Fisher and Raman (1996), Fisher et al. (2001) propose to let a number of experts within a company estimate the demand for a product. The demand is calculated as the average of the experts' estimates. The method is straightforward and very applicable for the speci?c company that we consider. Indeed, both Fisher and Raman (1996), Fisher et al. (2001) also proposed the method for a retailer in the apparel industry.

Mantrala and Rao (2001) also develop two forecasting methods based on experts' estimates for an apparel retailer (of the demand for mens walking shorts for the spring season). Their methods are more detailed than those of Fisher and Raman (1996), Fisher et al. (2001), since they divide the season into a number of periods and also consider different price levels. The ?rst method starts by asking each expert separately for the minimum, maximum, and most likely (modus) demand for each combination of period and price.

Subsequently, using the Delphi group method, the experts have to reach consensus on the minimum, maximum, and modus for each combination of period and price. Finally, for each combination of period and price, the forecast (for the mean) is calculated as the average of the minimum, the maximum, and the modus. The second method asks different input from the experts: an estimate of total (over all periods) demand at a single price, as well as a 95% con?dence interval; the expected percentage of total demand that will occur in each period; and an estimate of the price elasticity.

Based on these inputs, total demand is estimated using a rather complicated model including a log-normal disturbance term. The authors do not report any results on the quality of the resulting forecasts of the two methods.

Based on a survey among 240 ?rms, Sanders and Manrodt (2003) report judgmental forecasting methods to perform less well than quantitative methods. They offer two explanations for the poor performance of judgmental forecasting. First, there are a number of inherent biases, including optimism, wishful thinking, lack of consistency, politicalmanipulation, and overreacting to randomness. Second, people have a limited ability to consider and process large amounts of information.

On the other hand, judgmental methods are often preferred by practitioners, since they can incorporate special insights, trends, and macro-economic factors, which are hard, if not impossible, to quantify in practice and since practitioners are more acquainted with them. Moreover, a lack of data often rules out the use of complex forecasting methods. This is certainly true for the mail order retailer that we consider.