Analysing software quality attributes and user satisfaction in mobile banking applications

Purpose This study addresses a knowledge gap by showing how perceived software quality dimensions extracted from user-generated reviews influence user satisfaction in mobile banking applications. The paper extends existing theory by demonstrating how sentiment polarity moderates the relationship between perceived software quality dimensions and user ratings across both fully digital and hybrid digital–physical banks. Design/methodology/approach A large corpus of online reviews was analysed using a hybrid text-mining approach that integrates latent dirichlet allocation (LDA), sentiment analysis, large language model–assisted topic interpretation and sentence-BERT (SBERT) semantic similarity modelling. User-derived attributes were mapped to the software quality dimensions defined in ISO/IEC 25010:2023, resulting in a framework that operationalises perceived software quality. Regression analysis estimated their functional and emotional effects on user ratings. Findings Functional suitability, interaction capability and maintainability positively impact user satisfaction, while performance efficiency, reliability, flexibility, safety and security issues primarily drive dissatisfaction. Compatibility shows no significant effect. Sentiment polarity moderates these associations: under positive sentiment, negative quality effects intensify, while the moderation of positive effects varies across dimensions. Originality/value A novel framework based on ISO/IEC 25010:2023 is used to assess the perceived service quality of mobile banking applications for digital-only and banks with physical branches, by analysing user sentiment in online reviews. Further, the work examines the effect of perceived software quality on user satisfaction, specifically how this varies with sentiment polarity. The relationship between software quality and user satisfaction is found to be asymmetric and conditioned by review sentiment. Under positive sentiment, negatively associated quality dimensions exert stronger adverse effects on satisfaction, while positive effects vary across dimensions. Quality-satisfaction estimates must account for this asymmetric sentiment moderation. Two novel visualisation tools capture how quality dimensions differ in their impact, prominence and sentiment in shaping user satisfaction.