gh-lyndonkl-claude/skills/hypotheticals-counterfactuals/resources/methodology.md

# Hypotheticals and Counterfactuals: Advanced Methodology

This document provides advanced techniques for counterfactual reasoning, scenario planning, pre-mortem analysis, and extracting actionable insights from alternative futures.

## Table of Contents
1. [Counterfactual Reasoning](#1-counterfactual-reasoning)
2. [Scenario Planning Techniques](#2-scenario-planning-techniques)
3. [Extracting Insights from Scenarios](#3-extracting-insights-from-scenarios)
4. [Monitoring and Adaptation](#4-monitoring-and-adaptation)
5. [Advanced Topics](#5-advanced-topics)

---

## 1. Counterfactual Reasoning

### What Is Counterfactual Reasoning?

Counterfactual reasoning asks: **"What would have happened if X had been different?"** It's a form of causal inference through imagined alternatives.

**Core principle**: To understand causality, imagine the world with one factor changed and trace the consequences.

**Example**:
- **Actual**: Startup raised $5M Series A → burned through runway in 14 months → failed to reach profitability
- **Counterfactual**: "What if we had raised $3M instead?"
  - Hypothesis: Smaller team (8 vs 15 people), lower burn, forced focus on revenue, reached profitability in 12 months
  - Reasoning: Constraint forces discipline. Without $5M runway, couldn't afford large team. Would prioritize revenue over growth.
  - Insight: Raising more money enabled premature scaling. Constraint would have been beneficial.

### The Minimal Rewrite Principle

When constructing counterfactuals, change **as little as possible** to isolate the causal factor.

**Bad counterfactual** (too many changes):
- "What if we had raised $3M AND competitor had failed AND market had doubled?"
- Problem: Can't tell which factor caused different outcome

**Good counterfactual** (minimal change):
- "What if we had raised $3M (all else equal)?"
- Isolates the funding amount as causal variable

**Technique**: Hold everything constant except the factor you're testing. This reveals whether that specific factor was causal.

### Constructing Plausible Counterfactuals

Not all "what ifs" are useful. Counterfactuals must be **plausible** given what was known/possible at the time.

**Plausible**: "What if we had launched in EU first instead of US?"
- This was a real option available at decision time

**Implausible**: "What if we had magically known the pandemic was coming?"
- Requires impossible foreknowledge

**Test for plausibility**: Could a reasonable decision-maker have chosen this alternative given information available at the time?

### Specifying Causal Mechanisms

Don't just state outcome; explain **HOW** the change leads to different result.

**Weak counterfactual**: "Sales would be 2× higher"
**Strong counterfactual**: "Sales would be 2× higher because lower price ($50 vs $100) → 3× higher conversion rate (15% vs 5%) → more customers despite 50% lower margin per customer → net revenue impact +2×"

**Framework for causal chains**:
1. **Initial change**: What's different? (e.g., "Price is $50 instead of $100")
2. **Immediate effect**: What happens next? (e.g., "Conversion rate increases from 5% to 15%")
3. **Secondary effects**: What follows? (e.g., "Customer volume triples")
4. **Final outcome**: Net result? (e.g., "Revenue doubles despite lower margin")

### Using Counterfactuals for Learning

**Post-decision counterfactual analysis**:

After a decision plays out, ask:
1. **What did we decide?** (Actual decision)
2. **What was the outcome?** (Actual result)
3. **What else could we have done?** (Alternative decision)
4. **What would have happened?** (Counterfactual outcome)
5. **What was the key causal factor?** (Insight for future)

**Example**: Hired candidate A (strong technical, weak communication) → struggled, left after 6 months. Counterfactual: B (moderate technical, strong communication) would have stayed longer, collaborated better. **Insight**: For this role, communication > pure technical skill.

### Avoiding Hindsight Bias

**Hindsight bias**: "I knew it all along" — outcome seems inevitable after the fact.

**Problem**: Makes counterfactual analysis distorted. "Of course it failed, we should have known."

**Mitigation**:
- **Re-inhabit decision context**: What information was available then (not now)?
- **List alternatives considered**: What options were on the table at the time?
- **Acknowledge uncertainty**: How predictable was outcome given information available?

**Technique**: Write counterfactual analysis in past tense but from perspective of decision-maker at the time, without benefit of hindsight.

