What Should You Do When the Literature Suggests Your Planned Intervention Is Unlikely to Work?

What if previous research gives you little reason to expect your intervention to succeed?

You have identified a problem and perhaps already designed an intervention to address it. Then the literature review becomes uncomfortable. Similar interventions have produced small, inconsistent, or negligible effects. Some have failed repeatedly.

The tempting response is to emphasize the few favorable studies, argue that your context is different, and continue with the original plan. A more useful response is to ask why the intervention should work at all, what previous evaluations actually tested, and whether your version changes anything consequential.

Intervention research should not begin with the assumption that an intervention deserves to succeed. The literature may be telling you to refine it, test a different uncertainty first, or stop investing in the original approach.

Diagnose why the intervention is unlikely to work before testing it again

An intervention can fail for reasons that require very different responses

First, determine what “unlikely to work” actually means

Do not reduce the literature to a count of successful and unsuccessful studies. Examine what previous evaluations estimated, how precise those estimates were, the certainty and directness of the evidence, and whether the interventions were sufficiently similar to yours.

Evidence synthesis frameworks such as GRADE consider factors including risk of bias, inconsistency, indirectness, imprecision, and publication bias when judging confidence in a body of evidence. An intervention supported only by uncertain evidence is different from an intervention for which strong evidence consistently indicates negligible benefit.

If the strongest relevant evidence precisely estimates effects too small to matter, your planned study needs a stronger justification than “this has not been tested in my institution.” If previous evidence is weak or highly indirect, considerably more uncertainty may remain.

Distinguish intervention failure from implementation failure

An intervention may fail because its underlying mechanism is ineffective. It may also fail because participants never received enough of it, implementers deviated from the intended procedure, uptake was poor, necessary resources were unavailable, or the surrounding environment prevented the intended mechanism from operating.

These possibilities are not interchangeable.

Process evaluation can help examine implementation, mechanisms, and context rather than relying on the outcome estimate alone. Guidance developed for complex interventions specifically recognizes that understanding what was delivered and how, how change was expected to occur, and how context influenced implementation and outcomes can help explain apparently unsuccessful interventions.

Examine the intervention's programme theory

An intervention should have a plausible account of how its activities are expected to produce the intended outcomes. The updated Medical Research Council framework describes programme theory as an account of how an intervention is expected to generate effects and under what conditions. That theory should be tested and refined throughout intervention development and evaluation.

Trace the causal sequence. What does the intervention change first? How is that change supposed to affect the next step? Which assumptions must hold before the final outcome changes?

For example, providing students with automated feedback cannot improve revision quality through that feedback if students do not read it, cannot interpret it, distrust it, or have no meaningful opportunity to revise their work. A weak link earlier in the causal chain can prevent the intended final outcome even when the intervention itself appears sensible.

Ask whether your intervention meaningfully differs from previous unsuccessful versions

Researchers frequently respond to unfavorable literature by arguing that their intervention is “different.” The relevant question is whether the difference should plausibly alter the outcome.

Changing the interface, duration, institution, or delivery platform may matter, but not automatically. Identify the mechanism through which the modification should change implementation or effectiveness.

If previous interventions failed because participants rarely engaged with them, for instance, a redesigned intervention that directly addresses engagement may justify further evaluation. Simply delivering essentially the same intervention to another convenient sample provides a weaker reason for expecting a different result.

Context can alter intervention effects, but context is not a universal escape clause

Interventions interact with organizational, social, cultural, economic, physical, and other contextual conditions. Contemporary intervention frameworks therefore treat context as part of understanding how an intervention generates outcomes rather than merely as background information.

An intervention that works in one setting may perform differently elsewhere. Likewise, an intervention that previously failed could conceivably work under conditions better suited to its mechanism.

However, claiming that “our context is different” is not enough. Identify which contextual feature matters, how it interacts with the intervention, and why it changes the expected causal process. Otherwise, context becomes an unfalsifiable explanation for repeating an intervention despite unfavorable evidence.

Consider whether the intervention needs development rather than another effectiveness test

Sometimes the literature does not support moving directly to a full evaluation. The intervention may need further development, feasibility work, refinement, or assessment of key uncertainties first.

The Medical Research Council framework does not treat intervention research as an automatic march from development to evaluation to implementation. Depending on what is learned, researchers may proceed, repeat a phase, return to an earlier phase, or stop. Feasibility work can therefore be used to determine whether an intervention and its evaluation are workable before committing resources to a definitive test.

An unsuccessful intervention may expose a mistaken assumption about the problem

Sometimes the intervention is not the deepest problem. Its failure may reveal that the assumed cause of the research problem is wrong.

If an intervention designed to improve outcome Y by changing X repeatedly changes X but leaves Y largely unaffected, the literature may be challenging the presumed causal importance of X. At that point, continuing to optimize the intervention may be less informative than reconsidering the causal model behind it.

This connects intervention evaluation to the broader problem of what to do when the literature contradicts your research assumptions.

Sometimes the intervention should be abandoned

Not every intervention deserves another trial. If credible, directly applicable evidence repeatedly shows negligible benefits, important harms, unacceptable burdens, poor feasibility, or a failed mechanism, continuing simply because the intervention is available or attractive may have little scientific value.

The relevant question is not whether you can construct another study around it. It is whether another study can resolve an important uncertainty. If the evidence has already reduced that uncertainty sufficiently, resources may be better directed elsewhere.

When another effectiveness study is not the logical next step

How weak interventions survive unfavorable evidence

Choose the next study according to the uncertainty that remains

Do not ask only whether you can test the intervention. Ask what researchers and decision makers still need to know about it. Contemporary guidance for complex intervention research similarly emphasizes identifying key uncertainties and choosing research questions according to what further evidence would contribute.

If the intervention literature also undermines the effect you originally predicted, reconsider whether the expected effect remains a defensible basis for the study. A substantial change to the intervention may also require you to revise the research question rather than keeping the original question attached to a fundamentally different study.

Before testing an intervention that previous research makes doubtful, check:

Questions about interventions that previous research does not strongly support

Do not run another intervention study merely because you already designed the intervention

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.