scenario-planning

name	scenario-planning
description	Scenario planning toolkit — 2x2 matrices, four-case canvas, Monte Carlo setup, sensitivity & tornado, decision-tree EV, war-game, reverse stress test, optionality scoring.
allowed-tools	["Read","Write","Glob","Grep"]

Scenario Planning

Used by cso, chief-risk-officer, cfo, crisis-warroom, capital-allocation. Provides the analytic toolkit for thinking under uncertainty beyond single-point forecasts.

2x2 Scenario Matrix Construction

1. Identify drivers — brainstorm 15–30 forces shaping the future relevant to the decision
2. Cluster & cull — reduce to 8–12 driving forces
3. Score — each on impact (high / low) × uncertainty (high / low)
4. Pick the top 2 — highest impact × highest uncertainty pair
5. Build the 2x2 — four scenarios at the corners
6. Name each scenario — vivid, memorable
7. Build the narrative — what would the world look like in this corner?
8. Identify: common-denominator bets (robust across all 4); scenario-specific options; canary indicators (which scenario is unfolding?)

Four-Case Canvas

For any material decision:

Case	Probability	Description	NPV / outcome	Key drivers
Base	(typically 50–60%)	Most-likely path
Upside	(typically 15–25%)	Drivers break favorably
Downside	(typically 15–25%)	Drivers break unfavorably
Black-swan	(typically 1–5%)	Tail event; structural break

Recommendation lens: would we take this action even in the downside case (or only if upside)? What's our regret if black-swan hits?

Monte Carlo Setup

Step	What to do
1. Identify drivers	5–15 key uncertain inputs (revenue, margin, cost, tax, working capital)
2. Choose distribution per driver	Normal (symmetric), lognormal (asymmetric, multiplicative), triangular (bounded), beta (constrained), empirical (fit history)
3. Specify correlations	Revenue × margin is rarely independent. Use covariance matrix
4. Run iterations	10k+ for stable tails; 50k+ for P1 / P99
5. Report distribution	P5 / P25 / P50 / P75 / P95 — not just mean
6. Tail decomposition	Which driver(s) dominate the bad tail?

Sensitivity & Tornado

• One-at-a-time swings (±10%, ±25%) around base; rank by NPV delta
• Tornado chart — bars sorted by impact magnitude
• Output: top 5 break-the-case variables named in every memo
• Limitation: misses interactions (Monte Carlo complements)

Decision-Tree Expected Value

For sequential decisions with uncertainty:

Decision node [□]
 ├── Action A
 │    Chance node (○)
 │     ├── Outcome A1 (prob × value)
 │     └── Outcome A2 (prob × value)
 │     EV(A) = Σ p · v
 └── Action B
      ...

Choose the action with highest EV — but report variance / downside too.

War-Game / Red-Team Alternate Future

For strategy / competitive decisions:

1. Form Red Team (competitor / disruptor) + Blue Team (us) + Control (referee)
2. Each team plays 2–3 moves into the future, reacting to the other
3. Outcomes scored on shared metrics (market share, margin, customer retention)
4. Debrief identifies: blind spots, our most-vulnerable assumptions, counter-moves
5. Output: revised strategy with named counter-moves and trigger indicators

Reverse Stress Test

Standard approach: scenario → outcome. Reverse: outcome → scenario.

1. Define the failure outcome (e.g., cash exhausted in 6 mo; covenant breach; customer-flight reputation event)
2. Work backwards: what scenario would cause this?
3. Identify the leading indicators that would precede that scenario
4. Stand up monitoring + a pre-defined response playbook

The research doc highlights reverse stress as critical for black-swan preparedness.

Strategic Optionality Scoring

For any bet, compute:

Total value = Static NPV + Option value − Reversibility cost

Option type	Question	Approximate value
Defer	Can we wait for more information?	Black-Scholes-style on volatility
Expand	Can we scale up if it works?	Lattice on success probability
Abandon	Can we exit if it doesn't?	Salvage value relative to commitment
Stage	Can we phase commitment?	Multi-period lattice
Switch	Can we change use?	Spread option
Compound	Does this unlock further options?	Option-on-option

Bets with high option value justify lower static NPV, especially under high uncertainty.

Scenario-Planning Cadence

• Annual — full scenario refresh (cso leads)
• Quarterly — driver-status review; scenario-probability update
• Event-triggered — material change in any high-uncertainty driver triggers re-base

Output

Scenario artifacts saved per the requesting role (typically output/strategy/, output/finance/, output/risk/).