---

## 2. Scenario Planning Techniques

### Three-Scenario Framework

**Structure**: Optimistic, Baseline, Pessimistic futures

**When to use**: General strategic planning, forecasting, resource allocation

**Process**:

1. **Define time horizon**: 6 months? 1 year? 3 years? 5 years?
   - Shorter horizons: More specific, quantitative
   - Longer horizons: More qualitative, exploratory

2. **Identify key uncertainties**: What 2-3 factors most shape the future?
   - Market adoption rate
   - Competitive intensity
   - Regulatory environment
   - Technology maturity
   - Economic conditions

3. **Develop three scenarios**:
   - **Optimistic** (15-30% probability): Best-case assumptions on key uncertainties
   - **Baseline** (40-60% probability): Expected-case, extrapolating current trends
   - **Pessimistic** (15-30% probability): Worst-case assumptions on key uncertainties

4. **Describe each vividly**: Write 2-4 paragraph narrative making each world feel real

5. **Extract implications**: What should we do in each scenario?

**Example (SaaS startup, 2-year horizon)**:

**Key uncertainties**: (1) Market adoption rate, (2) Competition intensity

**Optimistic scenario** (20% probability): "Market Leader"
- Adoption: 40% market share, viral growth, $10M ARR
- Competition: Weak, no major new entrants
- Drivers: Product-market fit strong, word-of-mouth, early mover advantage
- Implications: Invest heavily in scale infrastructure, expand to adjacent markets

**Baseline scenario** (60% probability): "Steady Climb"
- Adoption: 15% market share, steady growth, $3M ARR
- Competition: Moderate, 2-3 well-funded competitors
- Drivers: Expected adoption curve, competitive but differentiated
- Implications: Focus on core product, maintain burn discipline, build moat

**Pessimistic scenario** (20% probability): "Survival Mode"
- Adoption: 5% market share, slow growth, $500k ARR
- Competition: Intense, major player launches competing product
- Drivers: Slow adoption, strong competition, pivot needed
- Implications: Cut burn, extend runway, explore pivot or acquisition

### 2×2 Scenario Matrix

**Structure**: Two key uncertainties create four quadrants (scenarios)

**When to use**: When two specific uncertainties dominate strategic decision

**Process**:

1. **Identify two critical uncertainties**: Factors that:
   - Are genuinely uncertain (not predictable)
   - Have major impact on outcomes
   - Are independent (not correlated)

2. **Define axes extremes**:
   - Uncertainty 1: [Low extreme] ←→ [High extreme]
   - Uncertainty 2: [Low extreme] ←→ [High extreme]

3. **Name four quadrants**: Give each world a memorable name

4. **Develop narratives**: Describe what each world looks like

5. **Identify strategic implications**: What works in each quadrant?

**Example (Market entry decision)**:

**Uncertainty 1**: Regulatory environment (Strict ←→ Loose)
**Uncertainty 2**: Market adoption rate (Slow ←→ Fast)

|  | **Slow Adoption** | **Fast Adoption** |
|---|---|---|
| **Strict Regulation** | **"Constrained Growth"**: Premium focus, compliance differentiator | **"Regulated Scale"**: Invest in compliance infrastructure early |
| **Loose Regulation** | **"Patient Build"**: Bootstrap, iterate slowly | **"Wild West Growth"**: Fast growth, grab market share |

**Strategic insight**: Common actions (build product), hedges (low burn), options (compliance prep), monitoring (regulation + adoption)

### Cone of Uncertainty

**Structure**: Range of outcomes widens over time

**When to use**: Long-term planning (5-10+ years), high uncertainty domains (technology, policy)

**Visualization**:
```
Present → 1 year: Narrow cone (±20%)
        → 3 years: Medium cone (±50%)
        → 5 years: Wide cone (±100%)
        → 10 years: Very wide cone (±200%+)
```

**Technique**:
1. **Start with trend**: Current trajectory (e.g., "10% annual growth")
2. **Add uncertainty bands**: Upper and lower bounds that widen over time
3. **Identify branch points**: Key decisions or events that shift trajectory
4. **Track leading indicators**: Signals that show which path we're on

