Decoding E-commerce repurchase signals: a mixed-method analysis of large-scale online review data

Purpose This study aims to investigate how product attributes and emotional cues in online reviews jointly influence consumers’ repeat purchase intentions, and how these effects differ between hedonic and utilitarian products. While prior research has largely examined attributes or emotions in isolation, this study clarifies their interactive mechanisms across product types in e-commerce contexts. Design/methodology/approach Using 976,483 verified Amazon reviews from six product categories, the study adopts a two-stage research design. Study 1 applies latent Dirichlet allocation, lexicon-based emotion measurement and random forest models with Shapley additive explanations analysis to identify five core product attributes: performance, functionality, physical appearance, value for money and customer service. Study 2 uses logistic regression models with interaction terms to examine how emotional valence and arousal moderate attribute effects across hedonic and utilitarian products. Findings The results reveal pronounced asymmetries in e-commerce contexts. For hedonic products, all five attributes positively influence repeat purchase intentions; emotional arousal amplifies these effects, while emotional valence weakens some attribute impacts. For utilitarian products, only value for money shows a stable positive effect on e-commerce repeat purchases, whereas other attributes exert negative effects. Emotional valence uniformly dampens attribute influences, while arousal selectively strengthens the effects of performance, value for money and physical appearance. Originality/value By integrating the valence–arousal–dominance emotion model with hedonic–utilitarian product theory, this study offers a unified framework explaining how multi-attribute information and emotional cues jointly shape online repeat purchase behavior. The two-stage combination of machine learning and econometric analysis provides both predictive and explanatory insights for online review research in the field of e-commerce.