Howard’s Value of Information (VoI)

Table of Contents

  1. Definition
  2. How VoI Relates to Marketing
  3. How to Calculate VoI
  4. How to Utilize VoI
  5. Comparison to Similar Approaches
  6. Best Practices
  7. Future Trends
  8. FAQs
  9. Related Terms
  10. Sources

Definition

Howard’s Value of Information (VoI) is a decision-analysis concept formalized by Ronald A. Howard in his 1966 paper Information Value Theory. It puts a dollar figure on how much a decision-maker should be willing to pay for information about an uncertain factor before committing to a choice. Howard’s framing combines probabilistic and economic factors, so numerical values can be assigned to the elimination or reduction of any uncertainty. SciSpace

The concept matters because Shannon’s earlier information theory measured information purely by probability, ignoring consequences. Howard argued that no theory considering only the probabilities of outcomes — without their consequences — could adequately describe the importance of uncertainty to a decision maker. VoI fixes that by tying the worth of information to what a decision-maker stands to gain or lose. Scribd

In marketing, almost every meaningful choice happens under uncertainty. Will a campaign convert? Will a segment respond to a price test? Should the team commission a $40,000 brand study or trust the pilot data already in hand? VoI gives marketers a way to answer the question that sits underneath all of these: is buying more information actually worth it, or would the money be better spent acting now?

Back to Top

How VoI Relates to Marketing

Marketers are professional buyers of information. They commission surveys, A/B tests, focus groups, brand trackers, attribution studies, segmentation work, and concept tests — and every one of those is, in Howard’s terms, an information-gathering activity with a cost and an expected payoff.

VoI reframes research spend as an investment with a ceiling. In risk-sensitive settings the value of information has been defined as the buying price (BPI, the most one should pay) for the revelation, as it provides an upper bound on the value of any information-gathering activity related to an uncertainty of interest, and therefore measures the importance of that uncertainty to the decision problem. If a $200,000 media bet has an EVPI of $8,000, no piece of research about that bet — however clever — is worth more than $8,000. That’s a useful number to have before signing the SOW. INFORMS

The framework also separates two ideas marketers often blur: how uncertain a number is, and how much that uncertainty matters. A wildly uncertain input that doesn’t change the decision has zero VoI. A modestly uncertain input that flips the choice between launch and kill can be worth a great deal.

Back to Top

How to Calculate VoI

The most common form is the Expected Value of Perfect Information (EVPI), which assumes the information would resolve the uncertainty completely.

Formula:

EVPI = EV|PI − EMV

Where:

  • EV|PI = Expected Value given Perfect Information (the average payoff if you always picked the best option for each possible state of the world)
  • EMV = Expected Monetary Value of the best available choice under current uncertainty

A simplified marketing example. A team is deciding whether to launch a new product line. Two possible states: high demand (probability 0.6, payoff $500K) and low demand (probability 0.4, payoff −$200K). The expected value of launching is (0.6 × $500K) + (0.4 × −$200K) = $220K. Not launching pays $0.

With perfect information, the team would launch only when demand is high and skip it when demand is low: EV|PI = (0.6 × $500K) + (0.4 × $0) = $300K.

EVPI = $300K − $220K = $80K. That’s the ceiling on what any market study about demand should cost.

Two properties matter. The value of information can never be less than zero since the decision-maker can always ignore the additional information and make a decision as if such information is not available. No other information gathering/sharing activities can be more valuable than that quantified by EVPI. Wikipedia

Real research rarely gives perfect information, so analysts also calculate Expected Value of Sample Information (EVSI) — the value of imperfect, partial information from a specific study design — and Expected Value of Partial Perfect Information (EVPPI) for resolving just one variable while others remain uncertain. The expected value of partial perfect information (EVPPI) is the expected reduction in loss if the exact value of a particular parameter or parameters were learnt, also interpreted as the amount of decision uncertainty that is due to that parameter. The expected value of sample information (EVSI) is the expected reduction in loss from a study of a specific design. nih

Back to Top

How to Utilize VoI

Common marketing use cases:

Sizing research budgets. Before commissioning a brand health study, segmentation, or pricing research, estimate the EVPI on the decision the study is meant to inform. If the ceiling is $15K and the proposal comes in at $60K, that’s a real conversation to have.

Prioritizing tests. With a roadmap of twenty possible A/B tests, VoI ranks them not by interestingness but by how much the answer would change the next decision. Tests on the highest-traffic pages with the largest decision stakes usually win.

Stage-gating product launches. In product marketing, EVPPI helps decide which uncertainty — price sensitivity, awareness, channel mix — to resolve first. Pick the variable whose resolution would most change the launch plan.

Justifying (or killing) attribution projects. Multi-touch attribution platforms can run six figures annually. VoI asks: how often would better attribution change a budget allocation decision, and by how much? The answer is sometimes lower than expected.

Deciding when to stop researching. A team that has run three rounds of concept testing can use VoI to check whether a fourth round would meaningfully sharpen the decision, or just confirm what’s already known.

