Using Text Mining to Improve Service Quality Effort: a Case on Indonesia Beauty E-Commerce

Abstract

This study aimed to investigate the quality performance of beauty e-commerce services in Indonesia to find some appropriate improvements according to customer perceptions. Identification is conducted based on electronic service quality dimensions using text mining approaches, namely multiclass classification, sentiment analysis, and text network analysis methods. We gathered data from Twitter as a form of user generated content, because it is essential for companies to perceive what customers feel and need. This study utilized Sociolla, the leading beauty e-commerce in Indonesia, as an object. The results indicate several factors considered necessary by customers, i.e., efficiency, system availability, fulfillment, and responsiveness obtained from classifying customer opinions using the Naive Bayes Classifier model with an accuracy rate of 89%. The study also discovered many complaints regarding incompatible goods stock information on the website, app crash problems, difficulties in tracking orders, and difficulties in selecting payment methods. Moreover, several services are to be maintained, including faster delivery, more discounts, promos, and giveaway events, as well as quick response from customer service in handling complaints.