icp-persona-builder

# ICP Persona Builder

Research a company's buyer segments and build detailed synthetic personas that model their ideal customers. These personas become a reusable client asset — once built, any skill can load them to evaluate content, messaging, websites, or campaigns through buyer eyes.

## Quick Start

```
Build ICP personas for [company]. Their site is [url].
```

With known ICPs:
```
Build personas for [company]. Their ICPs are: [ICP 1], [ICP 2], [ICP 3].
```

## Inputs

| Input | Required | Source |
|-------|----------|--------|
| **Company name** | Yes | User provides |
| **Company URL** | Recommended | Helps with research |
| **Known ICPs** | No | User provides, or discovered via research |
| **Client context file** | No | `clients/<client>/context.md` if available |

## Process

### Phase 1: Company Research

Understand what the company does and who they serve:

1. **WebFetch their website** — homepage, product/solutions pages, pricing, "who it's for" pages
2. **WebSearch** for:
   - "[company] customers" / "[company] case studies"
   - "[company] reviews" (G2, Capterra, TrustRadius)
   - "[company] vs" (comparison searches reveal buyer segments)
   - "[company] jobs" (who they're hiring to sell to / support)
3. **Extract signals:**
   - What problem do they solve?
   - What's their pricing/packaging? (Signals ACV and buyer type)
   - What industries/verticals do they serve?
   - What company sizes do they target?
   - What roles/titles appear in case studies and testimonials?
   - What's their go-to-market motion? (Self-serve, sales-led, hybrid)

### Phase 2: Identify ICP Segments

From the research, identify **4-6 distinct buyer segments**. Each segment should represent a meaningfully different type of buyer — different role, different company profile, or different buying motivation.

For each segment, define:

| Attribute | Description |
|-----------|-------------|
| **Segment name** | Short label (e.g., "Enterprise IT Leader", "Startup Founder", "Agency Operator") |
| **Role/titles** | Typical job titles in this segment |
| **Company profile** | Size, stage, industry, tech stack |
| **Core pain point** | The #1 problem driving them to look for a solution |
| **Buying trigger** | What event makes them start searching NOW |
| **Decision criteria** | What matters most when evaluating (ranked) |
| **Sophistication** | How well they understand the problem space and solution landscape |
| **Alternatives** | What else they'd consider (competitors, DIY, status quo) |
| **Segment size estimate** | Rough sense of how big this segment is for the company (primary, secondary, emerging) |

**Segment diversity rules:**
- At least one **technical** buyer (evaluates capabilities, architecture, integrations)
- At least one **business** buyer (evaluates ROI, outcomes, competitive advantage)
- At least one **skeptical** profile (has been burned before, hard to convince)
- At least one **junior/researcher** (doing initial research for a decision-maker)
- Try to cover different company sizes if the company serves multiple tiers

### Phase 3: Build Synthetic Personas

For each segment, create a detailed synthetic persona. The persona should feel like a real, specific person — not a marketing abstraction.

**Persona structure:**

```json
{
  "id": "persona-slug",
  "name": "Jordan Chen",
  "segment": "Enterprise IT Leader",
  "title": "VP of Engineering",
  "company": {
    "type": "Mid-market SaaS company",
    "size": "200-500 employees",
    "stage": "Series B, scaling fast",
    "industry": "Financial services technology"
  },
  "demographics": {
    "experience_years": 12,
    "reports_to": "CTO",
    "team_size": 35,
    "budget_authority": "$50K-200K without board approval"
  },
  "situation": "Jordan's team is growing faster than their tooling can support. They've been using a patchwork of internal scripts and are losing engineering hours to maintenance. The CTO has asked Jordan to evaluate modern solutions before next quarter's planning cycle.",
  "pain_points": [
    "Team productivity is dropping as they scale",
    "Current tools don't integrate well",
    "Onboarding new engineers takes too long"
  ],
  "buying_trigger": "CTO mandate to evaluate solutions before Q3 planning",
  "decision_criteria_ranked": [
    "Enterprise security and compliance (SOC2, SSO)",
    "Integration with existing stack (GitHub, Jira, Datadog)",
    "Scalability — will this work at 2x team size?",
    "Total cost of ownership, not just sticker price",
    "Implementation timeline — needs to be live in 6 weeks"
  ],
  "skepticism_profile": {
    "trust_level": "Low — has been burned by vendor promises before",
    "research_style": "Deep dive. Reads docs, checks GitHub issues, asks peers in Slack communities",
    "key_objections": [
      "Will this actually scale or will we outgrow it in a year?",
      "What's the real implementation cost beyond the license?",
      "How good is the support when things break at 2am?"
    ]
  },
  "technical_sophistication": "High — understands the technical landscape well, can evaluate architecture decisions, wants to see under the hood",
  "language": {
    "describes_problem_as": "We need to consolidate our toolchain and reduce operational overhead",
    "searches_for": ["engineering productivity platform", "developer tools consolidation", "[competitor] alternative enterprise"],
    "red_flag_words": ["revolutionary", "AI-powered", "seamless" — overpromising triggers skepticism],
    "trust_signals": ["SOC2 badge", "customer logos in their industry", "transparent pricing", "public changelog"]
  },
  "evaluation_behavior": {
    "first_visit": "Scans headline, checks if it's for their company size, looks for enterprise/security page",
    "deep_evaluation": "Reads docs, checks integrations list, looks for case studies from similar companies",
    "social_proof_needs": "Wants to see companies their size in their industry, not just FAANG logos",
    "deal_breakers": ["No SSO/SAML", "No self-hosted option", "Pricing only available via sales call"]
  }
}
```

### Phase 4: Save Persona Assets

Save to the client directory as reusable assets:

**`clients/<client>/personas/personas.json`** — Machine-readable, all personas in an array:
```json
{
  "company": "Acme Corp",
  "url": "https://acme.com",
  "created": "2026-02-26",
  "segment_count": 5,
  "personas": [ ... ]
}
```

**`clients/<client>/personas/personas.md`** — Human-readable Markdown with all personas written out in prose form, easy to review and share.

**`clients/<client>/personas/segments.md`** — Summary table of all segments with key attributes, useful as a quick reference.

## Output Summary

After building, present:
1. **Segment overview table** — All segments with key attributes at a glance
2. **Persona summaries** — 2-3 sentence summary of each persona
3. **Coverage check** — Confirm diversity rules are met (technical, business, skeptical, researcher)
4. **Next steps** — Suggest running `icp-website-review` or other skills that can use the personas

## Tips

- **Research depth matters.** Spend real time in Phase 1. The better you understand the company's actual customers, the more realistic the personas. Don't just read the homepage — dig into reviews, case studies, job postings.
- **Make personas specific.** "Marketing Manager" is too generic. "Sarah, Senior Demand Gen Manager at a 50-person B2B SaaS startup who just lost her SDR team to budget cuts" tells you exactly how she'll evaluate a tool.
- **Include the language dimension.** How the persona describes their problem is often completely different from how the vendor describes their solution. This gap is where messaging fails.
- **Skepticism is the most important trait.** Every persona needs a clear skepticism profile. What would make them NOT buy? What's their default assumption about vendors?
- **This skill has no code script.** It's agent-executed using WebSearch and WebFetch. The structured process above guides the research and persona creation.

## Dependencies

- Web search capability (for company and ICP research)
- Web fetch capability (for reading website pages)
- No API keys or paid tools required