The demand planning process in semiconductor supply chains faces many challenges. In this process, individuals, their properties such as sensing capabilities and their interactions play a crucial role. This paper shows how agent-based modelling (ABM) can provide a computational testbed to investigate these aspects with respect to forecast accuracy. Based on the requirements of the demand planning context, we develop an empirically validated agent-based model of the demand planning process. In this model, we incorporate different concepts from behavioural science and the distributed cognition perspective. We show the usefulness of this agent-based computational testbed by using a case study from the semiconductor industry. Our model shows that demand planning accuracy does not depend on the planning capabilities of planners alone, but that the interactions of the individuals, emerging from the planning process design, may both positively and negatively affect accuracy.
