Learning Pages

Delphi Consultation Process


Home | Articles | Management | Delphi Consultation Process

In antiquity, Delphi, a village in central Greece, was the principal sanctuary and oracle of the god Apollo. The oracle's advice about religion, morality, commerce, and colonial projects, spoken by a priestess in a trance, was widely sought by individuals and the Greek states.

Today, we no longer seek the answers to momentous questions from priestesses. Instead, we turn to consultants or experts. But we don't seem to have the same level of confidence in our experts as the ancient Greeks had in their oracles because we often call in another expert and — not surprisingly — we get another answer. We then consult ten experts and get — twelve answers! We need a method for helping experts (and others) home-in on a high quality, agreed conclusion.

The traditional response to this is to lock the experts in a room and refuse to let them out until they come up with 'the' answer. In reality, it is difficult to get all the experts together at the same time, time is usually limited because of other commitments, there is limited reflection, and the stronger members of the group will often ignore the opinions of the less forthright members.

Most group processes will usually give an answer that is of a higher quality than would be achieved individually, but the Delphi consultation process does not need to have the experts in the same room or have them consider the problem at the same time. Because of this it is ideally suited for use on the Internet or through e-mail messages.

The following is a summary of how the process works:

  1. The experts individually give an initial answer.
  2. The answers are analysed and feedback is given to the experts.
  3. The experts then give their revised answers.

Steps 2 and 3 are repeated for two more rounds unless agreement has already been reached.The following example will give you an idea of how the process works in practice and can also be used as an exercise for teaching the Delphi consultation process.

Fill a large glass or transparent plastic jar with beans (having counted them first) and invite the 'experts' to individually estimate, to the nearest 50, the number of beans there are in the jar. If the experts are not in the same room, you will need to send a picture of the jar via email.

The experts return their estimates to you...

ExpertEstimate
13000
21000
32500
42500
52000
62500
71500
85000

...and then you collate and rank the results:

EstimateValueExpert
110002
215007
320005
425003
525004
625006
730001
850008

You then identify the middle, or median, value which divides the range of estimates in half. If there were an odd number of values, the middle value would be the median. But in this case there is an even number, so you have to use the average of the middle two values (estimates 4 and 5)

Both these estimates are 2500, so the median value is also 2500.

You also identify the interquartile range (IQR) which contains the middle 50 per cent of the answers. In this case, estimates 3, 4, 5 and 6 make up the middle 50 percent, so the interquartile range is from 2000 to 2500.

This information is then fed back to the experts in the following format:

MedianIQRYour First AnswerYour New Answer
25002000-2500

If your new answer lies outside the IQR please give your reason:



If you have a particular insight or expertise in this subject please give details:




The collated responses to this form are then given to everybody as shown below:

MedianIQRYour Previous AnswerYour New Answer
23002100-2400

Arguments in favour of a higher estimate
- The beans are very small.
- The jar is larger than it looks.

Arguments in favour of a lower estimate
- The beans are not packed very tightly.
- The jar is smaller than it looks.

Particular insights and expertise
- The expert who has done this exercise before, estimates there are 2300 beans.
- A bean has a volume of 2 cubic centimetres.

If you disagree strongly with any of these arguments please comment here:





The results and arguments are collated again and are presented to the experts for their final answer:

MedianIQRYour Last Answer
23002200-2400

Arguments defending a higher estimate
- An estimate of 2 cubic centimetres for a bean's volume takes packing 'tightness' into account.

Arguments defending a lower estimate
- The people who do not agree that the jar is smaller than it looks have not taken into account that the base is very thick.

Having considered the counter arguments, the respondents make their last forecast and return it for the final collation, which in this case might give a median of 2250 and an IQR of 2200-2400. This is then said to be the consensus for the Delphi consultation process.

Buy this article in PDF or ePub format for only £2.00. Your email will only be used to send you the article as an attachment.

Email:
Format:

Access to all published PDF/eBook articles is included in our Level 3 subscription.

973   127

Become a member or supporter to get early access to new articles.




Support the Learning Pages project | ☕️ Buy me a coffee