01 · The Question
What if the effect you find is smaller than your study was built to detect?
You planned the study carefully, collected the data, and estimated an effect. The problem is that the effect appears smaller than the magnitude your design had a reasonable chance of detecting reliably. What can you conclude?
This situation is easy to mishandle. A researcher may see a non-significant result and conclude that nothing happened. Another may notice a small point estimate and conclude that a small effect definitely exists. Neither conclusion necessarily follows from the data.
The central issue is not simply whether the observed effect falls below a particular numerical threshold. It is whether the study provides enough information to distinguish effects that matter from effects that do not.
03 · What You Need to Know
Detection limits are about what the design can reliably distinguish
What does "detect reliably" actually mean?
During study planning, researchers often determine a sample size by specifying an effect size of interest, a significance level, a desired statistical power, and other design assumptions. Statistical power is the probability that a specified analysis will reject the null hypothesis when a particular alternative is true, given those assumptions.
Suppose a study is planned to have 80% power when the true standardized effect is 0.40. That does not create a hard boundary at 0.40. Effects of 0.39 do not suddenly become invisible, and effects of 0.41 are not guaranteed to be detected. Rather, smaller true effects will generally have a lower probability of producing the criterion used to declare evidence under that design.
Watch Out
A minimum detectable effect is not a measurement instrument's literal detection limit. It is a design-dependent statistical quantity. Its interpretation depends on the analysis, sample size, variability, significance criterion, desired power, and other assumptions used to calculate it.
The observed effect is not the same thing as the true effect
After the data are collected, you observe an estimate. The underlying population effect remains unknown. Sampling variation means the estimate may be larger or smaller than the true effect.
This distinction matters because saying "the observed effect was below the detectable effect" does not establish that the true effect is below that value. Nor does it establish that the study failed. The estimate needs to be interpreted together with an appropriate measure of uncertainty.
This is one reason post-study interpretation should not be reduced to asking whether the study had enough power. Once you have the data, the estimated effect and its uncertainty usually provide much more direct information about what the results support.
A non-significant result does not establish that the effect is zero
A study with limited sensitivity to small effects may produce a non-significant result even when a real effect exists. This is the familiar distinction between absence of evidence and evidence supporting the absence of an effect.
Consider two studies that both fail to reject a null hypothesis. In the first, the confidence interval is narrow and excludes effects large enough to matter. In the second, the interval is wide and includes substantial benefit, negligible effects, and perhaps harm. Calling both studies simply "negative" discards the most consequential difference between them.
Evidence consistent with little or no meaningful effect
The uncertainty is sufficiently narrow to exclude effects that would matter for the research question.
Insufficient evidence to determine the effect
The uncertainty remains wide enough that both negligible and meaningful effects are compatible with the data.
The confidence interval can reveal what the study has and has not ruled out
A confidence interval can help you assess the range of effect sizes compatible with the estimate under the statistical model and confidence procedure being used. If the interval remains wide, the study may have failed to narrow the relevant uncertainty enough for the intended conclusion.
The important comparison is often between that interval and substantively meaningful thresholds. Does the interval include effects large enough to change theory, practice, policy, design decisions, or future research? Does it also include effects so small that they would make little difference?
If both possibilities remain credible under your inferential framework, the scientifically appropriate conclusion may be that the study has not discriminated adequately between them.
Statistical detectability and practical importance are different questions
An effect can be difficult to detect statistically yet still matter in practice. Conversely, a sufficiently large study can provide strong statistical evidence for an effect too small to be consequential.
That distinction becomes especially important when the outcome accumulates across many people, repeated exposures, long periods, or high-stakes decisions. A numerically small effect is not inherently trivial. Its importance depends on the research context.
Before deciding that an effect is "too small," ask what magnitude would actually matter. This connects statistical design to the broader question of whether a result would change what someone should do , rather than treating statistical significance as the destination.
The smallest effect that matters should ideally be considered before data collection
A useful design starts with a substantive question: what is the smallest effect that would be important enough to distinguish from effects that are practically negligible? Depending on the field, researchers may describe this using concepts such as a smallest effect size of interest or a minimum clinically, educationally, scientifically, or practically important difference.
That threshold should have a defensible substantive basis. Choosing an effect merely because the available sample can detect it reverses the logic. The research question should inform the design, rather than allowing a convenient sample size to define what counts as important.
Thinking about what an informative result would look like before conducting the study can expose this mismatch while there is still time to redesign the research.
A smaller-than-expected effect does not necessarily make the study useless
Suppose the study was designed around an effect of 0.40, but the estimated effect is substantially smaller. The result may still be informative if the estimate is precise enough to exclude effects that would have changed the substantive conclusion.
For example, if the research question is whether an intervention produces an improvement large enough to justify an expensive implementation, a precise estimate near zero may be highly informative. It may show that effects of the magnitude needed to justify implementation are incompatible with the data, subject to the assumptions and limitations of the study.
