Value of Information

Decision Analysis

Medium
Value of information asks whether learning more before deciding is actually worth the time, money, delay, or effort required. It helps you avoid both rushing blindly and over-researching questions whose answers will not change the action.
Reasoning type
Decision-theoretic
Certainty level
Expected improvement estimate
Cognitive load
Medium
Formality
Medium

Core Idea

Definition

Value of information is the expected decision benefit of obtaining additional information before acting, relative to deciding now without it.

In Plain English

Information is valuable when it changes what you would do, not merely when it is interesting.

Framework Structure

Components

Current Decision
Decision-Relevant Uncertainty
Potential New Information
Expected Improvement vs Cost

Flow

Identify the live decision -> Ask what uncertainty matters -> Estimate whether new information would change the choice -> Compare that benefit with the cost of getting it

How to Apply

  • 1.Define the decision you are trying to improve
  • 2.Identify which uncertainties could actually change the choice
  • 3.Specify what information you could gather to reduce those uncertainties
  • 4.Estimate how much that information would improve the decision quality
  • 5.Compare the benefit with the cost, delay, and opportunity cost of obtaining it

When to Use

  • Choosing between acting now and learning more first
  • Planning experiments, interviews, research, or due diligence
  • High-stakes decisions with reducible uncertainty
  • Situations where waiting may improve choice quality
  • Breaking analysis paralysis with a decision-relevant research lens

When NOT to Use

  • When the decision is reversible and low-stakes
  • When no realistic information could materially change the action
  • When delay costs are higher than any likely learning benefit
  • When teams are using research as a substitute for commitment

Example

Problem

A product team is unsure whether to build a major workflow feature now or first run customer interviews.

Application

  • 1.Define the live decision: build now, delay, or discard
  • 2.Identify the uncertainty that matters most: whether users truly have the workflow pain at meaningful intensity
  • 3.Estimate whether a week of interviews could change the roadmap decision
  • 4.Decide the interviews are worth it because the feature is expensive and the missing knowledge is highly decision-relevant

Conclusion

The team learns before committing because the expected improvement in decision quality exceeds the short delay cost.

Takeaway

The right question is not do we want more information, but will more information change the decision enough to justify its cost.

Common Mistakes

  • Gathering interesting but decision-irrelevant information
  • Failing to state what exact evidence would change the choice
  • Researching too long on reversible decisions
  • Ignoring the opportunity cost of waiting
  • Assuming more data always improves judgment

How to Practice

what would change my mind

Before researching, name the exact evidence that would lead you to choose differently.

decision relevance filter

Screen possible research tasks by whether their answers would change action, timing, or confidence.

learning vs delay check

Compare the likely value of the new information with the cost of waiting to obtain it.

Related Cognitive Biases

analysis paralysis

People keep gathering information even after the expected benefit has largely vanished.

overconfidence

People may act too soon because they assume they already know enough.

curiosity bias

Interesting questions can distract from information that would actually change the choice.

Related Frameworks

Related Skills

option evaluation
hypothesis generation
evaluating credibility
probabilistic reasoning

Variants & Extensions

Expected value of perfect information
Expected value of sample information
Decision-focused research planning
Learning-before-commitment analysis

Typical Failure Modes

  • Research without decision linkage
  • Ignored delay cost
  • Overestimating learning value

Further Reading

  • How to Measure Anything by Douglas W. Hubbard
  • Thinking in Bets by Annie Duke
  • Superforecasting by Philip E. Tetlock and Dan Gardner