Delphi Method
The Delphi method is a multi-round, anonymous expert survey with controlled feedback: after each round, participants receive the group’s aggregated estimates and arguments and may revise their own judgement. The aim is a well-founded group assessment of uncertain questions about the future — without the distortions of open discussion.
Origin
The method was developed in the 1950s at the RAND Corporation for military forecasting questions; the foundational account is Norman Dalkey and Olaf Helmer, An Experimental Application of the Delphi Method to the Use of Experts (Management Science, 1963). The standard reference on the methodology and its variants is the volume edited by Harold Linstone and Murray Turoff, The Delphi Method (Addison-Wesley, 1975). The core idea: a group’s judgement improves when arguments circulate but status and rhetoric do not.
Typical use
Delphi is used when robust data are missing and judgement is the best available source: in technology foresight with horizons of ten years and more, in estimating arrival dates and diffusion speeds of new technologies, in policy advice, and to ground assumptions that feed into the scenario technique or technology roadmapping. Two to four survey rounds with fifteen to fifty experts are typical.
Procedure
-
Sharpen the questions and assemble the panel
The questions about the future are formulated so that they can be estimated — as an arrival date or a probability, for instance. The expert panel is deliberately heterogeneous: different disciplines, institutions and interests, to avoid collective blind spots.
-
Run the first round
The experts answer independently and anonymously, with a quantitative estimate and a rationale. Anonymity is constitutive: it separates the argument from the person and their status.
-
Provide controlled feedback and iterate
The facilitators mirror the aggregated results — median, spread, central arguments — back to the panel. Whoever deviates markedly from the group is asked to give reasons; everyone may revise their judgement. The cycle is repeated until the estimates converge or the disagreements are stable and well argued.
-
Evaluate and document the dissent
The result is not only the consensus value but equally the map of reasoned dissent: where experts remain permanently apart lies the uncertainty that is actually strategically relevant.
Limits and typical mistakes
Consensus is not correctness: even a converged expert group can be collectively wrong, as the history of technology forecasting amply demonstrates. The quality of the result depends entirely on panel selection — a homogeneous panel reproduces the convictions of its milieu with the appearance of methodological objectivity. The effort is considerable: several rounds over weeks tire the participants, and the drop-out rate rises with every round. A typical craft error is pressing for convergence for its own sake; whoever moderates deviating judgements away destroys exactly the information the survey was conducted to obtain. Finally, the quantitative form tempts readers to treat median values as forecasts rather than as structured pictures of expert opinion.
Relation to the Innovator’s Dilemma
In the Innovator’s Dilemma, assessments of disruptive developments rarely fail for lack of knowledge, but because of the social structure of the committees in which they would have to be voiced: whoever declares in a strategy meeting that the core business will be undermined within ten years by a technology ridiculed today contradicts the board, the best customers and their own career at once. The Delphi method neutralises this hierarchy and core-business bias through anonymity: estimates of disruptive developments become more honest when nobody is personally liable for them and no loss of status threatens. It is precisely the documented dissent of a Delphi panel — the question on which the experts remain permanently apart — that often marks exactly the spot where a disruption analysis should begin.