By contrast, a small estimate accompanied by broad uncertainty may leave the question open. The issue is therefore not simply the size of the point estimate. It is what range of scientifically relevant possibilities remains after seeing the data.
04 · A Practical Example
When a promising intervention produces a smaller effect than expected
Hypothetical Example
An educational intervention expected to produce a moderate improvement
A research team evaluates a digital learning intervention. During planning, they determine a sample size intended to provide reasonable power for an effect of approximately 0.40 standardized units. They regard an effect around that magnitude as sufficiently important to justify institution-wide adoption.
After the study, the estimated effect is only 0.18 standardized units and does not meet the study's conventional statistical-significance criterion.
Do not conclude: "The intervention has no effect." The study has not necessarily established a zero effect. A non-significant result alone cannot support that statement.
Inspect the uncertainty around 0.18. If the interval remains wide enough to include effects near 0.40 as well as zero or effects in the opposite direction, the study has not resolved whether an important effect exists.
Compare plausible effects with the substantive threshold. If the interval is narrow enough to exclude effects near the magnitude required for institution-wide adoption, the study can still provide useful evidence against that particular practical expectation.
State what remains uncertain. The researchers might reasonably conclude that the data do not support the originally anticipated effect while acknowledging that smaller effects have not necessarily been ruled out.
The distinction changes the scientific message. "We did not detect the planned effect" is not equivalent to "there is no effect." At the same time, researchers should not rescue every non-significant result by claiming that a smaller effect might exist. The uncertainty around the estimate determines how much room remains for that interpretation.
06 · What This Means for You
Design around the effects you need to distinguish
If you are still planning the study, do not begin with the question, "What effect can my available sample detect?" Begin with the substantive problem. Determine which effect sizes would lead to meaningfully different interpretations or decisions, then assess whether the proposed design can distinguish among them with adequate precision or operating characteristics.
If the required sample is unrealistic, that is useful information. It may mean the design needs stronger measurements, lower variability, repeated observations, a more efficient design, additional sites, a different outcome, or a more focused research question. In some circumstances, it may indicate that the proposed study cannot answer the intended question credibly with the available resources.
If the study has already been completed, avoid retroactively changing the meaning of "important" merely to accommodate the observed result. Interpret the estimate against substantively defensible thresholds and report the remaining uncertainty clearly.
A simple decision framework
If the interval excludes effects large enough to matter
The study may provide useful evidence against effects of that practically important magnitude, even if it cannot prove an exactly zero effect.
If the interval includes both negligible and important effects
Treat the result as insufficiently precise for the substantive question. Avoid both "no effect" and confident claims of benefit or harm.
If the likely meaningful effect is smaller than the design can study adequately
Reconsider the design before data collection. Increasing sample size is one option, but design efficiency and measurement quality may also matter.
If even a very small effect would change an important decision
Plan the study around that smaller threshold rather than around a more convenient effect size.
The broader objective is to design a study whose plausible outcomes can support useful interpretation. If the most likely effects would leave you unable to interpret the plausible outcomes meaningfully , the problem should be addressed at the design stage rather than discovered after months of data collection.
And if the expected combination of effect size and uncertainty would make the study inconclusive almost regardless of what happens, that is a reason to consider whether you should redesign the research idea before proceeding .
07 · A Quick Checklist
Before relying on a study to detect a small effect, check these points
Before finalizing the design or interpreting the result, check:
Define the smallest effect that would matter scientifically, practically, clinically, educationally, or for the decision at hand.
Verify that the planned sample size and analysis provide adequate operating characteristics for the effect sizes you actually care about.
Do not treat the minimum detectable effect as a hard boundary between effects the study can and cannot observe.
After data collection, interpret the effect estimate together with an appropriate measure of uncertainty rather than relying only on the significance decision.
Ask whether the uncertainty interval excludes effects large enough to change your substantive conclusion.
Distinguish failure to find evidence of an effect from evidence that substantively important effects are absent.
Avoid redefining a smaller observed effect as important after seeing the data unless that interpretation has an independent substantive justification.
If meaningful effects would probably remain too uncertain, reconsider the design before committing resources to the study.
09 · The Bottom Line
A study should be able to distinguish the effects that matter
The Bottom Line
If the effect is smaller than the study was designed to detect reliably, you cannot automatically conclude that no effect exists; you need to determine whether the estimate is precise enough to rule out effects that would matter.
Plan studies around substantively meaningful effects rather than convenient detection thresholds, and interpret completed studies using effect estimates and uncertainty. Sometimes a small estimate is genuinely informative. Sometimes it simply reveals that the study could not resolve the question it was meant to answer.
11 · Cite this Guide
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.
Recommended (Field Guide)
APA
MLA
Chicago
Copy Citation
