Labor is major cost component at Taco Bell, a large fast food chain specializing in Mexican food averaging 30 percent of sales. Scheduling the proper number of workers for a given time period is therefore critical in the highly competitive fast-food industry. Too many workers results in excessive costs and reduced profits. Too few workers on the other hand results in lost sales and/or poor service. with the-demand being highly variable throughout the day (52 percent of daily sales occur between 11:00 m. and 2:00 PM), Taco Bell needs to accurately forecast sales to schedule the proper number of workers. Weighted Moving Average After evaluating several forecasting techniques, Taco Bell adopted a six-week moving average. The number of customer transactions are recorded in 15-minute time intervals for each day of the week. For e ample, the fore castes
number of customers to be served next Friday between 10:30 m. and 10:45 AM is the six week average of the number of customers served in that same time period for the previous six Fridays.
This forecasting model is a major element in Taco Bell’s labor-management system”which is urinated to have saved Taco Bell $16.4 million in labor costs in 1996 (in comparison to the previously existing management system).
As noted in the above OM in Practice box, it is important to select the proper number of periods to include in the moving average. In determining the “right” number of periods to use, management must take into consideration several conflicting effects. As noted in Exhibit 9.6, the larger the number of periods included in the average. the greater the random elements are “smoothed,” which may be desirable in some cases. However. if a trend exists in the data–either increasing or decreasing-the resulting moving average has the adverse affect of constantly lagging this trend. Therefore, while a smaller number of periods in the moving average produces more oscillation, the resulting forecast will more closely follow the existing trend. Conversely, the inclusion of more periods in the moving average will give a smoother forecast but at the same time it will lag the trend by a greater amount. Exhibit 9.7 graphs the data shown in Exhibit 9.6, illustrating how the number of periods that are used in the moving average can impact the forecast. Note that the growth trend appears to level off at about the 23rd week. The three-week moving average responds better in following this change than the nine-week. although overall. the nine-week average is smoother.
The main disadvantage in calculating a moving average is that all the individual elements used in the average must be carried as data since a new forecast period involves adding the newest data and dropping the oldest data. For a three- or six period moving average, this is not too severe; but plotting a 6O-day moving average of the daily demand for each of 20,000 items in inventory would involve a significant amount of data. At the same time. today’s PCs are fast and efficient at doing multi product moving averages for a long time period. For example, using a 60-day moving average for 20,000 products would require 61 calculations per product x 20,000 products, or 1.220,000 calculations. With today’s PCs these calculations would probably take one or two seconds. at most. to complete.