CASE STUDY: VBW Catering VBW Catering is a small company that makes and delivers lunchtime fast -food
to local businesses. It has two principal product ranges for each of which daily demand can be aggregated into common units for the purposes of planning. The first range consists of basic sandwich products sold to individual customers. The second range comprises pre-assembled light lunch menus ordered for example, for lunchtime business meetings. The owners of the business currently have difficulty in forecasting demand on a day-to-day basis. This creates problems for them in planning the purchase of materials and in the allocation of work to staff. They have been advised to consider using an exponentially weighted moving average technique in order to predict future demand. The idea being, that once a suitable model is selected, it can be used each evening to predict a demand figure for the following working day. In order to investigate the suitability of this method of forecasting to their business and to develop an appropriate model, the owners have collected data on demand over a typical period of 20 days. This data is given in the following table. You are required to use this data to test two forecasting models - in the first, the value of the smoothing constant () is to be 0.1 and in the second it is to be 0.5. In order to start the modelling process, in both cases, you may use an initialising forecast value for working day 1 of 50 units.
Product range 1 (basic sandwiches) Working day number Demand 1 50 2 47 3 48 4 55 5 51 6 46 7 52 8 47 9 57 10 50 11 44 12 52 13 48 14 47 15 44 16 46 17 56 18 42 19 55 20 49 Product range 2 (pre-assembled lunch menus) Working day number Demand 1 78 2 81 3 91 4 22 5 10 6 82 7 15 8 18 9 28 10 79 11 72 12 87 13 16 14 20 15 86 16 22 17 89 18 76 19 13 20 15
1. For each value of the smoothing constant ( ) separately, use the method given below to calculate EWMA forecast figures for days 2 to 20 for both product ranges. (This means that you should develop 4 different forecasts two each for each product range). EWMA forecasting method: The forecast figure for tomorrow = * demand today + (1 - ) * demand figure forecasted for today NB. In carrying out this exercise, you need to remember that an EWMA forecast figure is calculated for the following day each evening once the current days actual demand is known. At this point in time, the demand figure for the following day is not known this is what you are forecasting! You have been given all the demand figures for days 1 to 20 at the start of the exercise, but, in order to simulate the actual process, you will need to work through the data day-by-day ignoring future demand figures. 2. Plot the actual demand figures for days 1 to 20 for each product range against the forecasts that you have developed using the two different models. 3. For each of the four forecasts, calculate the mean forecast error and the mean absolute deviation. 4. Comment on the results of the exercise that you have carried out and develop conclusions about: the suitability of the EWMA approach to forecasting in this case the most suitable model (value of ) to use in this case the influence of the value of on any EWMA forecast