Improving grocery demand forecasting- a design science approach
Loading...
URL
Journal Title
Journal ISSN
Volume Title
Perustieteiden korkeakoulu |
Master's thesis
Authors
Date
Department
Major/Subject
Mcode
SCI3049
Degree programme
Language
en
Pages
56 + 15
Series
Abstract
The grocery retail industry is growing rapidly while becoming increasingly competitive. At the same time, customers are demanding higher product availability and lower prices. In this business environment, competitive advantage can be built on effective demand forecasting, which not only increases company revenue, but also improves customer satisfaction and sustainability. In this thesis I use a design science approach to develop a new solution for grocery demand forecasting. The new solution, which is a method of automatic daily forecasting using TBATS modeling, is evaluated for both fast moving and intermittent demand based on real data obtained from a forecasting solution provider (case company). The new solution comprises two main approaches to tackle different weekly demand profiles, and a modified approach for intermittent demand. For fast moving products, the new solution showed better performance, especially in the reduction of extreme biases (over 50% better than the current solution). For intermittent demand, the current solution showed better performance on several measures, however, the new solution gave comparable/ better performance on the two most practically useful measures. While the new solution’s modified approach performed better than the current solution for intermittent demand, its applicability in the industry needs to be further researched. The increased complexity of the new solution was deemed a justifiable tradeoff when considering the improvement in forecasting performance and efficiency due to automatic daily forecasting. Daily forecasting methods are still uncommon in grocery retail. This thesis shows that TBATS modeling is feasible when moving towards daily forecasting. It also utilizes the concepts of design science, which is a comparatively new field of research. This thesis introduces extreme biases as a forecast performance measure and presents a method of aggregating forecast errors as well. The new solution provides a starting point to create a new generation of efficient forecasting tools on which grocery retailers may build a forecasting-based competitive advantage.Description
Supervisor
Holmström, JanThesis advisor
Öhman , MikaelSillanpää, Ville