EMV treats every dollar the same regardless of size or context. Expected utility weights outcomes by how much they actually matter to the decision-maker, so a $1M loss that bankrupts a firm carries far more disutility than a $1M gain adds utility. For risk-averse decision-makers or bet-the-company decisions, expected utility (or certainty equivalents) is the more defensible criterion.

A decision tree alternates choice nodes (squares) with chance nodes (circles). At each chance node you compute the EMV of its branches; at each choice node you pick the branch with the highest downstream EMV. Solving back from the leaves to the root (rollback) gives you the optimal policy and the EMV of the whole decision, including the value of staged information.

Frank Knight's distinction still applies: risk is when probabilities are known or estimable from data; uncertainty is when they are not. EMV is built for risk. Under deep uncertainty (novel technology, regulatory shifts, geopolitics) point probabilities are often guesses, so analysts pair EMV with sensitivity analysis, scenario planning, or robust decision methods.

EMV underperforms when outcomes are non-repeatable and one tail is catastrophic (ruin risk), when probabilities are anchored to weak data, when payoffs are non-linear in wealth, or when decision-makers face strict budget or covenant constraints. In those cases, supplement EMV with worst-case analysis, Monte Carlo simulation, or utility-based criteria.

Standard EMV treats a decision as now-or-never. Real options recognize the value of waiting, staging, or abandoning a project as information arrives. A small pilot with the option to scale often has a higher option-adjusted EMV than a full commitment, because flexibility caps the downside while preserving upside.

Common inputs include historical project data, base rates from industry benchmarks, structured expert elicitation (Delphi, calibrated estimation), market-implied probabilities, and reference-class forecasting. Whatever the source, document the basis and stress-test the EMV against reasonable shifts in each probability.

PMI's PMBOK Guide recommends quantifying each identified risk by EMV (probability × impact) and summing the negative EMVs to set a contingency reserve for known risks. Unknown risks are covered by a separate management reserve. This turns subjective risk registers into a defensible dollar number stakeholders can approve.