Predictive Customer Analytics - Order Predictions in Manufacturing and Business-to-Business Settings

Most order prediction approaches focus on retail and business-to-consumer settings with a limited forecast horizon (Sheil, Rana & Reilly, 2018; Romov & Sokolov, 2015; Wen, Yeh, Tsai, et al., 2018; Bernhard, Leung, Reimer, et al., 2016) creating a scientific gap concerning similar approaches for the manufacturing and business-to-business context. While it is important for business-to-consumer companies to plan and predict demand, the same is true for business-to-business manufacturing companies. It might be even more important for the later, especially predicting the needs of specific customers. This thesis aims to find out if the same or similar techniques as are in use in the B2C context can also be applied in the manufacturing and B2B context. For this purpose, currently used order prediction and related behavior prediction approaches are being analyzed to identify eligible models for a prototypic model implementation proofing the feasibility of this approach. To intensify the integration with the manufacturing and B2B context, data was sourced, analyzed and prepared in collaboration with a worldwide operating manufacturing company. By the prototypic model implementation, it could be shown that it is possible to reach high levels of accuracy and acceptable error rates for certain predictions. It however, seems that some models suffer from the reduced data available in the B2B context while other models perform remarkable. This work represents the initial proof of feasibility, while further research in this area remains necessary.