The rapid advancement of spatial data acquisition technologies in the past decade has enabled the efficient capture of high-resolution, high-accuracy point clouds in a non-invasive manner. These datasets have facilitated research across multiple disciplines, particularly in the analysis of complex architectural heritage structures. This study focuses on the geometric analysis of formal anomalies in historical buildings, specifically examining the webs of ribbed vaults. Despite their straightforward conceptual framework, the geometric parameterization of webs remains a complex task. The research evaluates two large vaults: one from the main nave of the Cathedral of Santa Maria de Girona, characterized by a straight-profiled ribbing, and another from the central nave of the Cathedral of Santa Maria de Palma de Mallorca, exhibiting a curved profile. A methodological workflow is proposed to analyse the shape of webs by comparing point cloud data against a reference shape. Given the irregular and large-scale nature of these vaults, standard automatic or semi-automatic modelling approaches introduce errors. Thus, two strategies for reference shape determination are compared: manual modelling based on webs’ profiles and automated modelling using robust primitive shape extraction techniques. The results have revealed that the best approximation is conditioned to the profile of the ridged rib, which can be curved or straight, and it determines the shape of the webs. The findings contribute to optimizing workflows for architectural heritage analysis, enhancing the accuracy and efficiency of vault geometry assessment.