<![CDATA[User retention is a critical factor in the sustainability of social commerce platforms; however, the relationship between usability and retention remains inconsistent across various studies. This study aims to analyze the effect of feature-level usability on user retention in TikTok Shop among Generation Z users, as well as to identify retention patterns using a predictive approach. The study adopts a quantitative approach with an explanatory–predictive research design. Data were collected from 115 respondents through a questionnaire based on the System Usability Scale (SUS), focusing on the usability of live shopping and product review features. The analysis was conducted using binary logistic regression and the Naïve Bayes algorithm. The results show that both features fall into the marginal-low usability category, with scores below the acceptable threshold. However, the logistic regression results indicate that usability does not have a significant effect on user retention. On the other hand, the Naïve Bayes model achieved an accuracy of 88.64%, with a macro-average F1-score of 58.71%, but showed limitations in detecting non-retained users, as indicated by a minority-class recall of 14.29% due to class imbalance. These findings suggest that feature-level usability has limited explanatory power in predicting user retention within the social commerce context. This study contributes by reinforcing the argument that user retention is not solely determined by usability and demonstrates that predictive approaches can provide additional insights that are not captured by explanatory models.]]>
