|A to Z of |
| Next |
The Delphi technique was developed in the 1950’s by the RAND Corporation as a tool for harnessing the views of a group of experts to forecast the potential damage from atom bomb attacks.
Other users for Delphi are in the surfacing and judging components of messy issues. Its main disadvantage being its high administrative overhead, however the method has been successfully incorporated in some computerised problem solving systems.
Between 2 – 5 consecutive questionnaires to a group of perhaps 15 – 25 people (occasionally up to 100) selected either as experts in the matter being investigate (if the intention of the exercise is to gather expert opinions on some issue) or as people directly concerned in some issue (if the purpose is to surface social or organisational concerns). E.g. a business creation agency used their voluntary steering group of local small business experts as a Delphi panel when trying to identify the psychological barriers inhibiting people from starting up their own businesses.
- Nominate the Panel; assuming they are experts and busy people, it is likely that they will require reassurance that there are advantages to their accepting the considerable commitment involved.
- Develop, send out, and get back the opening questionnaire; one or two broad open-ended questions are sent out initially and responses are preferred in the form of a list of separate sentences or short paragraphs rather than continuous text. A reminder letter may be required to encourage late responders.
- Develop, send out, and get back the second questionnaire; this subsequent document is created in light of the responses to the initial questionnaire. The responses to the first questionnaire are collated into a single anonymous list (using the original wording since participants will recognise their own contributions), the respondents’ are the asked to rate every item in the list (e.g. on a five point scale of importance, priority, feasibility, relevance, validity…) and finally to include any additional items suggested by the combined listing.
- A brief Delphi might end at this point; (in which case conclude), however a more extended Delphi may profit from additional rounds. The response ratings to questionnaire 2 are averaged and questionnaire 3 may ask the panel members to indicate where they felt the order of ratings need could be improved. There is no reason why this cannot be repeated for further questionnaires until a steady pattern materialises, but few expert panels have the patience for many further rounds unless the issue is crucial to them. Alternatively, the items rated above a certain threshold could be printed on separate cards, with a request for each panel member to sort the cards into related clusters.
- Thank the participants; the panel members will have been selected for their expertise and/or direct involvement, they are likely to have strong interest in the outcome, so a summary report and letter of thanks is usually forwarded to each member following the project. Some individuals may have given up substantial amounts of time to the project in which case a suitable ‘executive gift’ is often appropriate.
When the Delphi method is used to address a single, well defined, problem (such as its original use in estimating likely damage levels from nuclear war) the outcome may be easily summarised.
However, when used to surface and prioritise concerns, the output can be quite large (a panel of 20 can easily generate 15 – 20 concerns each – perhaps 2 – 300 distinct items) so as in any form of brainstorming or brain writing, some type of convergent post-Delphi analysis may be needed.