**Example (Revenue forecasting)**:
- Year 1: $1M ± 20% ($800k - $1.2M) — narrow range, short-term visibility
- Year 3: $3M ± 50% ($1.5M - $4.5M) — wider range, product-market fit uncertain
- Year 5: $10M ± 100% ($5M - $20M) — very wide range, market evolution unknown

### Pre-Mortem Process (Prospective Hindsight)

**What is it?**: Imagine future failure, work backward to identify causes

**Why it works**: "Prospective hindsight" — imagining outcome has occurred unlocks insights impossible from forward planning

**Research foundation**: Gary Klein, "Performing a Project Premortem" (HBR 2007)

**6-Step Process**:

**Step 1: Set the scene**
- Future date: "It is [6 months / 1 year / 2 years] from now..."
- Assumed outcome: "...and the [project has failed completely / decision was disastrous]."
- Make it vivid: "The product has been shut down. The team disbanded. We lost $X."

**Step 2: Individual brainstorm (5-10 minutes, silent)**
- Each person writes 3-5 reasons WHY it failed
- Silent writing prevents groupthink
- Encourage wild ideas, non-obvious causes

**Step 3: Round-robin sharing**
- Each person shares one reason (rotating until all shared)
- No debate yet, just capture ideas
- Scribe records all items

**Step 4: Consolidate and cluster**
- Group similar causes together
- Look for themes (technical, market, team, execution, external)

**Step 5: Vote on top risks**
- Dot voting: Each person gets 3-5 votes
- Distribute votes across risks
- Identify top 5-7 risks by vote count

**Step 6: Develop mitigations**
- For each top risk, assign:
  - **Mitigation action**: Specific step to prevent or reduce risk
  - **Owner**: Who is responsible
  - **Deadline**: When mitigation must be in place
  - **Success metric**: How to know mitigation worked

**Pre-mortem psychology**:
- **Permission to dissent**: Failure assumption gives license to voice concerns without seeming negative
- **Cognitive relief**: Easier to imagine specific failure than abstract "what could go wrong?"
- **Team alignment**: Surfaces hidden concerns before they become real problems

---

## 3. Extracting Insights from Scenarios

### Moving from Stories to Actions

Scenarios are useless without actionable implications. After developing scenarios, ask:

**Core questions**:
1. **What should we do regardless of which scenario unfolds?** (Common actions)
2. **What hedges should we take against downside scenarios?** (Risk mitigation)
3. **What options should we create for upside scenarios?** (Opportunity capture)
4. **What should we monitor to track which scenario is unfolding?** (Leading indicators)

### Identifying Common Actions

**Common actions** ("no-regrets moves"): Work across all scenarios

**Technique**: List actions that make sense in optimistic, baseline, AND pessimistic scenarios

**Example (Product launch scenarios)**:

| Scenario | Build Core Product | Hire Marketing | Raise Series B |
|----------|-------------------|----------------|----------------|
| Optimistic | ✓ Essential | ✓ Essential | ✓ Essential |
| Baseline | ✓ Essential | ✓ Essential | △ Maybe |
| Pessimistic | ✓ Essential | △ Maybe | ✗ Too risky |

**Common actions**: Build core product, hire marketing (work in all scenarios)
**Not common**: Raise Series B (only makes sense in optimistic/baseline)

### Designing Hedges

**Hedges**: Actions that reduce downside risk if pessimistic scenario unfolds

**Principle**: Pay small cost now to protect against large cost later

**Examples**:
- **Pessimistic scenario**: "Competitor launches free version, our revenue drops 50%"
- **Hedge**: Keep burn low, maintain 18-month runway (vs. 12-month)
  - Cost: Hire 2 fewer people now
  - Benefit: Survive revenue shock if it happens

- **Pessimistic scenario**: "Regulatory crackdown makes our business model illegal"
- **Hedge**: Develop alternative revenue model in parallel
  - Cost: 10% of eng time on alternative
  - Benefit: Can pivot quickly if regulation hits

**Hedge evaluation**: Compare cost of hedge vs. expected loss × probability
- Hedge cost: $X
- Loss if scenario occurs: $Y
- Probability of scenario: P
- Expected value of hedge: (P × $Y) - $X

