After the 2009 L’Aquila earthquake, a procedure was developed for assigning reconstruction grants to buildings with usability ratings of B, C, or E following the outcome of the post-earthquake usability assessment (AeDES forms). Although the procedure is complex due to the need to account for various scenarios, the calculation of reconstruction costs depends on two main factors: the observed damage and the assessed vulnerability level. For masonry buildings, which make up more than 90% of structures in smaller centres, vulnerability is determined by combining nine contributions that represent key factors influencing the overall vulnerability assessment. This work investigates whether a classification model with satisfactory performance metrics can be developed to predict cost classes using only synthetic vulnerability data. The goal is to calibrate a model that can reasonably estimate reconstruction costs using only vulnerability features, which can be assessed even without damage information following a seismic event. Such models are crucial for calculating earthquake insurance premiums, which are expected to depend on a building’s vulnerability.