<![CDATA[The development of digital gold-based financial services in Indonesia has shown significant growth; however, the quality of user experience remains a critical factor in ensuring long-term application usage. Although thousands of user reviews are available across various digital platforms, this information has not been fully utilized as a foundation for service improvement. This study aims to conduct an in-depth analysis of user sentiment and generate application development recommendations based on real user perceptions and experiences. The methodology includes collecting a large number of user reviews, performing text preprocessing, and applying sentiment classification using several machine learning models. IndoBERT demonstrated the best performance with an accuracy of 0.88, precision of 0.84, recall of 0.88, and F1-score of 0.86 based on 7,113 test data, indicating strong capability in identifying positive and negative sentiment, although neutral sentiment remains challenging. The analysis reveals that positive sentiment is associated with the ease of investing in gold, intuitive interface design, and overall user convenience. Conversely, negative sentiment is predominantly linked to technical issues such as transaction errors, login disruptions, slow authentication, and system instability. The WordCloud visualization also highlights the dominance of terms such as error, login, and verification. Based on these sentiment patterns, the study proposes several priority recommendations, including enhancing system stability, optimizing authentication mechanisms, refining transaction flows, and strengthening customer support. This study provides a structured mapping of user issues that can serve as a strategic foundation for developing more reliable, secure, and user-aligned digital gold financial applications.]]>
