What Is the Delphi Method? Building Expert Consensus Round by Round
What Is the Delphi Method?
The Delphi method is a structured consensus procedure in which experts who cannot easily be assembled in one place are surveyed anonymously and repeatedly, with each round's results fed back to converge their opinions. Anonymity reduces the influence of a few authoritative voices and the pressure to conform.
When Do You Use the Delphi Method?
Use it when objective data are too scarce for prediction, or when new concepts, indicators, competencies, or criteria must be established through expert consensus. It is also frequently used to validate the content validity of scale items.
How Does a Delphi Study Proceed?
It runs over several rounds, from panel composition to confirming consensus.
Step 1: Compose the Expert Panel
Select panelists on the basis of subject expertise and experience, considering both representativeness and diversity.
Step 2: Round One (Open-Ended)
Collect broad opinions and candidate items with open-ended questions. When structured items already exist, the study sometimes begins in closed-ended form.
Step 3: Iterative Rounds (Convergence)
Organize the collected opinions and have panelists rate them again, presenting the previous round's group statistics alongside each panelist's own response so opinions converge.
Step 4: Confirm Consensus and Close
End the study once the consensus criteria are met or opinions have stabilized.
How Many Experts Should the Panel Have?
There is no absolute rule, but 10 to 20 panelists is common, and a more homogeneous panel can work with fewer. Recruit with a margin to allow for attrition across rounds.
How Is Consensus Judged?
Through the content validity ratio (CVR), interquartile range (IQR), convergence, and stability. CVR is commonly used for rating item necessity, applying the minimum CVR threshold that corresponds to the panel size.
Summary
The Delphi method derives expert consensus through panel composition, an open-ended first round, iterative convergence, and confirmation of consensus. Anonymity and feedback are essential, and consensus should be judged with explicit criteria such as CVR and convergence. It differs from meta-analysis, which statistically pools the results of existing studies.