id: delphi-method
title: Delphi Method
description: >-
  A structured method for gathering expert judgement through rounds of
  anonymous questionnaires with controlled feedback, used when direct
  meeting is impractical or would be distorted by dominance. Developed at
  the RAND Corporation in the 1950s.
source: RAND Corporation
source_url: https://www.rand.org/
tags: [delphi, expert, decision-making, foresight]
forked_from: null
image: null
steps:
  - headline: Define the question
    description: >-
      The coordinators define the question and recruit a panel of experts
      who will participate across several rounds.
  - headline: First round questionnaire
    description: >-
      Each panellist responds independently and anonymously to the first
      questionnaire, providing their judgement and reasoning.
  - headline: Summarise and feed back
    description: >-
      The coordinators summarise all responses — including areas of
      agreement and disagreement — and feed them back to the panel
      without identifying who said what.
  - headline: Revise in light of others
    description: >-
      Each panellist revises their judgement in light of the summary,
      optionally explaining why they hold their view against the group.
  - headline: Iterate to convergence
    description: >-
      Rounds continue (typically three or four) until judgement stabilises.
      The final summary reports the distribution of views, not a forced
      consensus.
outcome: >-
  A reasoned, anonymised convergence of expert judgement that resists the
  dominance and groupthink of face-to-face meetings.
practice: >-
  Anonymity is the engine — without it, the most senior or loudest voice
  distorts the rest. Do not force consensus; report the spread of
  judgement honestly.