<p>In industrial sheet metal production, variations in material properties across different batches often lead to significant differences in the subsequent forming behavior and overall product quality. These fluctuations of material properties pose challenges for process robustness and require strategies to ensure reliable forming results. Despite the existence of numerous methodologies for the expeditious characterization of the forming behavior and forming properties of sheet metals, in addition to standardized characterization methods, these techniques have not yet achieved widespread industrial implementation. Most existing methods rely on data-driven models, or surrogate (meta-) models, to predict forming behavior and material response. However, generating representative and sufficiently diverse training data remains a major obstacle to the practical use of such models. This study introduces a new methodology for generating high-quality training data for data-based prediction models, combining experimental testing, numerical simulations, and machine learning. The methodology is developed using the ‘IWU-Werkstofftester’, which is a compact and flexible testing system designed to rapidly evaluate the properties of sheet metal. The focus is on identifying and validating suitable substitute tests, defining an appropriate finite element (FE) model and introducing variability into the input parameters systematically to ensure sufficient data diversity. Furthermore, the resulting datasets are evaluated in terms of quality, representativeness and the effort required for data generation. Finally, the study investigates the predictive performance of an exemplary machine learning model. The results emphasize the potential of combining targeted experiments and simulations to speed up the development of predictive models for sheet metal forming, thereby supporting the industrial adoption of data-driven material characterization methods.</p>