Utilizing Cluster Analysis for Decision Support in Material Planning

Abstract

Currently, in the manufacturing industry, planning parameters are infrequently updated. This entails that they over time will become outdated and inaccurate, which in turn results in inefficient operations. To increase the frequency of parameter updates, reducing the amount of work required would be useful. A way of doing this is to group items that have similar demand patterns, or what this study calls item behavior. Grouping the items opens the possibility of handling them in bulk. Therefore, the purpose of this study was to identify the characteristics that make up the item behavior and use these in a cluster analysis to find suitable groups of items. These groups should then be used as decision-support in material planning. This study was carried out in collaboration with Meridion AB which is a consultancy company that specializes in improving organizations' supply chains. The methodology used to conduct the study included quantitative and qualitative data collection. The main methodology used was cluster analysis, along with supporting methods like Z-score, min-max scaling, and the DeD method. In addition, several other methods were used to conduct the analysis, including an interview, discussions, and a workshop. The study resulted in five item behavior characteristics that were identified and then combined with each other as well as other parameters to group items based on their behavior. Three cluster analyses were then conducted and validated. The first was regarding safety stock levels, the second was used for ABC-XYZ classification, and the third was used to identify items where production quantities deviated from EOQ. These serve as examples of how cluster analysis can be used to increase the efficiency regarding the process of performing parameter updates, and how it can be applied to material planning issues. All three of these examples successfully provided decision support in material planning, although in different ways. In addition, there is also some more general knowledge regarding the use of cluster analysis for this purpose included that was gathered during the course of the study.