Discovery of Product Features for Redesign from User Implicit Feedback

Abstract

Online reviews are key sources to help manufacturers design and improve their products. However, using online reviews for product redesign is challenging because many online reviews contain useless information, and customers rarely provide explicit information about the products. The few methods available to recognize the customers’ demands from online reviews have several limitations, including failing to consider the redundancy of useless reviews, confusion in providing targeted information among different aspects of product design, and disregarding the manufacturing limitation. Hence, considering customer needs and manufacturing constraints for product improvement, this study proposes a new model called DPFR to discover the potential product features for redesign. In DPFR, it first analyzes and predicts helpfulness of reviews for more effective decision of product design and improvement, which is neglected in previous research. Then, a structured model based on topic modeling is conducted. Overlooked by previous researchers, DPFR also introduces a feature importance matric to measure the significance of each commented feature to product redesign. Third, a target features selection model is conducted further considering the limitations of manufacturers rather than only focusing on customer satisfaction. The effectiveness of the proposed model was assessed using a case study of data acquired from Bilibili, a large short-video platform in China. Focusing only on customer satisfaction, the results confirm the effectiveness of DPFR. Finally, including the constraints of product redesign, DPFR balances customer satisfaction and the manufacturing constraints such as redesign cost, redesign time, and carbon emissions. This study provides more practical suggestions for manufacturers to discover product redesign features that result in strategic product improvement decisions.