Why a simple majority is not necessarily Delphi consensus
A bare majority can decide which side wins a vote. It does not necessarily show that an expert panel reached strong, stable, or broadly shared agreement.
A narrow win is not necessarily broad agreement
A Delphi consensus rule should reflect the study purpose and be defined before the results are known. When the threshold is only just above 50%, label the finding accurately and examine the full response distribution, disagreement, stakeholder differences, and stability before calling it consensus.
Consensus does not require unanimity. But it should mean more than whichever side happened to receive one additional rating.
A decision rule and a consensus claim are not the same
In a yes-or-no vote, 51% may be enough to choose an option. Delphi studies usually make a different claim: that repeated, structured expert judgment has produced a meaningful level of agreement. A result close to an even split may satisfy a voting rule while still showing substantial disagreement.
Simple example
If 11 of 20 panelists support an item, the item has 55% support—but nine panelists do not support it. Calling that result “consensus” without qualification may overstate what the panel concluded.
The same percentage can also mean different things in panels of different sizes. Always report the numerator and denominator, not only the rounded percentage.
A low threshold may still be defensible for an exploratory screening step or an explicitly defined majority decision. The reporting should match the claim: “endorsed by a majority” is more precise than “strong consensus” when the panel remains nearly divided.
Recent research shows why the threshold needs justification
A 2026 meta-research study examined 1,904 Delphi studies and found that percentage thresholds ranged from 50% to 100%, with 80% used most often. This variation does not make 80% a universal rule; it shows why authors should explain the threshold selected for their own study. Read the Journal of Clinical Epidemiology study.
A recent three-round study developing and validating a questionnaire for postamputation pain used a greater-than-50% endorsement criterion and identified the comparatively low threshold as a limitation. Its transparency is valuable: the example shows that a method can be reported carefully while still inviting a cautious interpretation of what “consensus” means. Read the applied Delphi study.
The practical lesson is not to copy the most common percentage automatically. Define what level of agreement is meaningful for the decision, explain why, and preserve the evidence needed to interpret the result.
Look beyond one agreement percentage
A defensible analysis asks whether the apparent agreement survives other views of the data. The appropriate combination depends on the scale and research question.
| Measure | Question it answers |
|---|---|
| Response distribution | Are ratings clustered, dispersed, or polarized? |
| Median and interquartile range | Where is the center, and how tightly are ratings grouped? |
| Disagreement rule | Do meaningful numbers of panelists occupy opposing parts of the scale? |
| Stakeholder analysis | Does pooled agreement conceal a group-specific objection? |
| Round-to-round stability | Did judgments converge, remain stable, or continue to move? |
| Qualitative comments | Why do panelists support, reject, or interpret the item differently? |
A high percentage can still hide a divided stakeholder group, while a lower percentage may reflect two defensible context-dependent positions. The label should follow the evidence rather than forcing every item into “consensus” or “no consensus.”
How to choose and report a threshold
1. Match the rule to the construct
Agreement, importance, feasibility, validity, and appropriateness are not interchangeable. Define what the rating is intended to measure before choosing the qualifying band.
2. Define it prospectively
Record the qualifying rating band, required percentage, denominator, treatment of missing or unable-to-rate responses, and any disagreement rule before launch.
3. Consider the consequences
A final clinical recommendation may warrant stronger evidence than an exploratory list of candidate ideas.
4. Use several classifications when useful
Consensus in, consensus out, near consensus, unresolved, and stable disagreement can be more informative than a binary label.
5. Describe deviations honestly
If a rule changes after data are reviewed, report the original rule, the change, who approved it, and why.
How Surveylet fits the methodology
Surveylet supports configurable Delphi questionnaires, rating scales, multiple rounds, controlled feedback, stakeholder analysis, and consensus-oriented reporting. The research team defines the study-specific threshold and interpretation rather than relying on an unexplained universal percentage.
For detailed examples of thresholds, inclusion and exclusion rules, subgroup findings, and disagreement, see our guide to consensus thresholds.
Common questions
Does Delphi consensus always require 75% or 80% agreement?
No universal percentage fits every study. The threshold should be justified for the research purpose, scale, panel, and consequences of the classification. Surveylet commonly supports studies using a predefined percentage threshold together with a disagreement rule.
Can a simple majority ever be appropriate?
Yes, when the purpose is explicitly majority endorsement or an exploratory screening decision. The study should describe it that way and avoid implying stronger agreement than the data show.
What if the overall threshold is met but one stakeholder group disagrees?
Report both findings. Overall consensus and stakeholder-specific consensus answer different questions. The group difference may be substantively important even when the pooled result passes.
Define consensus before the first round
We can help align the rating scale, threshold, disagreement rule, stakeholder analysis, and reporting plan with the purpose of your Delphi study.