Forecasting food waste: The case of a Swedish grocery company - A quantitative & exploratory study of how to forecast food waste for a Swedish retailer

Abstract: In the retail stage of a food supply chain, the occurrence of food waste is often caused by inaccurate forecasting of sales, which leads to incorrect ordering of products. Previous literature has suggested using more accurate demand forecasts on the sales data in order to combat the problem. However, a few authors have suggested applying forecasting methods to the waste data in order to reduce the food waste. This thesis explores the potential predictive power that forecasting methods possess in explaining food waste by comparing more advanced methods to the simplest form of forecasting, namely the Naïve forecast. Comparisons are made between a Stepwise Regression method, with a set of explanatory variables, lags and manually constructed variables, Exponential Smoothing (ES) methods and a combination model of the Stepwise Regression method and the ES methods, were used in order to assess which of these are the most accurate in terms of predicting food waste data at different aggregated levels. By using a set of different error metrics, more nuanced conclusions can be drawn regarding which aspects of the food waste data is explained by the models. The variables from the Regression model can further describe if and what factors actually explain food waste and if these differ for aggregated levels of the data. At the highest aggregated level, the Combination model has the most accurate forecast. At lower aggregated levels the results show that the ES has the most accurate predictive power in terms of explaining seasonal structure, whereas the Stepwise Regression model is, on all aggregated levels, most proficient in terms of explaining outliers. Some of the Stepwise Regression model's explanatory variables display consistent, significant correlation to food waste, indicating that these could be leveraged by practitioners. Our study suggests that it is feasible to forecast food waste data and that distinct waste-affecting variables exist. These variables can be used to explain the causes of waste, and subsequently minimize it. Grocery retailers can, by adopting the conclusions drawn in this study, and integrating them into their strategy, effectively reduce their overall food waste. This helps retailers to save money, mitigates environmental effects and aids them in reaching critical sustainability goals.