Combining textual and rating data from online reviews to cluster consumers

In today’s competitive global market, businesses seek a profound understanding of consumer behavior to gain a competitive edge. Leveraging Big Data and Web 2.0, companies can efficiently gather vast amounts of timely online information at minimal cost. Online review platforms like TripAdvisor ask users to provide textual reviews and ratings regarding the overall product/service and its key aspects. The challenge is identifying the best tools to extract as much reliable information as possible combining all data available. This paper proposes the Clustering Online Evaluations using Rating, Topic, and Sentiment (COERTS) method, allowing businesses to gain comprehensive insights into customer satisfaction by analyzing textual and rating data within a cluster framework. By identifying groups of similar clients, businesses can extract valuable information about customer experiences, enabling targeted improvements for unsatisfied consumers. Techniques such as sentiment analysis, emotional analysis, and the Latent Dirichlet Allocation model preprocess textual data. Additionally, Likert-type variables representing overall ratings and specific aspects are fuzzified to address response uncertainty and heterogeneity. After preprocessing, a suitable clustering algorithm organizes the data effectively. The presented procedure is theoretically explained and illustrated through a case study, showcasing its advantages in enhancing customer satisfaction, increasing retention, and re-attracting former customers.