Reframing User Acceptance with User-Generated Content: Insights from Digital Banking in Indonesia

Abstract

This study examines user acceptance of BNI’s newly launched digital banking application, WONDR, by leveraging user-generated content (UGC) obtained from the Google Play Store. A total of 60,434 user reviews were collected and analyzed using a supervised machine learning approach, specifically the Random Forest algorithm, to classify sentiments into positive and negative categories. The dataset underwent a rigorous preprocessing pipeline, including text normalization, tokenization, stopword removal, and stemming, followed by feature extraction using the Term Frequency–Inverse Document Frequency (TF–IDF) method. Model performance was evaluated through a 7-Fold Cross-Validation strategy and standard metrics such as accuracy, precision, recall, and F1-score, achieving strong and consistent results with an overall accuracy of 0.888. The findings indicate that UGC-based sentiment analysis provides a scalable and interpretable method for assessing user acceptance, offering actionable insights for strategic application development. Theoretically, this study extends traditional acceptance models (e.g., TAM and UTAUT) by operationalizing user perceptions through naturally occurring feedback. Practically, the proposed approach supports data-driven decision-making for enhancing digital service quality in the banking sector.