Description
Calculates how much a piece of research is worth by estimating how likely it is to change a decision and how much that change would gain.
Research has value only if it might change a decision. The expected value of perfect information (EVPI) is the expected gain from knowing the true state before deciding, an upper bound on what any study is worth. The expected value of sample information (EVSI) estimates the gain from a specific study design with its sample size and noise. Health technology assessment uses VOI to prioritise trials. For UX research operations, VOI provides a principled way to prioritise studies, size them and decide when to stop researching and ship, using rough distributions of uncertainty and decision payoffs, as Hubbard advocates.
Attributes
- Discipline
- Game Theory & Decision Science
- Loss type
- Unmigrated
- Goal
- Support a decision, Prioritise
- Stage
- Strategy
- Data
- Analytical / expert
- Effort
- Days
Steps
- Define the decision, alternatives and the outcome metric in value terms (revenue, retention, harm).
- Quantify current uncertainty about key parameters with calibrated estimates (e.g. 90% intervals).
- Model payoffs of each alternative as a function of uncertain parameters.
- Compute EVPI by simulation: expected payoff with perfect information minus expected payoff of the best current choice.
- For candidate studies, compute EVSI given sample size and measurement error, and compare with research cost.
- Prioritise studies with highest net value and identify parameters worth measuring (partial EVPI).
Sources
- Raiffa & Schlaifer, Applied Statistical Decision Theory (1961)
- Howard, Information value theory (1966), IEEE Transactions on Systems Science and Cybernetics
- Hubbard, How to Measure Anything (3rd ed., 2014)
Also known as
- EVI
- EVPI
- EVSI
- Value of information analysis
- VOI
Categories
- Strategy & Collaboration
- Game Theory & Decision Science