When Delphi should produce two answers, not one
Delphi is designed to identify consensus—but it should not manufacture it. When two well-supported positions remain stable after feedback and reconsideration, preserving both may be more informative and defensible than forcing one combined conclusion.
Disagreement can be a finding
A single median or agreement percentage can hide a polarized distribution. For example, two groups may rate the same statement differently because they work in different settings, represent different stakeholders, or are applying the statement under different conditions.
Signals that two positions may need to be preserved
A bimodal distribution
Ratings cluster at two separated parts of the scale rather than spreading randomly around the middle.
Stability after feedback
The same division remains after panelists review group results and have a genuine opportunity to reconsider.
Coherent rationales
Qualitative comments show that each cluster is supported by clear reasoning rather than misunderstanding or ambiguous wording.
Stakeholder differences
Clinicians, patients, caregivers, researchers, or policy leaders may prioritize different outcomes or risks.
Context-dependent answers
Different settings, populations, resources, or implementation conditions may make different conclusions appropriate.
No obvious design defect
The split persists after checking wording, scale direction, missing information, and whether participants interpreted the item consistently.
How to analyze and report stable disagreement
- Show the full distribution instead of relying only on a median or overall percentage.
- Compare rounds to determine whether the split is stable, narrowing, or widening.
- Review anonymized qualitative reasoning for each position.
- Examine relevant stakeholder groups while reporting denominators and avoiding strong claims from very small subgroups.
- State whether the two positions depend on setting, population, role, or another interpretable condition.
- Report both positions and the circumstances in which each may apply.
A recent four-round Delphi study on climate-change acceptance retained two context-dependent definitions after stable bipolarity persisted. The broader lesson is methodological: a transparent Delphi can clarify the structure of disagreement instead of erasing it.
How Surveylet and Calibrum can help
Surveylet supports response distributions, round-to-round feedback, anonymized comments, stakeholder grouping, and analysis-ready exports. Calibrum’s reporting services can integrate quantitative patterns with qualitative reasoning so consensus, unresolved items, and stable divergent positions are represented accurately.
Let the evidence show agreement—and meaningful disagreement
If your results contain polarization, stakeholder differences, or items that resist convergence, we can help determine whether they reflect a design issue, uncertainty, or distinct positions that deserve to be reported.