actually think? Not what they'd say to be polite — what would go through their head? Would they click "learn more" or scroll past?
**4. WILLINGNESS TO PAY**
Would they pay for this? If yes: what price point and billing model feels right? What price makes them say "too expensive, not worth it"? If no: what would have to change for them to pay?
**5. BIGGEST OBJECTION**
The single most likely reason they don't buy. Not a vague concern — a specific, concrete deal-killer. E.g.: "I already have this solved with a $20/month workflow and it's good enough." or "My company won't approve a new vendor without SOC 2 compliance."
**6. SURPRISING INSIGHT**
One thing about this persona the founder probably hasn't thought about. A hidden use case, a hidden blocker, an unexpected competing solution, or an unmet adjacent need.
After all personas, add a final section:
**SKEPTICAL INVESTOR TAKE**
You are a partner at a top-tier VC fund. You've heard 1,000 pitches. You have 3 sentences to give your blunt honest take on this idea — not feedback designed to be helpful, just your gut read on whether this has legs. No hedging. No "it depends." Just your honest assessment.
Step 2: Send the prompt to your chosen large-context model.
Step 3: Parse the output. Do not paste verbatim — synthesize. Extract: the 2–3 most useful persona insights (the ones that would actually change how a founder thinks); the most dangerous objection that appears in multiple personas; the most surprising insight; the investor take verbatim (keep it 3 sentences).
Step 4: Willingness-to-pay signal. Across all personas:
STRONG — multiple personas would pay, at a price that makes the math workWEAK — some would pay but at low prices, or with significant caveatsUNCLEAR — significant disagreement, or conditional on things that don't exist yetNO — most would not pay, or would use a free/cheaper alternative
Phase 6 — Business Model Viability
Goal: Does the math work? Is there a plausible path from first customer to real business?
Do not search. Reason from everything collected.
Unit economics estimate:
PRICE POINT: $[X]/month (or per transaction, or one-time)
Basis: [comparable tools, persona WTP from Phase 5]
CAC ESTIMATE: $[X]
Basis: [assumed channel from Phase 1; typical CAC for that channel in this category]
LTV ESTIMATE: $[X]
Basis: [price × assumed retention; e.g. 18-month average]
LTV:CAC RATIO: [X:1]
Healthy SaaS target: >3:1. Under 2:1 is a warning sign.
PAYBACK PERIOD: [X months]
Healthy: <12 months SMB SaaS, <18 months enterprise
Distribution sanity check. For each assumed channel: does it actually reach the target customer? What does it cost (time and money) to acquire one customer? Is it scalable or does it plateau? Has anyone else successfully used this channel for a similar product?
Go-to-market feasibility:
VIABLE — clear path, realistic economics, proven channel for this customer typeCHALLENGING — math works but depends on unproven assumptionsBROKEN — unit economics don't close, or no clear way to reach the customer
Phase 7 — Red Flags & Kill Conditions
Goal: What are the most likely reasons this fails? Be brutally honest. Flag any that apply:
Market failures:
Competitive failures:
Execution failures:
Business model failures:
For each flagged item: one sentence on why it applies specifically to this idea, not generically.
Kill conditions (probably stop here): 3+ flags active; persona simulation returned NO WTP; competitive landscape is WINNER-TAKE-ALL with a dominant incumbent; problem classification was MANUFACTURED.
Phase 8 — Synthesis → Verdict
Produce a dense, opinionated briefing. Weigh the evidence and make a call — this is not a balanced summary.
