Uncovering Customer Issues in E-Commerce: Sentiment Analysis and Topic Modeling Approach

The rise of digital technology has revolutionized consumer behavior, leading to a notable transition toward online shopping. Ensuring the provision of advantages and satisfaction to customers is crucial, along with the competitive nature of the e-commerce industry. The emergence of digital media, particularly social media platforms, provides a channel for customers to express their experiences, encompassing expressions of praise and complaints regarding perceived dissatisfaction with the services offered. Parties are presented with valuable opportunities to actively delve into issues that contribute to user dissatisfaction through this open media and support the targeted data-driven considerations for decision-making. In this study, we use text mining methods to uncover user issues based on textual customer traces on Twitter. We use IndoBERT sentiment classification to define the satisfaction polarity. This research study centers around the examination of negative expressions. Subsequently, a two-stage topic modeling methodology is employed to investigate and identify the primary issues underlying customer dissatisfaction. We uncover five primary issues in customer dissatisfaction, including application issues (25%), delivery issues (21%), express courier issues (18%), cost-related issues (17%), and customer service issues (17%). We comprehensively examine sub-topics on each primary issue to understand the underlying problems better. This method offers a more profound approach to uncovering customer satisfaction issues and supporting targeted strategies in enhancing customer experience by responding to emerging customer satisfaction issues.