Booking-to-redemption barriers in tourism E-commerce live streaming: explanation-prediction via multimodal data fusion

Order cancellation is prevalent in tourism e-commerce live streaming and significantly undermines marketing effectiveness. After making impulsive purchases during live streams, users often revisit external product information to review descriptions and user comments, which is reported to be a crucial stage that leads to revised decisions and order cancellations. From an information consistency perspective, and grounded in Expectation Confirmation Theory and the Working Memory Model, this study employs regression analysis and interpretable machine learning to investigate how textual, visual, and integrated consistency across modalities influence order cancellation behavior. The results indicate that textual consistency and integrated consistency exert significant positive effects on order cancellation behavior, whereas visual consistency exerts a significant negative effect. The machine learning predictive models further confirm that consistency features play a critical role in identifying order cancellation risk. Overall, this study reveals the differentiated mechanisms through which information consistency across modalities shapes post-purchase behavior in the tourism e-commerce live streaming.