What can this evidence actually support?

Evidence is not strong or weak in the abstract

It is suitable or unsuitable for a particular conclusion and use.

A blinded sensory study may be excellent evidence that two formulations are perceptibly different under controlled conditions. It may be poor evidence that one formulation will produce greater long-term adoption. A qualitative study may be excellent for language, mechanisms and hypotheses. It cannot estimate prevalence. An actual-use study can reveal behaviour under relevant conditions while giving up some causal isolation.

The evidence question becomes more demanding as the intended conclusion travels farther from the design.

Six checks before translating a result

1. What is being measured: Did the measure actually represent the perception, behaviour, mechanism or outcome named in the conclusion?
2. Population: To whom can the result be generalized?
3. Comparator: Against what baseline or alternative is the result meaningful?
4. Context and time: Under which conditions and duration must the conclusion hold?
5. Analysis and uncertainty: Are effect size, precision, multiplicity, missingness and practical thresholds visible?
6. Documentation and boundary: Can another person see how the conclusion was reached and what the design does not support?

Reporting frameworks such as PRISMA can improve transparency in evidence synthesis, and systems such as GRADE can help make judgments about bodies of evidence explicit (Guyatt et al., 2008; Page et al., 2021). They do not remove the need for scientific judgment. Nor does statistical significance settle the importance or intended use of a result (Wasserstein et al., 2019).

Claims make the inference public

This is especially important for claims and regulated products. A sentence in marketing, a label, a standard, a regulatory position or a scientific paper is an inference made public. The concept measured, population, comparison, outcome, analysis and conditions have to support the wording. Health-product guidance similarly emphasizes that the evidence must fit the specific claim being communicated.

The useful question is not merely “Did the study work?” It is “What can this design legitimately allow us to say, to whom, under what conditions and for what decision?”

Good scientific translation does not make evidence sound bigger. It makes the boundary clearer.