Accurate sales forecasting plays a crucial role in improving operational efficiencies in sales and distribution using Enterprise Resource Planning (ERP) systems in the ever-changing retail management industry. Although there have been significant improvements in predictive analytics, current models frequently struggle with a compromise between accuracy and interpretability, which is essential for making strategic decisions. Addressing these challenges, our study introduces a refined predictive modeling approach that employs advanced regularization techniques-Ridge and Lasso regression-within a linear regression framework, combined with systematic hyperparameter tuning. This methodology enhances the model’s ability to generalize across diverse datasets while maintaining simplicity and interpretability, making it more adaptable for real-world applications. The results underscore the effectiveness of our approach. Initially, the basic linear regression model showed limited predictive accuracy, with an RMSE of 529,338.33 and an R-squared of 14.76%. Implementing hyperparameter tuning significantly improved these metrics, reducing the RMSE to 345,386.93 and raising the R-squared to 63.71%. The final model, incorporating Ridge and Lasso regression techniques, achieved a dramatically lower RMSE of approximately 22,150.78 and an R-squared value of 97.44%, demonstrating near-perfect predictive performance. Integrating regularization techniques with hyperparameter tuning represents a significant advancement in balancing model complexity with interpretability, offering valuable insights for academic researchers and retail managers. This research makes substantial contributions to sales forecasting in retail, providing a robust framework that advances the predictive capabilities of ERP systems and ensures that the outputs are user-friendly and actionable.