When Is More Research Actually Needed Before a Decision Can Be Made?

How do you know when the evidence is not yet sufficient to decide?

Researchers frequently conclude that “more research is needed.” Sometimes that conclusion is well justified. Existing studies may leave unresolved uncertainty that could materially affect which action should be taken.

But uncertainty is present in almost every evidence base. If a decision had to wait until uncertainty disappeared, many decisions would never be made. The more useful question is therefore not whether uncertainty remains, but whether reducing that uncertainty is important enough to justify additional research before acting.

That distinction matters because research itself takes time and resources. Waiting for better evidence can also have consequences.

More research is needed when remaining uncertainty could change what should be done

The need for more research depends on decision-relevant uncertainty

A decision does not require certainty

Research evidence rarely produces complete certainty. Sampling variation, measurement limitations, model assumptions, differences among populations, and incomplete knowledge about relevant outcomes can all leave uncertainty after substantial research has already been conducted.

The existence of uncertainty therefore cannot serve as the stopping rule for decision-making. Instead, a decision-maker generally has to choose the best available option given current information while considering the possibility that additional evidence could alter that choice.

This distinction is central to value-of-information reasoning. The question is not simply, “How uncertain are we?” It is, “What are the consequences of this uncertainty, and could obtaining more information improve the decision?”

The important uncertainty is uncertainty about the decision

A parameter, effect estimate, mechanism, or outcome can be uncertain without creating much uncertainty about what should be done.

Suppose two interventions are being compared. Researchers remain unsure about the exact magnitude of the benefit of Intervention A, but across nearly all plausible values A remains preferable to B. More precise evidence might improve scientific knowledge without changing the decision.

Now suppose relatively modest changes in the estimated effect would reverse the preferred option. The same amount of statistical uncertainty could then have much greater practical importance because the evidence leaves genuine uncertainty about the decision itself.

This is why the size of an uncertainty should be interpreted in relation to the decision it could affect, rather than treated as an independent reason to conduct another study.

Ask whether plausible new evidence could change the preferred action

One practical test is to consider the range of findings that a credible new study might produce. Would some plausible results favor one action while others favor another?

If so, additional evidence may have substantial decision value. If the same action remains preferable across the plausible range, the argument for postponing a decision becomes weaker, even though additional research could still have other scientific benefits.

This also explains why greater precision can have limited decision value when it would not change what should be done.

The consequences of making the wrong decision matter

The possibility that new evidence could change a decision is only part of the problem. The consequences of choosing incorrectly also matter.

A decision affecting a large population, substantial resources, serious harms, or an intervention that is difficult to reverse may warrant a stronger evidential basis. By contrast, it may be reasonable to act with greater uncertainty when the decision has limited consequences, can easily be reversed, or can be revised as evidence accumulates.

Thus, a small probability of choosing incorrectly can still justify research when the consequences are large. A much larger probability of error may be tolerable when little is at stake.

Research must be capable of reducing the uncertainty that matters

Even consequential decision uncertainty does not automatically justify another study. The proposed research must be able to reduce the relevant uncertainty.

For example, suppose a policy decision depends primarily on uncertainty about long-term outcomes. A short study measuring only immediate satisfaction may be rigorous and publishable, but it would do little to resolve the uncertainty driving the decision.

This is an important reason to identify the specific uncertainty worth reducing before selecting a study design. Research should be aligned with the information the decision actually requires.

Perfect information is different from information a real study can provide

Formal value-of-information analysis makes a useful distinction here. The expected value of perfect information (EVPI) represents the expected value of eliminating all uncertainty relevant to a decision. It provides an upper bound on the potential value of resolving that uncertainty.

A real study will usually eliminate only part of it. The expected value of sample information (EVSI) instead estimates the expected benefit associated with the reduction in uncertainty that could result from a particular study design and sample size.

This distinction matters for research decisions. A problem may contain substantial valuable uncertainty while a proposed study has little value because it would resolve too little of that uncertainty.

The expected benefit of research should be compared with its cost

If additional evidence could improve a decision, the next question is whether obtaining it is worthwhile.

Formal value-of-information approaches can compare the expected value of information from a proposed study with the expected cost of obtaining it. One relevant measure is the expected net benefit of sampling (ENBS), which considers the value of the information generated by a particular study relative to its expected costs.

Outside settings where formal modelling is feasible, the same logic can still guide qualitative reasoning: what could be gained from knowing more, what would it cost to learn it, and is the expected improvement sufficiently important?

Waiting for research also has consequences

“Research first, decide later” can sound cautious, but postponement is itself a choice. People may continue receiving an inferior intervention, an effective policy may remain unavailable, resources may continue to be allocated inefficiently, or an avoidable problem may persist while evidence is collected.

The appropriate comparison is therefore not between acting under uncertainty and a costless future state of perfect knowledge. It is between realistic alternatives, including acting now, gathering information while acting, or delaying a decision while conducting research.

In some situations, it may be preferable to act with imperfect evidence rather than wait for additional research.

Sometimes research and action can occur together

The choice is not always binary. Some decisions can be implemented provisionally while additional evidence is collected. Policies can sometimes be piloted, interventions can be monitored, and decisions can be revisited as new information becomes available.

Whether such an approach is appropriate depends on reversibility, ethical constraints, implementation costs, the feasibility of learning after adoption, and the consequences of exposing people to an option that may later prove inferior.

Should a university wait for another study before adopting a learning intervention?

Now change one fact. Suppose every plausible estimate from the existing evidence still makes the program preferable to the available alternative. Another study could improve precision, but it would be unlikely to alter the adoption decision. The argument that the university must wait for more evidence would then be considerably weaker.

Why “more research is needed” can be an inadequate conclusion

Replace the generic call for more research with a decision argument

When you conclude that additional research is needed, specify why. Identify the unresolved uncertainty, explain how it affects the decision, describe what evidence would reduce it, and consider whether obtaining that evidence is worth its costs.

This produces a much stronger justification than simply observing that the literature remains incomplete.

Ultimately, the issue is not whether researchers could learn more. They almost always could. The more discriminating question is whether the information gained from answering the unresolved question would be valuable enough to justify postponing or modifying the decision.

Before recommending more research before a decision

Questions about deciding whether more research is needed

Research is needed when learning more is worth more than deciding now

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.