Delphi panel size: how many experts do you need?

Start with coverage, not a universal minimum

A narrowly defined technical question and an international core outcome set study need different recruitment plans. List the perspectives required to answer the question before choosing the total number of invitations. A larger panel cannot compensate for excluding a group whose experience matters.

Methodological literature describes substantial variation in Delphi panels and designs. This makes a transparent study-specific rationale more useful than borrowing a number without its context. Niederberger and Spranger, Delphi Technique in Health Sciences: A Map.

What a review of 1,904 Delphi studies found

A 2026 meta-research study found that only 19% of the reviewed Delphi studies reported a planned sample size, and only 16% explained the rationale. Panel sizes and consensus methods varied widely.

This evidence does not establish a universal ideal panel size. It supports a more practical conclusion: define a reasoned target before recruitment, preserve the rationale, and report actual participation and attrition transparently. Read the Journal of Clinical Epidemiology study.

Plan stakeholder groups separately

If the study needs distinct patient, clinician, and researcher perspectives, define the desired coverage within each group. Then decide whether consensus will be assessed across all responses, within groups, or through both views.

An overall total can look adequate while a key group has very few remaining participants. See our stakeholder balance guide for the implications of unequal group sizes.

Allow for recruitment and attrition

Distinguish the number invited, the number who start, and the number who provide usable final-round responses. Use realistic recruitment and retention assumptions, informed by your setting where possible.

Remember that percentages behave differently in small panels

One person's rating changes the percentage by 10 points when an item has 10 numeric respondents, and by 5 points when it has 20. This does not make a smaller panel automatically invalid, but it affects how a threshold should be interpreted and reported.

Report the actual numerator and denominator, not only a rounded percentage. Explain how missing answers and unable-to-judge responses were treated. Our consensus-threshold guide covers the rule-setting decisions.

What should the panel-size justification report?

• The expertise and stakeholder perspectives required.
• How eligibility, recruitment, and desired group coverage were defined.
• The starting target and the assumptions about retention.
• Actual participation by round and relevant group.
• How attrition or missing perspectives may limit interpretation.

Describe the panel as recruited. A purposive expert panel should not be presented as a population-representative probability sample. Also consider whether small subgroup summaries could make individual participants recognizable.

Use monitoring to protect the study plan

Track participation during each round and use consistent reminders. Review whether dropout has changed the balance of perspectives before treating a higher agreement percentage as evidence that the original panel moved closer together. If a later round contains different respondents, explain that limitation.

Surveylet supports panel management, response tracking, and Stakeholder Analysis. The research team defines eligibility, recruitment targets, and how the composition of the panel will be considered in the analysis.