Contact-Finding Subskill

Purpose

Identify decision-makers at target companies and provide contact research strategy (LinkedIn queries, email patterns, phone research methods) based on ICP buyer personas.

Operates on: ONE product per invocation Input: Qualified prospect list + ICP buyer personas Output: Contact research strategy with LinkedIn queries, email patterns, phone methods

Context

• Reads from: research/customer/prospects/[segment]-prospects-{date}.csv (target companies)
• Reads from: research/customer/icp/[segment]-icp.md (personas - decision-maker titles)
• Writes to: research/customer/prospects/[segment]-contacts-{date}.csv
• Uses WebSearch for LinkedIn discovery, email pattern detection, phone research

Key Workflows

1. Load Target Companies

Read prospect list:

• Company names and domains from prospect CSV
• Filter by rows parameter (default: 1-20)
• Priority: Research Tier 1 prospects first

Read ICP personas:

• Extract buyer personas section from ICP YAML
• Identify decision-maker titles by buyer type:
  • Economic buyer (budget authority)
  • Technical buyer (evaluation authority)
  • End user (implementation user)
• Map persona focus parameter to titles

Example ({Product Name}):

buyer_personas:
  economic_buyer:
    - Founder/CEO
    - CFO
    - CMO
  technical_buyer:
    - {Technical role 1}
    - {Technical role 2}
  end_user:
    - {End user role 1}
    - {End user role 2}

2. Identify Decision-Maker Titles

Per company, determine target titles:

If persona_focus = "economic":

• Target only economic buyer titles
• Example: Founder, CFO, CMO

If persona_focus = "technical":

• Target only technical buyer titles
• Example: Ecommerce Director, Head of Operations

If persona_focus = "end_user":

• Target only end user titles
• Example: Customer Service Manager

If persona_focus = "all" (default):

• Start with economic buyer (primary)
• Add technical buyer if product requires technical evaluation
• Optionally add end user for product-led growth

Prioritization:

1. Economic buyer first (budget authority)
2. Technical buyer second (if product is technical)
3. End user third (for PLG or bottoms-up)

3. Research Contacts

For each company + title combination:

LinkedIn Search Strategy

Primary pattern:

site:linkedin.com/in "{company name}" "{title}"

Examples:

site:linkedin.com/in "{Company Name}" "Founder"
site:linkedin.com/in "{Company Name}" "CMO"
site:linkedin.com/in "{Company Name}" "{Technical Role}"

Verification indicators (look for in search results):

• Current position at target company
• Recent LinkedIn activity (posted/commented in last 90 days)
• Profile completeness (photo, headline, experience)
• Connection count (500+ = active user)

Confidence scoring:

• High: Multiple matching profiles, recent activity, complete profiles
• Medium: 1-2 matching profiles, some activity
• Low: No clear matches, incomplete profiles, inactive accounts

Email Pattern Detection

Pattern discovery methods:

1. Company website contact page:

site:{domain} "contact" OR "team" OR "about"

2. Public email signatures:

"{domain}" "email" "@{domain}"

3. Press releases or news:

"{company name}" "contact" "@{domain}"

Common B2B patterns:

• firstname.lastname@domain.com (most common)
• first.last@domain.com
• flast@domain.com
• firstnamel@domain.com
• firstname@domain.com (small companies)

Pattern verification:

• If you find 2+ emails with same pattern → High confidence
• If you find 1 email showing pattern → Medium confidence
• If no emails found, infer from company size:
  • <50 employees: firstname@domain.com
  • 50-200 employees: firstname.lastname@domain.com

  • 200 employees: first.last@domain.com or flast@domain.com

  • Low confidence for inferred patterns

Phone Research Methods

Phone discovery strategy:

1. Company website:

site:{domain} "phone" OR "call" OR "contact us"

2. LinkedIn company page:

site:linkedin.com/company "{company name}"

Look for: Company phone number in About section

3. Industry directories:

• Chamber of Commerce listings
• Professional association directories
• Regulatory databases (for regulated industries)

Phone research output:

• Document WHERE to find phone (company page, LinkedIn, directory)
• Do NOT include actual phone numbers (privacy)
• Note: "Company phone: {source}", "LinkedIn profile: {person name}"

Confidence scoring:

