The article presents a data-driven approach to predict the success of bank telemarketing calls for selling long-term deposits. The study uses data from a Portuguese retail bank collected between 2008 and 2013, including 52,944 phone contacts. The dataset contains 150 features related to bank clients, products, and social-economic attributes. A semi-automated feature selection process was applied, reducing the number of features to 22. Four data mining models—logistic regression (LR), decision trees (DT), neural networks (NN), and support vector machines (SVM)—were compared using two metrics: area under the receiver operating characteristic curve (AUC) and area under the lift cumulative curve (ALIFT). The NN model achieved the best results, with an AUC of 0.8 and ALIFT of 0.7, allowing the identification of 79% of successful contacts by selecting the top half of classified clients. Knowledge extraction methods, including sensitivity analysis and DT, were applied to the NN model, revealing key attributes such as Euribor rate, call direction, and agent experience. These findings confirmed the model's credibility and value for telemarketing campaign managers. The study emphasizes the importance of feature engineering and the use of realistic evaluation methods, such as rolling windows, to improve predictive accuracy. The NN model outperformed other models in both AUC and ALIFT metrics, demonstrating its effectiveness in predicting telemarketing success. The results highlight the potential of data mining techniques in enhancing decision-making processes in banking, particularly in targeting clients for marketing campaigns. The study also underscores the need for robust validation and the importance of considering temporal dimensions in predictive modeling. Overall, the research contributes to the field of decision support systems by providing a practical and effective approach to predicting the success of bank telemarketing calls.The article presents a data-driven approach to predict the success of bank telemarketing calls for selling long-term deposits. The study uses data from a Portuguese retail bank collected between 2008 and 2013, including 52,944 phone contacts. The dataset contains 150 features related to bank clients, products, and social-economic attributes. A semi-automated feature selection process was applied, reducing the number of features to 22. Four data mining models—logistic regression (LR), decision trees (DT), neural networks (NN), and support vector machines (SVM)—were compared using two metrics: area under the receiver operating characteristic curve (AUC) and area under the lift cumulative curve (ALIFT). The NN model achieved the best results, with an AUC of 0.8 and ALIFT of 0.7, allowing the identification of 79% of successful contacts by selecting the top half of classified clients. Knowledge extraction methods, including sensitivity analysis and DT, were applied to the NN model, revealing key attributes such as Euribor rate, call direction, and agent experience. These findings confirmed the model's credibility and value for telemarketing campaign managers. The study emphasizes the importance of feature engineering and the use of realistic evaluation methods, such as rolling windows, to improve predictive accuracy. The NN model outperformed other models in both AUC and ALIFT metrics, demonstrating its effectiveness in predicting telemarketing success. The results highlight the potential of data mining techniques in enhancing decision-making processes in banking, particularly in targeting clients for marketing campaigns. The study also underscores the need for robust validation and the importance of considering temporal dimensions in predictive modeling. Overall, the research contributes to the field of decision support systems by providing a practical and effective approach to predicting the success of bank telemarketing calls.