All Concepts
Methodology

Evidence Ladder

Every finding in our analysis comes with an evidence grade (E0-E3). This tells you how much confidence you should place in each claim.

Why This Matters
In a world of AI-generated content, knowing the difference between “plausible-sounding” and “actually verified” is critical. The Evidence Ladder forces transparency: every claim must show its receipts.

The Four Tiers

E0Assertion

Claim without specific evidence

Key Test: No source cited
E1Referenced

Claim with source reference but unverifiable

Key Test: Source cited but can't be independently checked
E2Evidenced

Claim supported by specific, verifiable data

Key Test: Third party CAN verify the underlying facts
E3Triangulated

Claim supported by multiple, cross-verified data points

Key Test: Multiple independent sources OR multiple time periods

Expected Distribution

A well-executed analysis has most findings at E2 or E3. Too much E0/E1 suggests shallow research; too little E3 suggests missing the most important conclusions.

E0
E1
E2
E3
< 5%
10-20%
50-70%
20-30%
GradeTargetNotes
E0< 5%Avoid assertions without evidence
E110-20%Context, but shouldn't drive conclusions
E250-70%Majority of substantive findings
E320-30%Ideal for key conclusions

Calibration Examples

These examples show how we grade real-world findings. Use them to calibrate your own judgment.

E0

Management is aggressive with revenue recognition

Evidence: None

Why: No evidence provided

E3

Revenue grew 23% YoY

Evidence: 10-Q, page 8

Why: Quantified from audited filing

E2

CFO resigned unexpectedly

Evidence: 8-K filing

Why: Verifiable primary source

E1

Former employees report quality issues

Evidence: 3 Glassdoor reviews

Why: Referenced but unverifiable

E3

Short interest is 18% of float

Evidence: Bloomberg, dated

Why: Quantified, verifiable

E2

Auditor noted material weakness

Evidence: 10-K Item 9A

Why: Documented primary source

E3

Top 3 customers = 67% of revenue

Evidence: 10-K Risk Factors

Why: Quantified concentration

E3

DSO up 15 days YoY

Evidence: Calculated from 10-Q

Why: Quantified, calculated

E2

May struggle to refinance

Evidence: Credit spreads +450bps

Why: Data verifiable, conclusion probabilistic

E3

Insider selling accelerated: $12M Q4 vs $2M Q3

Evidence: Form 4 filings

Why: Quantified from regulatory filings

Boundary Rules

E1 vs E2 Boundary

Test: Can a third party verify the underlying facts?

  • Verifiable primary source (10-K, 8-K, transcript) → E2
  • Unverifiable source (channel checks, anonymous tips, 'industry experts') → E1

E2 vs E3 Boundary

Test: Multiple independent sources or time periods?

  • Single primary source, single data point → E2
  • Multiple sources confirming same finding, OR pattern across time → E3

Special Cases

Source TypeGradeRationale
Short-seller reportsE1Unless primary evidence independently verified
Rating agency commentaryE1Unless primary data directly cited
News articlesE1Unless quoting primary source directly
Management commentary (in filing)E2It's their statement, verifiable as 'what they said'
Calculated from filing dataE3Quantified, verifiable, reproducible
Channel checksE1Valuable but unverifiable by third party
Forward-looking inference with dataE2Data verifiable even if conclusion probabilistic
Key Principle
Grade based on the factual basis, not the conclusion. A well-reasoned interpretation of verified facts can still be E2 even if the conclusion is probabilistic. An assertion without evidence is E0 regardless of how reasonable it sounds.