• High: Company phone on website, LinkedIn profile shows contact info
• Medium: Company phone only (no direct line)
• Low: No phone found, will need outbound discovery call

4. Output Contact Report

CSV structure:

company_name,domain,target_title,buyer_persona,linkedin_search_query,email_pattern,phone_research_method,confidence

Example rows:

{Company A},{companya.com},Founder,economic,"site:linkedin.com/in ""{Company A}"" ""Founder""",firstname.lastname@companya.com,Company website contact page,high
{Company A},{companya.com},CMO,economic,"site:linkedin.com/in ""{Company A}"" ""CMO""",firstname.lastname@companya.com,Company website contact page,medium
{Company B},{companyb.com},{Technical Role},technical,"site:linkedin.com/in ""{Company B}"" ""{Technical Role}""",first.last@companyb.com,LinkedIn company page,high

Confidence calculation:

• High: LinkedIn high + Email high + Phone high/medium
• Medium: LinkedIn medium or Email medium or Phone low
• Low: LinkedIn low or Email low or Phone not found

Input Parameters

Required:

• product: Product name (e.g., "{Your Product}")
• prospect_list_path: Path to prospect CSV (default: research/customer/prospects/{segment}-prospects-{date}.csv, use most recent)
• icp_path: Path to ICP (default: research/customer/icp/{segment}-icp.md)

Optional:

• rows: Which prospects to research (default: "1-20")
  • "1-10": First 10 prospects
  • "1-50": First 50 prospects
  • "all": All prospects (use with caution for large lists)
• persona_focus: Which buyer personas to target (default: "all")
  • "economic": Economic buyers only (Founder, CFO, CMO)
  • "technical": Technical buyers only (CTO, VP Eng, Head of Ops)
  • "end_user": End users only (Manager, Coordinator)
  • "all": All personas (prioritize economic, then technical, then end user)

Output

File: research/customer/prospects/{segment}-contacts-{date}.csv

Columns:

1. company_name: Target company name
2. domain: Company domain
3. target_title: Decision-maker title to search for
4. buyer_persona: economic/technical/end_user
5. linkedin_search_query: Exact LinkedIn search query to run
6. email_pattern: Detected or inferred email format
7. phone_research_method: Where to find phone (source description)
8. confidence: high/medium/low (overall contact research confidence)

Web Search Patterns

LinkedIn Discovery

Pattern 1: Title-based search:

site:linkedin.com/in "{company name}" "{title}"

Pattern 2: Seniority-based search (if title yields no results):

site:linkedin.com/in "{company name}" "Director"
site:linkedin.com/in "{company name}" "VP"
site:linkedin.com/in "{company name}" "Head of"

Pattern 3: Department-based search:

site:linkedin.com/in "{company name}" "Marketing"
site:linkedin.com/in "{company name}" "Operations"
site:linkedin.com/in "{company name}" "Finance"

Email Pattern Detection

Pattern 1: Direct contact page:

site:{domain} "contact" OR "team"

Pattern 2: About/Team page:

site:{domain} "about" OR "our team"

Pattern 3: Press/Media page:

site:{domain} "press" OR "media" OR "news"

Pattern 4: Public email signatures:

"{domain}" "@{domain}"

Phone Research

Pattern 1: Contact page:

site:{domain} "phone" OR "call us"

Pattern 2: LinkedIn company page:

site:linkedin.com/company "{company name}"

Pattern 3: Business directories:

"{company name}" "phone" OR "contact"

Contact Research Strategy

Priority Order (Pete Kazanjy Method)

1. LinkedIn first (most reliable for B2B):

   • Current employment verification
   • Profile activity indicates engagement
   • Direct messaging capability (for Sales Navigator users)
   • Connection requests as outreach channel

2. Email pattern second (primary outreach channel):

   • Verify pattern from company website/public sources
   • If unverifiable, use conservative inferred pattern
   • Avoid guessing (use email validation tools if needed)

3. Phone third (supplementary, critical per Kazanjy):

   • Company phone as starting point
   • Ask for decision-maker by name/title
   • LinkedIn profile contact info (if available)
   • Direct line research (post-connection)

Verification Best Practices

Before outputting contact:

