Enhancing E-commerce Chatbot Design with Customer Journey Mapping: Survey and NLP Analysis of User Inquiries

Despite adoption of chatbots in e-commerce, chatbots struggle to maintain context, provide complete policy guidance, or respond appropriately to post-purchase problems. This study maps user questions as shoppers’ needs at different points in the Customer Journey Map (CJM). A cross-sectional survey collected users’ opinions on expected chatbot interactions in an e-commerce setting. Sixty participants provided responses on which three questions they would ask an e-commerce chatbot during online shopping. Three unsupervised machine-learning approaches were used: Latent Dirichlet allocation (LDA) topic modeling, K -means clustering, and Lexicon-based analysis. We compared themes and clusters from the Lexicon and K -means methods with the topics generated by LDA. Mapping inquiries to the CJM showed the largest share in Consideration (35%), followed by Decision (30%), Awareness (20%), and Post-Purchase (15%). Overall sentiment of inquiries was observed to be predominantly neutral, Awareness questions had a largely positive tone, and Post-Purchase questions had a largely negative tone.