Interpretable machine learning-based approach for customer segmentation for new product development from online product reviews

For new product development, previous segmentation methods based on demographic, psychographic, and purchase behavior information cannot identify a group of customers with unsatisfied needs. Moreover, segmentation is limited to sales promotions in marketing. Although needs-based segmentation considering customer sentiments on product features can be conducted to develop a new product concept, it cannot identify commonalities among customers owing to their diverse preferences. Therefore, this paper proposes an interpretable machine learning-based approach for customer segmentation for new product development based on the importance of product features from online product reviews. The technical challenges of determining the importance of product features in each review are identifying and interpreting the nonlinear relations between satisfaction with product features and overall customer satisfaction. In this study, interpretable machine learning is used to identify these nonlinear relations with high performance and transparency. A case study on a wearable device is conducted to validate the proposed approach. Customer segmentation using the proposed approach based on importance is compared with that employing a previous approach based on sentiments. The results show that the proposed approach presents a higher clustering performance than the previous approach and offers opportunities to identify new product concepts. •An interpretable machine learning-based approach is proposed for customer segmentation.•Customer segmentation is performed based on the importance of product features.•This study employs SHAP method to explain black-box machine learning model.•Customer segmentation of a wearable device is conducted from 5045 reviews.•The proposed approach is validated by comparing a previous approach based on sentiments.