A national Consumer Packaged Goods (CPG) company supplies snack foods to consumers all over Canada. With sales heavily based on consumer purchasing behaviour, being able to accurately forecast sales is a large part of their continued success. According to their protocol on product replacement, any product that sits on shelf for more than seven days is considered to be spoiled and must be discarded. If the forecasted sales are greater than the actual sales, then the extra product will be discarded. If a forecast is much lower than the actual sales, there is wasted money and resources from the items produced that weren’t sold. The Supply Chain and Operations teams send sales forecasts to the Canadian plants to predict the demand for the various products for each week. The in-house built forecasting model currently being used requires an intuitive estimate of the sales for that week based on experience from the previous three to four years. The CPG has been using this particular inventory model for many years and was open to any ideas for improvement. Jesse Singh, a third year Management Engineering co-op student from University of Waterloo, was asked to compare various types of forecasting models to the current model and determine if there was an alternative model that could be used to increase demand forecast accuracy.
The primary teaching objective of this case study is to illustrate concepts of forecasting models in the targeted courses. Students will apply an appropriate forecasting model and decide on the best solution.