Systemising Data-driven Methods for Predicting Throughput Time within Production Planning & Control

Abstract

Predicting throughput times is of particular interest to production planners to schedule the production flow or communicate reliable delivery times to customers. Most established prediction methods are based on general assumptions, expert knowledge or simple statistical techniques. With the increasing use of data mining in production management, it is possible to provide more sophisticated predictions of throughput time. However, current research often does not describe the application or locate the particular prediction approach within the time and task structure of Production Planning and Control (PPC). Therefore, this paper aims to develop a systematisation approach to classify prediction models within the PPC task structure. To this end, applications along the order fulfilment process are first defined and then elaborated. A systematic literature review is conducted to classify current throughput time prediction approaches within the previously described application domains. In a case study, the application possibilities of throughput time predictions based on the provided systematisation are demonstrated, and differences in data availability and prediction quality are highlighted.