354. Decision Analysis
A framework for making decisions under uncertainty. Distinguishes:
- States of nature: random outcomes outside your control (demand, weather, exchange rates)
- Acts / decisions: things you choose (build big plant? buy insurance? launch product?)
- Payoffs: cost or reward depending on (act, state) combination
- Probabilities: belief about likelihood of each state
Decision analysis asks: given uncertainty, how should I choose?
354.1. Core components
- Decision tree — graphical representation of the decision problem
- EMV — Expected Monetary Value, the standard decision criterion
- EVPI — value of perfectly knowing the state in advance
- EVSI — value of imperfect (sample) information
- Decision criteria — alternatives to EMV under deep uncertainty
- Utility theory — risk-averse decision making
354.2. Steps in a decision analysis
- Frame: identify objectives, alternatives, uncertainties, payoffs
- Structure: build a decision tree (or influence diagram)
- Quantify: assign probabilities and payoff values
- Evaluate: compute expected value of each strategy
- Sensitivity analysis: how does the answer change with input changes?
- Recommend: highest expected value (or by other criterion)
354.3. When to use
- Strategic decisions (capital projects, market entry) with multi-million-dollar payoffs
- Decisions involving uncertain outcomes with quantifiable probabilities
- Sequence decisions (act, observe, decide again) — decision trees are natural
- Comparison of alternatives under uncertainty
354.4. Limits
- Probability estimation is hard — anchoring, overconfidence biases
- Risk aversion not captured by EMV alone — use utility theory
- Deep uncertainty (probabilities themselves unknown) — see non-probabilistic criteria
- Behavioral / strategic considerations (game theory) outside this framework