Back to Top

Comparison to Similar Approaches

ApproachWhat it measuresWhen marketers use itLimitation
Howard’s VoI (EVPI)Maximum a decision-maker should pay for perfect informationSizing research budgets, prioritizing studiesRequires explicit probabilities and payoffs
ROI on researchRealized return on past research spendPost-hoc justification, vendor evaluationBackward-looking; ignores counterfactuals
Cost-benefit analysisTotal expected benefits vs. total expected costsGo/no-go on large initiativesDoesn’t isolate the value of resolving specific uncertainties
Sensitivity analysisHow much outputs change when inputs changeStress-testing modelsIdentifies which inputs matter, but not what they’re worth
Bayesian updatingRevising probabilities as evidence arrivesSequential testing, iterative campaignsA method, not a valuation; pairs well with VoI
Real options analysisValue of flexibility to defer or change a decisionCapital-intensive launchesMore complex; useful for staged investments
Back to Top

Best Practices

Start with the decision, not the research. If a study won’t change any action under any plausible result, its VoI is zero — regardless of how interesting the findings would be. Run the VoI calculation before drafting the brief, not after.

Be honest about the priors. EVPI math is only as good as the probability estimates fed into it. Marketers tend to anchor on the outcome they expect or hope for. Sanity-check probabilities with someone who has no stake in the project.

Set EVPI as a ceiling, not a target. The actual value of imperfect information (EVSI) is always less than EVPI, often a lot less. A $20K EVPI doesn’t justify a $20K study; it caps it.

Don’t overweight precision. Howard observed that the joint elimination of uncertainty about a number of even independent factors in a problem can have a value that differs from the sum of the values of eliminating the uncertainty in each factor separately. Translation: bundling research questions sometimes pays more than the sum of the parts, and sometimes less. Worth modeling. SciSpace

Document the assumptions. A VoI estimate that can’t be reconstructed six months later is just a number. Keep the probabilities, payoffs, and decision structure in a one-pager attached to the research plan.

Back to Top

Three shifts are pushing VoI from a niche decision-analysis tool into mainstream marketing practice.

First, AI is making the math cheaper. Probabilistic modeling that used to require a decision-science consultant can now be drafted in a notebook in an afternoon. Some marketing-mix modeling platforms have begun exposing EVPI-style outputs alongside their forecasts.

Second, privacy changes have made every piece of customer data more expensive to acquire. As cookies disappear and consent requirements tighten, marketers need a defensible answer to “is this data worth what it costs to collect?” VoI gives them one.

Third, experimentation programs at scale — hundreds or thousands of tests per quarter — have created a real prioritization problem. VoI-based scoring is starting to show up in test backlogs the same way ICE and RICE scoring did a decade ago.

The likely direction: more marketing teams will calculate something like VoI without calling it that, embedding the logic into MMM platforms, experimentation tools, and research procurement workflows.

Back to Top

FAQs

Is VoI only useful for big-budget decisions? No. The math works at any scale. It just tends to be most worth doing when the cost of being wrong is high relative to the cost of the analysis itself.

Can VoI be negative? No. The value of information can never be less than zero since the decision-maker can always ignore the additional information and make a decision as if such information is not available. The information itself can have zero value, but never negative. Wikipedia

What’s the difference between EVPI and EVSI? EVPI assumes information that resolves the uncertainty completely — a theoretical upper bound. EVSI estimates the value of a specific, imperfect study. EVSI is always less than or equal to EVPI for the same uncertainty.

Do I need a decision tree to calculate VoI? Not for simple cases. A spreadsheet with probabilities and payoffs is enough. Decision trees help when there are multiple sequential decisions or several uncertain variables interacting.

Can VoI be applied to qualitative research? Yes, though the calculation is messier. The question is still the same: would the research output change a downstream decision, and by how much? Qualitative inputs often need to be translated into rough probability shifts.

How does VoI handle risk aversion? The basic EVPI formula assumes risk neutrality. For risk-averse decision-makers, the value is measured in expected utility rather than expected dollars. In risk-sensitive settings the value of information has been defined as the expected utility increase of the revelation, or the buying price for the revelation. INFORMS

Is VoI the same as ROI on research? No. ROI is backward-looking and measures realized returns. VoI is forward-looking and measures the maximum justifiable spend before research is done.

What if I don’t know the probabilities? Estimate them. The exercise of being forced to put a number on “how likely is this campaign to work” is often more useful than the resulting VoI calculation. Use ranges and sensitivity analysis if a single point estimate feels too false.

Does VoI work for ongoing measurement (brand trackers, dashboards)? It can, but the calculation is trickier because the “decision” is continuous rather than discrete. One workable approach: identify the specific decisions the tracker is meant to inform, then calculate VoI on those.

Where did the concept originate? Ronald A. Howard’s 1966 paper Information Value Theory paved the way for decision analysis and Information Value Theory, which provides a framework for anyone to calculate the “Value of Perfect Information.” Howard later expanded the framework in The Foundations of Decision Analysis (1968) and the textbook Foundations of Decision Analysis co-authored with Ali Abbas. Andrewclark

Back to Top
Back to Top

Sources

Back to Top

Was this helpful?

Previous

Hoshin Kanri (Policy Deployment)

Next

Jobs-to-be-Done (JTBD)