skills/bigdata-earnings-preview/references/epic-framework.md

# EPIC Framework for Factor Evaluation

## Overview

The EPIC framework is a rigorous filter for evaluating whether a potential investment factor or thesis component provides genuine edge. Before incorporating any insight into an investment decision, it must pass all four EPIC tests. A factor that fails even one test should be discarded or substantially reworked.

**Critical Insight**: A factor must pass ALL four tests to be useful. Partial passes do not create investment edge - they create the illusion of edge, which is more dangerous than acknowledging ignorance.

## Table of Contents
1. [The Four Tests](#the-four-tests)
2. [Application Examples](#application-examples)
3. [Implementation Guidelines](#implementation-guidelines)
4. [Integration with Investment Process](#integration-with-investment-process)

---

## The Four Tests

### E - Effect

**Question**: Does this factor have a material impact on value?

**Pass Criteria:**
- Quantifiable impact on earnings, cash flow, or asset value
- Magnitude sufficient to move intrinsic value by >10%
- Direct causal relationship to business fundamentals
- Not already fully reflected in market price

**Fail Examples:**
- Management changed office locations (immaterial)
- Company won industry award (sentiment, not value)
- Minor product update (insufficient magnitude)
- Well-known structural advantage (already priced)

**Evaluation Questions:**
| Question | Purpose |
|----------|---------|
| How does this factor translate to dollars of value? | Quantify impact |
| What is the magnitude relative to current valuation? | Assess materiality |
| Is this a one-time impact or recurring? | Duration of effect |
| What assumptions must hold for the effect to materialize? | Identify dependencies |

---

### P - Predictability

**Question**: Can you reliably forecast how this factor will evolve?

**Pass Criteria:**
- Historical patterns provide guidance for future behavior
- Underlying drivers are observable and trackable
- Reasonable confidence interval around outcomes
- Feedback mechanisms exist to validate predictions

**Fail Examples:**
- Commodity price speculation (inherently unpredictable)
- Regulatory outcome betting (binary, unknowable)
- Technological disruption timing (high uncertainty)
- Macroeconomic forecasting (poor track record)

**Evaluation Questions:**
| Question | Purpose |
|----------|---------|
| What is your base rate for predictions of this type? | Historical accuracy |
| What would make you change your view? | Falsifiability |
| What is your confidence interval? | Precision of estimate |
| How quickly will you know if you are wrong? | Feedback loop |

---

### I - Independence

**Question**: Is this factor independent of what is already priced into the stock?

**Pass Criteria:**
- Insight derived from proprietary research or analysis
- Not discussed in sell-side research or earnings calls
- Represents new information or novel interpretation
- Not correlated with factors already reflected in price

**Fail Examples:**
- Company has strong brand (everyone knows this)
- Industry is growing (consensus view)
- Management is excellent (reflected in premium multiple)
- Balance sheet is strong (visible to all)

**Evaluation Questions:**
| Question | Purpose |
|----------|---------|
| Where did this insight come from? | Source independence |
| Who else knows this? | Information diffusion |
| Why hasn't the market already priced this? | Edge identification |
| How would a sophisticated investor already model this? | Consensus check |

---

### C - Consensus

**Question**: Does your view differ meaningfully from what is priced into the market?

**Pass Criteria:**
- Explicit variance from sell-side consensus estimates
- Different from implied expectations in reverse DCF
- Contrary to prevailing market narrative
- Supported by evidence consensus has not considered

**Fail Examples:**
- Agreeing earnings will beat by 5% when that is consensus
- Expecting margin expansion when guidance implies it
- Believing in growth story when multiple reflects it
- Seeing risk when high short interest shows others see it too

**Evaluation Questions:**
| Question | Purpose |
|----------|---------|
| What is consensus on this factor? | Map expectations |
| How specifically does your view differ? | Define variance |
| What evidence supports your non-consensus view? | Substantiate thesis |
| Why will consensus eventually agree with you? | Catalyst identification |

---

## Application Examples

### Example 1: Passing All Four Tests

**Factor**: Undiscovered pricing power in niche industrial company

| Test | Assessment | Pass/Fail |
|------|------------|-----------|
| Effect | 300bps margin expansion = 40% earnings upside | Pass |
| Predictability | Customer surveys, contract analysis support thesis | Pass |
| Independence | Primary research not in public domain | Pass |
| Consensus | Street models flat margins; narrative is "commoditized" | Pass |

**Verdict**: Proceed with investment analysis

---

### Example 2: Failing One Test

**Factor**: Market share gains in competitive software market

| Test | Assessment | Pass/Fail |
|------|------------|-----------|
| Effect | 5pts share gain = 25% revenue upside | Pass |
| Predictability | Competitive dynamics uncertain; incumbents may respond | **Fail** |
| Independence | Based on proprietary channel checks | Pass |
| Consensus | Consensus assumes flat share | Pass |

**Verdict**: Insufficient predictability. Acknowledge uncertainty rather than false precision. May warrant smaller position size or options structure.

---

### Example 3: Failing Multiple Tests

**Factor**: Strong management team will create value

| Test | Assessment | Pass/Fail |
|------|------------|-----------|
| Effect | Difficult to quantify specific value impact | **Fail** |
| Predictability | Management quality is somewhat stable | Pass |
| Independence | Well-known; covered extensively by analysts | **Fail** |
| Consensus | Premium multiple already reflects quality | **Fail** |

**Verdict**: Not a source of investment edge. May support holding position but does not justify purchase.

---

## Implementation Guidelines

### Pre-Investment Checklist

Before adding any factor to your thesis:

- [ ] Quantified the dollar impact on intrinsic value (Effect)
- [ ] Documented historical accuracy for similar predictions (Predictability)
- [ ] Identified unique source of insight (Independence)
- [ ] Mapped specifically how view differs from consensus (Consensus)

### Common Traps to Avoid

| Trap | Description | Mitigation |
|------|-------------|------------|
| Complexity disguised as insight | Sophisticated analysis that restates consensus | Ask: "What do I know that others don't?" |
| Narrative seduction | Compelling story without quantifiable impact | Require dollar impact estimate |
| Overconfidence in unpredictable factors | Precision around inherently uncertain outcomes | Scenario analysis with wide ranges |
| Stale variant perception | View that was once non-consensus but is now mainstream | Regularly update consensus mapping |

---

## Integration with Investment Process

EPIC filtering should occur:

1. **Idea generation**: Before deep research, quick EPIC screen to prioritize time
2. **Thesis construction**: Each key thesis element must pass EPIC
3. **Position review**: Periodic reassessment of whether factors still pass
4. **Post-mortem**: Evaluate which EPIC failures led to poor outcomes