How to use expertsâ€™ opinion in forecasting demand? If the number of experts is just one, than whatever he / she tells could be your forecast, subject to of course own well found modification. When the number of experts is more than one, there are two approaches. One takes a simple or weighted average of the numbers given by various experts, temper it with own well conceived judgment and arrive at the needed forecasts. Under this situation, the more weight would obviously be attached to the persons having greater expertise than those having limited expertise. Two, follow the Delphi method.
Under Delphi method, opinions are first collected from experts and then instead of mathematical averaging efforts are made to match them. It is true that no two experts agree, but something could be done to bridge the gap. This is done through bringing the experts together, arranging one or more meetings and arriving at some narrow range for the forecast under attempt. The narrow range could, of course be used to give the interval forecast directly (i.e. the lower and upper limits within which the demand is forecast to be) and for arriving at a point (specific number) forecast by tampering it with the overall assessment of the researcher or the coordinator of the forecasting exercise. For example, if the demand forecast for, a product in India is being attempted for the year 2003 through this method, one could proceed through the following stages:
1.Request all the experts of product X to give their individual estimates for the likely demand for X in 2003. Suppose there are 15 experts and their requisite forecasts are the following:
100 110 190 200 210 225 240 240 250
250 260 275 300 310 400
2. If the difference in forecasts is significant, invite the experts for a conference on the subject, present the problem with regard to differences in their estimates without spelling out who said what, and request them to resolve the problem. In the debate, people who gave either too small numbers (e.g. 100 and 110) or too large numbers (e.g. 400) whose identity should of course, not be disclosed, would generally be cornered and others would go on arguing others or getting convinced by others, through exchanging new inputs from the colleagues in the conference. Hopefully, this would lend to narrowing the limits for likely demand in 2003.
3. If the coordinator of the conference (researcher) feels that the range of variations is still large, he could adjourn the conference for the time being, give lunch (dinner) break and invite them for a second round of discussion. This exercise should be repeated. Iteration of this would continue until the coordinator is able to arrive at an acceptable range or beyond which he does not see any scope for reduction. Suppose the exercise leads to the interval of 240 to 280.
4. Declare the so arrived range (240 to 280) as the interval forecast for the demand for product X in the year 2003.
5. Take a simple average of the lower (240) and upper (280) value of the forecast and declare that (260) as the point forecast for the variable under forecasting It should be emphasized that the coordinator would have also participated in the debates and so his judgment would have already been reflected in the consensus reached at stage Â© above
It is obvious from the above discussion that the Delphi method is quite sound but it could be tedious and costly. Thus in the situations where the number of experts is not too large and they are co-operative, and the researcher has the researcher has the necessary fund and the authority to perform the task the Delphi method could appropriate for demand forecasting.