When Is It Better to Act With Imperfect Evidence Than Wait for More Research?

Should you act on the best evidence available or wait until you know more?

Decision-makers often face an uncomfortable situation: the evidence is incomplete, but a decision cannot simply disappear while researchers collect more data.

Acting now creates the possibility of choosing incorrectly. Waiting may reduce that risk if useful research becomes available, but waiting has consequences too. Benefits may be postponed, existing harms may continue, resources may remain poorly allocated, and the eventual evidence may not change the decision anyway.

The choice is therefore not between a risky decision now and a perfectly informed decision later. It is between acting under current uncertainty and accepting the costs and consequences of waiting for information that may reduce only part of that uncertainty.

Acting now can be preferable when the expected cost of waiting exceeds the value of learning more first

Waiting for evidence is itself a decision under uncertainty

You do not need certainty before acting

Empirical evidence almost always contains uncertainty. Effect estimates have sampling uncertainty, measurements are imperfect, populations differ, models depend on assumptions, and future conditions may not exactly resemble those studied.

If action required complete certainty, many decisions would never be made. The practical standard is therefore not whether uncertainty has disappeared, but whether the available evidence is sufficient for choosing among the available actions given their expected consequences.

This is why the existence of uncertainty does not automatically mean more research is required before a decision.

Acting now and conducting research are separate decisions

A particularly useful distinction is between the adoption decision and the research decision. You may have enough evidence to prefer an action now while still believing that further research would be valuable for improving future decisions.

These questions can produce different answers. “Act now” does not necessarily mean “stop researching.” Likewise, “further research has value” does not necessarily mean “delay action until the research is complete.”

The cost of waiting can be substantial

Delay matters when the currently preferred action is expected to produce benefits that cannot be recovered later. If an effective intervention is postponed for two years while a study is conducted, the people who would have benefited during those two years may never recover those lost outcomes.

Similar opportunity costs can arise in education, public policy, organizational decisions, environmental management, and other settings. Students may pass through an educational system, infrastructure may deteriorate, preventable costs may accumulate, or an existing inferior practice may continue.

Waiting is therefore not the neutral alternative. It preserves current uncertainty while also preserving, at least temporarily, the consequences of the current course of action.

Ask whether the new evidence could realistically change the decision

The case for waiting is stronger when a feasible study could plausibly reverse the preferred action. Suppose current evidence slightly favors Option A, but credible new findings could readily make Option B preferable. Research may then prevent a consequential mistake.

If the current decision is robust across the plausible range of future evidence, waiting becomes harder to justify solely on decision-value grounds. Research might make estimates more precise without changing what should be done.

This connects directly to whether more precise evidence would have meaningful value if it would not change the decision.

Value of information helps quantify what waiting might gain

Value-of-information analysis provides a formal framework for evaluating the expected benefit of reducing uncertainty. The expected value of sample information (EVSI) estimates how much a particular proposed study is expected to improve decisions by reducing uncertainty.

The research decision can then consider whether that expected improvement justifies the cost of collecting the evidence. The expected net benefit of sampling (ENBS) compares the population EVSI of a study with its expected costs. In formal value-of-information analysis, research costs can include more than the direct study budget. Opportunity costs associated with research may also matter.

Formal value-of-information methods have been developed particularly extensively in health economics, although the underlying reasoning about information, uncertainty, cost, and decision consequences can be applied more broadly.

A valuable study does not automatically justify waiting for it

This is one of the most important distinctions in the problem. Additional research may have positive expected value while immediate adoption is also preferable.

Why? Because the new evidence can improve decisions for future populations even if people facing the decision today should receive the currently preferred option. In some circumstances, action and research can proceed together.

Decision frameworks in health technology assessment have explicitly distinguished options such as approval based on current evidence, approval while further research proceeds, delaying approval until research is conducted, and rejection based on existing evidence. The appropriate choice depends partly on the value of additional evidence and the consequences of adoption or delay.

