Delphi method is to consult the forecast opinions of the members of the expert group by way of back-to-back communication. After several rounds of consultation, the expert group's prediction opinions tend to be concentrated, and finally a prediction conclusion that conforms to the future development trend of the market is made. Delphi method, also known as expert opinion method or expert inquiry investigation method, is a management technology that expresses opinions anonymously according to systematic procedures, that is, team members are not allowed to discuss with each other, there is no horizontal connection, and only have relations with investigators, so as to fill in questionnaires repeatedly, so as to gather questioners' knowledge and collect opinions from all sides, which can be used to build a team communication process and deal with complex task problems.

Delphi method, also known as expert prescribed procedure investigation method. The method is mainly that the investigators draw up a questionnaire and consult with the members of the expert group according to the established procedures; Members of the expert group submit their opinions anonymously (by letter). After several times of repeated consultation and feedback, the opinions of the members of the expert group gradually tended to be concentrated, and finally a collective judgment result with high accuracy was obtained.

Delphi method is essentially a feedback anonymous query method. The general process is as follows: after obtaining the expert's opinion, sort out, summarize and count the problems to be predicted, and then give anonymous feedback to the experts, solicit opinions again, concentrate again and give feedback again until a consensus is reached. This process can be simply expressed as follows:

Anonymous experts suggest-induction, statistics-anonymous feedback-induction, statistics ... stop after several rounds.

Therefore, Delphi method is a process of collective anonymous exchange of ideas in the form of letters and inquiries. It has three characteristics that are obviously different from other expert forecasting methods, namely anonymity, multiple feedback and statistical answers of groups.

not reveal one's name

Because in this way, all members of the expert group do not meet directly, but exchange letters, which can eliminate the influence of authority. This is the main feature of this method. Anonymity is an extremely important feature of Delphi method. Experts engaged in forecasting don't know who else is involved in forecasting. They exchange ideas completely anonymously. Later, the improved Delphi method allowed experts to meet for special discussion.

feedback

This method needs 3-4 rounds of information feedback, and in each feedback, the investigation team and the expert group can conduct in-depth research, so that the final result can basically reflect the experts' basic ideas and understanding of information, so the result is more objective and credible. The communication of group members is realized by answering the organizer's questions, and it usually takes several rounds of feedback to complete the prediction.

(3) Statistics

The most typical group prediction result is to reflect the views of the majority, and the views of the minority are at most general, but this does not explain the disagreement of the group. However, the statistical answer is not the case. It reports 1 median and two quartiles, half of which fall within two quartiles and the other half fall outside two quartiles. In this way, every point of view is included in such statistics, avoiding the shortcoming that the expert meeting method only reflects the views of the majority.

workflow

In the implementation of Delphi method, there are always two people, on the one hand, the organizer of prediction, on the other hand, the selected experts. First of all, it should be noted that the questionnaire in Delphi method is different from the usual questionnaire. In addition to the usual questionnaire content, it also has the responsibility to provide information to the respondents, and it is a tool for experts to exchange ideas. The workflow of Delphi method can be roughly divided into four steps. In each step, organizers and experts have different tasks.

1. Open the first round of investigation.

(1) The first round of questionnaires sent by the organizer to experts is open-ended, without any framework, only asking the prediction questions, and inviting experts to ask the prediction events around the prediction questions. Because if there are too many restrictions, you will miss some important events.

(2) The organizer summarizes expert questionnaires, merges similar events, excludes minor events, and puts forward an accurate list of predicted events, which will be sent to experts as the second step questionnaire.

2. The second round of evaluation survey

(1) Experts evaluate each event listed in the second step questionnaire. For example, explain the time of the incident, the controversial issue, and the reasons for the incident sooner or later.

(2) The organizer statistically processed the expert opinions in the second step and sorted out the third questionnaire. The third questionnaire includes events, the median and upper and lower quartiles of events, and the reasons why events occur outside quartiles.

3. Re-examination of the third round of research

(1) Issue the third questionnaire and ask experts to re-examine the argument.

(2) Evaluate the objections beyond the upper and lower quartiles.

(3) Give your own new evaluation (especially the experts who exceed the upper and lower quartiles should restate the reasons).

(4) If you correct your point of view, you should also explain the reasons for the change.

(5) The organizer collects new comments and arguments from experts, and counts the median and upper and lower quartiles similarly to the second step.

(6) Summarize the opinions of experts and form the fourth questionnaire. The focus is on the opinions of both sides of the argument.

4. Review the fourth round of research.

(1) Send out the fourth questionnaire, and experts will re-evaluate the trade-offs and make new predictions. Whether to require new argumentation and evaluation depends on the requirements of the organizer.

(2) Recycle the fourth questionnaire, calculate the median and upper and lower quartiles of each event, and summarize the reasons and arguments of various viewpoints.

It is worth noting that not all forecasting events will go through four steps. Some events may be unified in the second step, but will not appear in the third step; Some events may be after the fourth step, and experts' predictions for all events may be inconsistent. Inconsistency can also be inferred from the median and upper and lower quartiles. In fact, there will always be many events with inconsistent prediction results.

/link? URL = 1ok 78 apyzwwyzlkpjbem 1 hqantkbclbcl _ lmsqu 0 _ wqzacrksatkuo-j 0it r5w

See Baidu Encyclopedia-Delphi Method for details.