Unlocking online product return behaviour: the influence of product attributes on customer interaction styles
本研究区分了主动与非主动退货者，利用亚马逊评论数据通过机器学习分析产品内在和外在属性如何影响不同风格的退货行为，为制造商和电商管理者提供策略参考。
Purpose Despite growing research on online product return behaviour (OPRB), customer behaviour remains complex and unpredictable. Some customers return products assertively with clear complaints, while others exhibit hesitation and silent dissatisfaction. This study distinguishes between assertive and non-assertive returners and examines the intrinsic (e.g. performance and reliability) and extrinsic (e.g. warranty and product information mismatch) product attributes influencing their behaviours. Design/methodology/approach Drawing on a large, global dataset of Amazon customer reviews – a reliable source of customer-perceived insights – we employ Latent Dirichlet Allocation topic modelling and (semi)-unsupervised machine learning models (e.g. gradient boosting and self-training) to analyse these reviews. These methods allow us to uncover behavioural patterns and explore the key product characteristics influencing OPRB. Findings We find that intrinsic attributes (performance and reliability), extrinsic attributes (warranty), high price and sales rank are key drivers of assertive OPRB. Durability has a heterogeneous effect, while low price and information mismatch are linked to non-assertive OPRB. Additionally, assertive OPRB can be triggered by joint effects between two product attributes. Practical implications The findings provide manufacturers with insights to prioritise quality issues in design and production, while e-commerce managers and operations professionals can manage returns more strategically and better address customer dissatisfaction. Originality/value This study contributes to attribution, prospect and planned behaviour theories by explaining how intrinsic and extrinsic attributes cause differences in assertive vs non-assertive OPRB, emphasising the role of customer feedback in product development and operational optimisation.