Reversibility changes the value of waiting

Some decisions can be changed cheaply when new evidence arrives. Others create commitments that are expensive or impossible to reverse.

If a decision is readily reversible, acting now can be more attractive because new information can be incorporated later. A pilot program, adjustable policy, or temporary implementation may allow decision-makers to obtain immediate benefits while retaining flexibility.

By contrast, waiting can become more valuable when acting now creates substantial irrecoverable costs. These might include expensive infrastructure, specialized training, contractual commitments, environmental changes, or other investments that cannot easily be recovered if later evidence reverses the decision.

The relevant issue is not simply whether an action can technically be stopped. It is how much is lost if the initial decision later needs to be reversed.

Some evidence can be collected while acting

The choice is sometimes presented as “act now” versus “research first,” but those are not always the only options.

Decision-makers may be able to implement an action conditionally, collect outcomes during implementation, use staged rollout, conduct a pilot, establish monitoring requirements, or revisit the decision at predetermined points.

Such strategies can be attractive when current evidence favors action but further information remains valuable. Their appropriateness depends on whether implementation permits credible learning, whether the action is reversible, and whether ethical and methodological requirements can be satisfied.

Acting now can destroy the opportunity to learn

There is an important counterargument to immediate adoption. Once an intervention becomes widely available, conducting informative comparative research may become more difficult. Participants or organizations may be unwilling to accept alternatives, existing practices may disappear, or implementation may alter the conditions needed for a rigorous study.

Thus, waiting can sometimes preserve the value of research. This is particularly relevant when the evidence gap is consequential, a high-quality study is feasible now, and widespread adoption would make that study much harder to conduct later.

The decision must therefore consider not only what is lost by waiting, but also what might be lost by acting immediately.

Urgency increases the opportunity cost of delay

When a problem is rapidly causing serious harm, the threshold for acting with imperfect evidence may reasonably differ from a situation in which the decision can wait at little cost.

Urgency does not make weak evidence strong. It changes the consequences of the available choices. A decision-maker may rationally accept greater uncertainty when waiting itself is expected to cause substantial harm.

Conversely, when there is little penalty for waiting and a highly informative study can be completed quickly, gathering evidence first may be preferable.

Irreversible harm requires particular caution

Reversibility applies not only to financial investments but also to outcomes. An action that creates potentially serious and irreversible harm demands a different level of caution from a low-risk decision that can be corrected easily.

Where harms cannot be undone, uncertainty about safety or other consequential outcomes may substantially increase the value of research before widespread action.

The best option may depend on who bears the consequences

Expected-value calculations can obscure distribution. The benefits of acting immediately may accrue to one group while the risks of uncertainty fall on another. Likewise, the burdens of research and delay may not be borne by the people who eventually benefit from better evidence.

Ethical, equity, and distributional considerations should therefore accompany decision analysis where relevant. A favorable average expected outcome does not automatically resolve questions about acceptable risk or fair allocation of benefits and burdens.

A decision can favor both immediate action and continued research

Now change the scenario. Suppose implementation requires a large irreversible infrastructure investment and there is meaningful uncertainty about serious adverse academic effects. Waiting for a highly informative study could then become the better option. The uncertainty has not necessarily increased, but the consequences of acting under it have changed.

Common mistakes when deciding whether to act or wait

Compare the consequences of acting now with the consequences of waiting

When faced with incomplete evidence, identify the currently preferred action first. Then ask whether a realistic study could change that preference and whether waiting for its results would create important costs or forgone benefits.

Also examine reversibility. Acting with uncertainty is easier to justify when the decision can be revised at low cost and evidence can continue to accumulate after implementation.

The central comparison is between the cost of obtaining more evidence and the consequences of remaining uncertain, while also accounting for what happens during the time required to learn more.

Before deciding whether to act now or wait for research

Questions about acting with imperfect evidence

Imperfect evidence can be sufficient when waiting is expected to cost more than learning first is worth

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.