Delphi Technique is a process whereby experts come to a consensus about future events without
face-to-face discussion is called the Delphi Technique. It is a means of Group Decision Making
representing one of the first formalized methods of systematically obtaining and aggregating group
judgments.  It was developed at the Rand Corporation  in late 1960s (Scott Sink, "Delphi Study on
Root Causes of Implementaoin and Deployment Problems of Quality and Productivity Improvement
Initiatives at Regional, National and Organizational Levels", 9th World Productivity Congress,
1995, Istanbul) . It is especially effective for technological forecasts, because precise data for
predicting technological break-throughs are not available. Technological experts, typically composed
of people who represent a wide range of professional modes of inquiry, fill a questionnaire about
future events, and the responses are summarized and returned to participants. They then complete
a new questionnaire based on their own previous responses and  the estimates of other experts.
The process continues until a consensus is reached. For example, a Delphi panel might be convened
to examine the future of the American steel industry.

The Delphi Technique promotes independent thought and precludes direct confrontations and participants'
defensiveness about their ideas. Its biggest advantage is that experts with widely different opinions can
share information with one another  and reach agreement about future predictions (Ivancevic et.al.,
Management, 1994, Irwin-Illinois). The method represents a systematic approach for including
multidisciplinary perspectives on planning problems. Conventional Delphi processes are characterized
by anonymity and controlled feedback. The anonymity of panel members is preserved by physical
separation and by the use of carefully selected questionnaires or other communication procedures such
as on-line computer communication (Thomas Saaty, Multiple Criteria Decision Making: The Analytical
Hierarchical Process, 1991, Pittsburgh). The panelists are (in the ideal case) experts in the special
field of application and their number has to be big enough to cover all aspects of the topic and will therefore
depend on the complexity of the questions and on the numbers of related fields. Anonymity is necessary to
guarantee that the ideas and arguments are not influenced by the reputation of panelists supporting them
and that there will be no cooperation and coordination of the panelists during the Delphi inquiry.

The results of a Delphi experiment may not only identify technological trends and development alternatives,
but also provide insights into the assumptions and methods used in driving judgments by the selected
experts. The technique is also educational because the built-in feedback provides the respondents with
information and opinions which may be useful in forming their informed judgments (Ernst Frankel,
Management of Technological Change, 1990).

A typical Delphi technological forecasting experiment consists after Frankel,
of the following steps:

1. Definition of process, product, or service technology to be forecast. This often takes the form of definition
of a technological problem or opportunity and not of a particular type of product or process with some
general performance characteristics.

2. Determination of the areas, range and depth of expertise and knowledge required fairly judge future
technological problems.

3. Identification and evaluation of experts in terms of credentials, credibility, reputation, communication
skills and bias.

4. Selection of respondent group (anonymous).

5. Identification of major issues affecting technological developments under consideration, including
constraints, exogenous, regulatory and institutional issues.

6. Preparation of a first questionnaire. This should be broad or general and elicit respondent comments,
information disclosure, and more.

7. Analysis of questionnaire results and development of general consensus including elimination of some
considerations and developments. Formulation of additional feedback and other information designed to force
convergence of expert opinions. Tabulation of results.

8. Preparation of second questionnaire and distribution.

9. Repetition of step (7) and test for consensus or at least adequate convergence of opinions for a reasonable
forecast of technological change.

10. Questionnaires are reissued until such consensus is reached, when final responses, as well as background
information, are compiled, and the results formulated as a technological forecast.

A Delphi experiment may result in more than one alternative technological forecast.

It is important to assure that experts are not influenced by the objective or objectives of the experiment.
As a measure against this, potential Delphi experiments usually use two or more separate groups of experts,
all made up of members of users, monitors and experts.

There are a lot of variations of the Delphi technique, such as more specialized Delphi applications for policy
forecasting and decision prediction.

To the Homepage