The Delphi Method

Expert consensus through anonymous rounds

The Delphi method is a structured research approach that builds consensus among experts through successive rounds of questionnaires. After each round, a summary of the group's responses is fed back to participants; anonymity reduces dominance effects, and iteration allows opinions to converge. The method is widely used for forecasting, priority-setting, and policy development, particularly in domains where empirical data are scarce or where expert judgment is the primary evidence base.

Conceptual Framework

The Delphi method is a systematic expert elicitation technique developed in the 1950s at the RAND Corporation by Norman Dalkey and Olaf Helmer. Its core logic rests on converting individual judgments into a group response through structured iteration. It was designed to address the problems created by face-to-face panels, including conformity pressure, hierarchical dominance, and premature consensus. The anonymous questionnaire format allows each expert to respond without knowing the identities of other participants, thereby minimizing social pressure and status effects.

Process and Main Types

Classical Delphi consists of three or four rounds: the first round collects open-ended expert opinions; in the second round, these are synthesized into a rating scale and the group mean and spread are fed back; in subsequent rounds participants revise their answers in light of group statistics. Policy Delphi aims to preserve diverse viewpoints rather than reach consensus. Real-Time Delphi accelerates the process by running rounds simultaneously on an online platform. The number of rounds ends when a pre-specified stability criterion is met, typically when opinion change between two consecutive rounds falls below a defined threshold.

Concrete Application Example

Health policy researchers might use Delphi to identify priority intervention areas for a national chronic disease prevention program. In the first round, a panel of public health experts, clinicians, and policymakers is asked to list possible interventions. In the second round, this list is standardized and participants rate each intervention on a 1-9 scale for impact and feasibility; the median and interquartile range are fed back. In the third round, participants share their reasoning for items that have not yet converged. The outcome reflects a normative priority ranking that could not be derived from empirical clinical data alone.

Common Pitfalls and Good Practice

The most common pitfalls include: homogeneity in panel selection (failing to include participants from diverse areas of expertise), reducing feedback summaries to the mean alone (ignoring minority viewpoints), not pre-specifying the number of rounds, and forcing consensus artificially. A well-conducted Delphi study defines the research question clearly, builds the panel through criterion-based sampling, shares both central tendency and dispersion statistics in each round, distinguishes genuine agreement from silent acquiescence, and typically keeps panel size between 10 and 30 participants.

Key terms

Controlled Feedback: Sharing summary statistics of group responses with participants after each round.
Anonymity: Participants respond without knowing each other's identity, reducing social pressure.
Iteration: Administering the questionnaire multiple times to allow opinions to converge.
Stability Criterion: Pre-specified threshold below which opinion change between two rounds signals convergence.
Policy Delphi: A Delphi variant aimed at surfacing diverse viewpoints rather than forcing consensus.