### Creating Options

**Options**: Prepare to capture upside if optimistic scenario unfolds, without committing resources now

**Real options theory**: Create flexibility to make future decisions when more information available

**Examples**:
- **Optimistic scenario**: "Adoption faster than expected, enterprise demand emerges"
- **Option**: Design product architecture with enterprise features in mind (multi-tenancy, SSO hooks), but don't build until demand confirmed
  - Low cost now: Design decisions
  - High value later: Fast enterprise launch if demand materializes

- **Optimistic scenario**: "International markets grow 3× faster than expected"
- **Option**: Hire one person with international experience, build relationships with international partners
  - Low cost now: One hire
  - High value later: Quick international expansion if opportunity emerges

### Defining Leading Indicators

**Leading indicators**: Early signals that show which scenario is unfolding

**Characteristics of good leading indicators**:
- **Observable**: Can be measured objectively
- **Early**: Visible before scenario fully plays out (6+ months advance notice)
- **Actionable**: If indicator triggers, we know what to do

**Example (Market adoption scenarios)**:

| Scenario | Leading Indicator | Threshold | Action if Triggered |
|----------|------------------|-----------|---------------------|
| Optimistic | Monthly adoption rate | >20% MoM for 3 months | Accelerate hiring, raise capital |
| Baseline | Monthly adoption rate | 10-20% MoM | Maintain plan |
| Pessimistic | Monthly adoption rate | <10% MoM for 3 months | Cut burn, explore pivot |

**Monitoring cadence**: Review indicators monthly or quarterly, update scenario probabilities based on new data

### Decision Points and Trigger Actions

**Decision points**: Pre-defined thresholds that trigger specific actions

**Format**: "If [indicator] crosses [threshold], then [action]"

**Examples**:
- "If monthly churn rate >8% for 2 consecutive months, then launch retention task force"
- "If competitor raises >$50M, then accelerate roadmap and increase marketing spend"
- "If regulation bill passes committee vote, then begin compliance implementation immediately"

**Benefits**:
- **Pre-commitment**: Decide now what to do later, avoids decision paralysis in moment
- **Speed**: Trigger action immediately when condition met
- **Alignment**: Team knows what to expect, can prepare

---

## 4. Monitoring and Adaptation

### Tracking Which Scenario Is Unfolding

**Reality ≠ any single scenario**: Real world is blend of scenarios, or something unexpected

**Monitoring approach**:
1. **Quarterly scenario review**: Update probabilities based on new evidence
2. **Indicator dashboard**: Track 5-10 leading indicators, visualize trends
3. **Surprise tracking**: Log unexpected events not captured by scenarios

**Example dashboard**:

| Indicator | Optimistic | Baseline | Pessimistic | Current | Trend |
|-----------|-----------|----------|-------------|---------|-------|
| Adoption rate | >20% MoM | 10-20% | <10% | 15% | ↑ |
| Churn rate | <3%/mo | 3-5% | >5% | 4% | → |
| Competitor funding | <$20M | $20-50M | >$50M | $30M | ↑ |
| NPS | >50 | 30-50 | <30 | 45 | ↑ |

**Interpretation**: Trending optimistic (adoption, NPS), watch competitor funding

### Updating Scenarios

**When to update**:
- **Scheduled**: Quarterly reviews
- **Triggered**: Major unexpected event (pandemic, regulation, acquisition, etc.)

**Update process**:
1. **Review what happened**: What changed since last review?
2. **Update probabilities**: Which scenario looking more/less likely?
3. **Revise scenarios**: Do scenarios still capture range of plausible futures? Add new ones if needed
4. **Adjust actions**: Change hedges, options, or common actions based on new information

**Example**: Before pandemic: Optimistic 20%, Baseline 60%, Pessimistic 20%. After: Add "Remote-first world" (30%), reduce Baseline to 40%. Action: shift from office expansion to remote tooling.

