A REVIEW OF MODELS FOR PREDICTING CUSTOMER SATISFACTION

Evaluation, modelling, and prediction of client satisfaction are crucial business strategies that address customer dissatisfaction, promote retention, and enhance organisational reputation. However, many organisations focus on acquiring new customers, a costly endeavour, often neglecting retention, leading to potential business failure. Those few organisations that do prioritise customer retention sometimes rely on inadequate models for predicting satisfaction, resulting in non-representative outcomes and missed opportunities to prevent client dissatisfaction. This study systematically reviews the literature on various predictive models of client satisfaction, employing the PRISMA methodology, which facilitates a rigorous process of review. Through a structured selection, 100 articles were critically examined, encompassing diverse sectors such as mobile banking, e-governance, healthcare, etc. Ultimately, 80 high-quality studies were analysed, highlighting the methods used, their effectiveness, and limitations. The results reveal that Structural Equation Modelling (SEM) is the most frequently employed method, accounting for 42% of reviewed studies, due to its robustness in modelling latent variables and causal relationships. Machine learning techniques (MLT), Artificial Neural Networks (ANNs), represent 18% of the studies and are recognised for their ability to capture complex, non-linear relationships. Hybrid SEM-ANN models, which combine the strengths of both methods, were featured in 10% of the publications, showcasing superior predictive performance. Conversely, classical statistical approaches, such as multiple regression and cluster analysis, are less common and used in 4% of cases, due to their inability to effectively handle intricate data structures. Furthermore, SEM with Deep Learning, advanced data mining, and predictive precision approach is gaining attention for its potential to provide deeper insights and more adaptive solutions in understanding client satisfaction. The review identifies the hybrid SEM-ANN approach and Deep Learning models as the most effective frameworks for predicting client satisfaction. This innovative integration provides a comprehensive and practical approach for organisations aiming to enhance client satisfaction and achieve sustainable growth.