Practical framework / 5 MIN READ
Ask whether better information can change a valuable decision
Compare the expected benefit of learning with the cost and delay of the study.
AI-assisted research and analysis for the SNS Creative Fieldbook
Sources checked 2026-11-06
More precise measurement is useful when it can improve a consequential choice. A value-of-information review focuses the research budget on uncertainty that could change action, while recognizing that a real study will provide imperfect evidence rather than certainty.
What you will leave with: Build a study-value screen with decision alternatives, uncertain states, research costs and a deadline.
Use this when
- A proposed measurement study costs nearly as much as the decision it informs.
- The team keeps collecting data without saying what result would change the plan.
This is a decision-analysis illustration, not a universal research valuation formula or authority to commission a study.
Define the choice and the uncertain facts
List the feasible actions and the outcomes that would make one preferable. Separate uncertainty about customer response from uncertainty about costs, execution or the model itself. State when the decision must be made and how long its consequences last.
Identify what would actually change after each plausible study result. If every result leads to the same action, this study has no decision-improvement value for that specified choice, although it may have another justified purpose. Do not invent a future use merely to defend an already preferred research purchase.
Use perfect information as a screening bound
The National Academies’ decision guidance distinguishes perfect from sample information and emphasizes research and delay costs. Perfect information provides a conceptual upper bound on the value of resolving the modeled uncertainty; a practical study normally resolves less and can still mislead.
For a simple monetary decision, an analyst can compare the best expected outcome using current information with the expected outcome from choosing the best action in each known state. Make probabilities and payoffs explicit and test their sensitivity. Do not present subjective inputs as observed facts or apply a risk-neutral calculation to every decision without discussion.
Assess what the proposed study can really learn
Map the study’s design to the uncertain input. A preference survey may not resolve actual purchase response, and an underpowered experiment may not distinguish the actions that matter. Consider outcome delay, implementation risk, bias and whether the evidence will transfer to the decision population.
Estimate the value of the imperfect information with qualified support where the stakes justify formal analysis. Compare it with study cost, staff effort, delay and the cost of changing course. A study priced below the perfect-information bound is not automatically worthwhile; its actual information may be much less valuable.
Choose a proportionate learning plan
Consider a smaller pilot, existing evidence, a reversible staged action or no new study. Explain which uncertainty each option reduces and what remains. Preserve customer protections and ethical research requirements rather than treating every possible measurement as available for purchase.
Obtain approval for the selected research scope and spending. Set a decision date and a rule for acting on inconclusive evidence. Afterward, assess whether the study informed the choice as intended, without pretending that a favorable business outcome proves the research paid for itself or that an unfavorable result made the information useless.
Make the reasoning concrete.
A hypothetical safe option yields 10,000 currency units in either state. A risky option yields 30,000 in a favorable state with assumed probability 0.4 and minus 10,000 otherwise. Current expected value favors the safe option at 10,000; perfect information would yield 18,000, so its modeled value is 8,000 before research and delay costs.
Buy any study below the bound
Opening: Approve a 7,000 study because it costs less than 8,000.
Hypothesis: Perfect information could be valuable.
Risk: The proposed study may provide much less than perfect information.
Review the actual learning
Opening: Assess whether the study can change the decision reliably and in time.
Hypothesis: Cost is compared with realistic decision improvement.
Risk: The probabilities, payoffs and study accuracy remain uncertain inputs.
Use 8,000 only as the illustrative upper bound under the stated model. A 9,000 study cannot be justified by this decision-improvement value alone; a cheaper study still needs an imperfect-information assessment.
Take the questions into your next brief.
Use these prompts in your own notes. Nothing is submitted from this page.
Decision
Which feasible action could change?
Uncertainty
What fact matters enough to alter the choice?
Bound
What is perfect information worth under explicit assumptions?
Study
How much can this design actually resolve?
Cost and timing
What research, delay and switching costs apply?
Watch for the wrong conclusion.
- Valuing precision without a decision consequence.
- Treating perfect information as what a survey will deliver.
- Ignoring the cost of waiting.
- Calling hypothetical probabilities measured evidence.
Your next action
Write the action that would follow each plausible study result before approving another measurement project.
Sources & scope
Frameworks, worksheets and hypothetical examples are original editorial synthesis. Linked primary sources support only the stated concepts. Source checking is not an independent audit of effectiveness. No named human author or reviewer is claimed.
- Institute of Medicine: Environmental Decisions in the Face of Uncertainty, Chapter 5 →Primary expert-committee methodological report · Supports: Explains expected value of perfect and sample information and the need to consider research and delay costs. The marketing payoff example is original and depends entirely on its hypothetical probabilities and values.