abilityai/trinity/dd-verification

Claim Verification Methodology

When analyzing claims from pitch decks, founder statements, or company materials, follow this rigorous verification process.

Verification Levels

Assign one of these confidence levels to every claim:

Verification Process

1. Identify the Claim

Extract the specific, testable assertion:

• ❌ "We're growing fast" (vague)
• ✅ "We grew 300% YoY in 2024" (specific, testable)

2. Find Independent Sources

Never rely solely on company-provided materials. Seek:

• Public filings (SEC, state records)
• Third-party data providers (Crunchbase, PitchBook, LinkedIn)
• Press coverage (with dates)
• Customer reviews and testimonials
• Industry reports from analysts

3. Cross-Reference

Compare across sources:

• Do numbers align within reasonable variance (±10%)?
• Are timelines consistent?
• Do different sources tell the same story?

4. Document Everything

For each verified claim, record:

{
  "claim": "The specific claim text",
  "confidence": "VERIFIED | PARTIALLY VERIFIED | UNVERIFIED | CONTRADICTED",
  "sources": [
    {"name": "Source name", "url": "URL if available", "date": "YYYY-MM-DD"}
  ],
  "notes": "Any caveats or context"
}

Red Flags

Watch for these warning signs:

• Round numbers without context ("$10M ARR exactly")
• Claims that can't be independently verified
• Metrics that don't match industry standards
• Vague timeframes ("recently", "soon")
• Comparisons without clear methodology

Output Format

Include a verification summary in your analysis:

{
  "verification_summary": {
    "total_claims_analyzed": 15,
    "verified": 8,
    "partially_verified": 4,
    "unverified": 2,
    "contradicted": 1
  },
  "key_concerns": [
    "Market size claim contradicted by industry reports",
    "Founder employment dates don't match LinkedIn"
  ]
}