### Dealing with Surprises

**Black swans**: High-impact, low-probability events not captured by scenarios (Taleb)

**Response protocol**:
1. **Acknowledge**: "This is outside our scenarios"
2. **Assess**: How does this change the landscape?
3. **Create emergency scenario**: Rapid scenario development (hours/days, not weeks)
4. **Decide**: What immediate actions needed?
5. **Update scenarios**: Incorporate new uncertainty into ongoing planning

**Example**: COVID-19 lockdowns (not in scenarios) → Assess: dining impossible → Emergency scenario: "Delivery-only world" (6-12 mo) → Actions: pivot to takeout, renegotiate leases → Update: add "Hybrid dining" scenario

### Scenario Planning as Organizational Learning

**Scenarios as shared language**: Team uses scenario names to communicate quickly
- "We're in Constrained Growth mode" → Everyone knows what that means

**Scenario-based planning**: Budgets, roadmaps reference scenarios
- "If we hit Optimistic scenario by Q3, we trigger hiring plan B"

**Cultural benefit**: Reduces certainty bias, maintains flexibility, normalizes uncertainty

---

## 5. Advanced Topics

### Counterfactual Probability Estimation

**Challenge**: How likely was counterfactual outcome?

**Approach**: Use base rates and analogies
1. **Find analogous cases**: What happened in similar situations?
2. **Calculate base rate**: Of N analogous cases, in how many did X occur?
3. **Adjust for specifics**: Is our case different? How?
4. **Estimate probability range**: Not point estimate, but range (40-60%)

**Example**: "What if we had launched in EU first?" — 20 similar startups: 8/20 chose EU-first (3/8 succeeded = 37.5%), 12/20 chose US-first (7/12 = 58%). Our product has EU features (+10%) → EU-first 35-50%, US-first 50-65%. **Insight**: US-first was better bet.

### Scenario Narrative Techniques

**Make scenarios memorable and vivid**:

**Technique 1: Use present tense**
- Bad: "Adoption will grow quickly"
- Good: "It's January 2026. Our user base has grown 10× in 12 months..."

**Technique 2: Add concrete details**
- Bad: "Competition is intense"
- Good: "Three well-funded competitors (FundedCo with $50M Series B, StartupX acquired by BigTech, OpenSource Project with 10k stars) are fighting for same customers..."

**Technique 3: Use personas/characters**
- "Sarah, our typical customer (marketing manager at 50-person B2B SaaS company), now has five alternatives to our product..."

**Technique 4: Include metrics**
- "Monthly churn rate: 8%, NPS: 25, CAC: $500 (up from $200)"

### Assumption Reversal for Innovation

**Technique**: Take core assumption, flip it, explore implications

**Process**:
1. **List key assumptions**: What do we take for granted?
2. **Reverse each**: "What if opposite is true?"
3. **Explore plausibility**: Could reversal be true?
4. **Identify implications**: What would we do differently?
5. **Test**: Can we experiment with reversal?

**Examples**:

| Current Assumption | Reversed Assumption | Implications | Action |
|-------------------|---------------------|--------------|--------|
| "Customers want more features" | "Customers want FEWER features" | Simplify product, remove rarely-used features, focus on core workflow | Survey: Would users pay same for product with 50% fewer features but better UX? |
| "Freemium is best model" | "Paid-only from day 1" | No free tier, premium positioning, higher LTV but lower top-of-funnel | Test: Launch premium SKU, measure willingness to pay |
| "We need to raise VC funding" | "Bootstrap and self-fund" | Slower growth, but control + profitability focus | Calculate: Can we reach profitability on current runway? |

### Timeboxing Scenario Work

**Problem**: Scenario planning can become endless theorizing

**Solution**: Timebox exercises

**Suggested time budgets**: Pre-mortem (60-90 min), Three scenarios (2-4 hrs), 2×2 matrix (3-5 hrs), Quarterly review (1-2 hrs)

**Principle**: Scenarios are decision tools, not academic exercises. Generate enough insight to decide, then act.

---

## Summary

**Counterfactual reasoning** reveals causality through minimal-change thought experiments. Focus on plausibility, specify mechanisms, avoid hindsight bias.

**Scenario planning** (three scenarios, 2×2 matrix, cone of uncertainty) explores alternative futures. Assign probabilities, make vivid, extract actions.

**Extract insights** by identifying common actions (no-regrets moves), hedges (downside protection), options (upside preparation), and leading indicators (early signals).

**Monitor and adapt** quarterly. Track indicators, update scenario probabilities, adjust strategy as reality unfolds. Treat surprises as learning opportunities.

**Advanced techniques** include counterfactual probability estimation, narrative crafting, assumption reversal, and rigorous timeboxing to avoid analysis paralysis.

**The goal**: Prepare for uncertainty, maintain strategic flexibility, and make better decisions by systematically exploring "what if?"