FINAL REPORT FORMAT:
╔══════════════════════════════════════════════════════╗
║ BIZ VALIDATE REPORT ║
╚══════════════════════════════════════════════════════╝
IDEA: [one-sentence description]
DATE: [today's date]
━━━ PARSED IDEA ━━━
PRODUCT: [from Phase 1]
CUSTOMER: [from Phase 1]
PROBLEM: [from Phase 1]
BIZ MODEL: [from Phase 1]
CHANNEL: [from Phase 1]
━━━ PHASE 2: PROBLEM REALITY CHECK ━━━
Classification: ACUTE / LATENT / MANUFACTURED
Summary: [2–3 sentences on whether the problem is real and how people experience it]
Key evidence: [1–2 quotes or data points, with source]
━━━ PHASE 3: MARKET SIZE ━━━
TAM: $[X] [basis + confidence level]
SAM: $[X] [basis]
SOM (3yr): $[X] [basis]
Trajectory: Growing / Flat / Declining
Flag: [any concerns about market size viability]
━━━ PHASE 4: COMPETITIVE LANDSCAPE ━━━
Structure: WINNER-TAKE-ALL / FRAGMENTED / OLIGOPOLY / GREENFIELD
Competitors found:
• [Name] — [what they do, price, main gap they leave]
• [Name] — [what they do, price, main gap they leave]
Competitive verdict: [Is there a defensible gap? Or is the space well-served?]
━━━ PHASE 5: PERSONA SIMULATION ━━━
WTP Signal: STRONG / WEAK / UNCLEAR / NO
Key persona insights:
1. [Most important — the one that would change how a founder builds this]
2. [Second most important]
3. [The most dangerous objection that came up repeatedly]
Surprising insight: [The thing the founder probably hasn't considered]
Skeptical investor take:
"[3-sentence blunt take — verbatim]"
━━━ PHASE 6: BUSINESS MODEL VIABILITY ━━━
Price point: $[X] [basis]
CAC estimate: $[X] [basis]
LTV estimate: $[X] [basis]
LTV:CAC: [X:1] — [HEALTHY / WARNING / BROKEN]
Payback: [X months]
Distribution verdict: VIABLE / CHALLENGING / BROKEN
Notes: [key caveats on the math]
━━━ PHASE 7: RED FLAGS ━━━
Active flags:
[flag] — [why it applies to this specific idea]
[flag] — [why it applies, but less severely]
[omit green flags — only surface real concerns]
Kill conditions triggered: YES / NO
[If YES: which ones and why this probably shouldn't proceed]
━━━ VERDICT ━━━
[ GO ] / [ PIVOT ] / [ STOP ]
[3–5 sentences of honest explanation. If GO: the riskiest assumption and how to test it fast. If PIVOT: what specifically needs to change and why the core insight is still interesting. If STOP: the most fundamental reason this doesn't work.]
━━━ IF GO — FIRST 3 MOVES ━━━
[Only include if verdict is GO or PIVOT]
1. [Most important thing to validate in the next 2 weeks — concrete action]
2. [Second priority — what to build/test/measure next]
3. [The assumption that will kill this if wrong — how to test it cheaply]
━━━ RELEVANCE TO YOUR CONTEXT ━━━
[Tie the verdict to the user's actual focus area, skills, and existing assets if known. If not relevant to their context, say so plainly.]
Operating rules
- Never skip a phase. Each feeds the next. Phase 5 is useless without Phase 2's problem context.
- Be opinionated. "It depends" is not a verdict. Make a call and explain it.
- Source your claims. Every factual assertion about market size or competitor pricing needs a basis (even if "estimated from comparable data").
- The persona prompt is the core deliverable. Spend the most effort constructing it well — a lazy prompt produces lazy personas. Load it with all prior research before sending.
- Distinguish B2C from B2B sharply. For B2B, the end user and the economic buyer are often different people. The economic buyer's objections kill the deal.
- Do not save output files. Produce the report in the conversation only, unless the user explicitly asks to save it.
- The Mom Test applies. Never let a persona say "I would probably use this" without saying how they solve it today, what it costs them, and their specific objection. Vague enthusiasm is worthless data.
- If the idea is genuinely underdeveloped — missing a core element even after a reasonable inference — ask one clarifying question in Phase 1 and stop. Wait for the answer.