• ✓ LinkedIn profile shows current position at target company
• ✓ Email pattern verified by 2+ examples OR is conservative inference
• ✓ Phone research method documented (even if "not found")
• ✓ Confidence score reflects weakest link in contact research

Red flags requiring manual review:

• No LinkedIn profiles found for any target title
• Email pattern conflicts (multiple patterns detected)
• No company contact information available (phone/email)
• All confidence scores "low"

Company Size Adjustments

Small companies (<50 employees):

• Focus on Founder/CEO (often wears multiple hats)
• Email: firstname@domain.com (simpler patterns)
• Phone: Company phone likely reaches decision-maker directly
• LinkedIn: Founder very active, high response rate

Medium companies (50-200 employees):

• Multiple decision-makers (CFO, CMO, CTO separate)
• Email: firstname.lastname@domain.com (standard corporate)
• Phone: Receptionist/gatekeeper, ask for specific person
• LinkedIn: Mix of active/inactive profiles

Large companies (200+ employees):

• Complex hierarchy, multiple approvers
• Email: first.last@domain.com or flast@domain.com (IT-managed)
• Phone: Complex phone tree, direct lines hard to find
• LinkedIn: High profile count, many inactive profiles

Constraints

Decision-Maker Title Mapping

Must map to ICP personas:

• All target titles must appear in ICP buyer personas section
• If title not in ICP, do not research (out of scope)
• If ICP has no personas, return error (ICP incomplete)

Example validation:

# ICP has these personas:
buyer_personas:
  economic_buyer: [Founder, CFO]
  technical_buyer: [CTO]

# Valid targets: Founder, CFO, CTO
# Invalid targets: COO, CMO (not in ICP)

Email Pattern Verification

Verified patterns (high/medium confidence):

• Found 2+ emails with same pattern on company website
• Found 1 email with pattern in press release/news
• Pattern confirmed by email verification tool

Inferred patterns (low confidence):

• No emails found, inferred from company size
• Must use conservative pattern (firstname.lastname@domain.com)
• Never guess creative patterns (flast@, f.lastname@, etc.)

Phone Research (Privacy)

Allowed:

• Document WHERE to find phone (source)
• General company phone from website
• LinkedIn profile shows "contact info available"

Not allowed:

• Direct phone numbers in CSV (privacy violation)
• Personal mobile numbers
• Unsolicited phone list scraping

Output format:

phone_research_method
"Company website contact page"
"LinkedIn company page"
"Not found - will need discovery call"

LinkedIn Query Specificity

Must be specific:

• Include company name AND title in query
• Use exact match quotes: "Company Name" "Title"
• Use site:linkedin.com/in for profile search

Examples:

✓ site:linkedin.com/in "{Company Name}" "Founder"
✓ site:linkedin.com/in "{Company Name}" "CMO"
✗ "CMO {industry}" (too broad)
✗ site:linkedin.com "{Technical Role}" (no company)

Error Handling

No prospects found:

• Check prospect_list_path exists
• Check rows parameter is valid range
• Return error: "No prospects found in {path} for rows {rows}"

No personas in ICP:

• Check icp_path has buyer_personas section
• Return error: "ICP missing buyer_personas section"

Invalid persona_focus:

• Check parameter is one of: economic, technical, end_user, all
• Return error: "Invalid persona_focus: {value}. Must be: economic, technical, end_user, or all"

No contacts found for company:

• Document in CSV with confidence: "low"
• phone_research_method: "Not found - manual research needed"
• linkedin_search_query: Still provide query to run manually

Quality Validation

Before finalizing CSV, verify:

• All companies from prospect list (rows range) are included
• All target titles map to ICP buyer personas
• LinkedIn queries use correct syntax (site:linkedin.com/in)
• Email patterns are verified or conservatively inferred
• Phone research methods documented (even if "not found")
• Confidence scores calculated correctly
• No actual phone numbers in CSV (only methods)

Output review:

• Total contacts researched: {count}
• High confidence: {count} ({percent}%)
• Medium confidence: {count} ({percent}%)
• Low confidence: {count} ({percent}%)
• Companies with no contacts found: {list}

References

See references/ directory for:

• contact-patterns.md: Decision-maker titles by industry and product type
• email-format-detection.md: Common email patterns and detection methods
• phone-research-methods.md: Where to find phone numbers, when to call