An innovative analytical model for predicting charge weld length in continuous extrusion is presented, with the goal of minimizing scrap and ensuring the structural performance of extruded profiles. By relying only on die and profile geometry, the approach offers major advantages over computationally expensive FEM simulations and experimental methods in terms of time and resources, while improving prediction accuracy compared with previous analytical formulations. The model covers a wider range of extrusion conditions and is applicable to complex industrially relevant profiles. In addition to parameters considered in earlier models, such as die volume, exit area and number of ports, the proposed method accounts for inlet area, feeding chamber area and angle, port area, exit area and profile wall thickness, providing a more complete description of material flow. A dataset from 12 experimentally analyzed extruded profiles was first acquired, then expanded through FEM simulations of 84 additional cases. Model coefficients were obtained via multivariate regression, achieving a mean absolute percentage error below 9%, significantly lower than previous analytical approaches. Based solely on geometric descriptors available early in the design stage, the model offers a fast and reliable tool for reducing charge weld length and supporting geometry-driven die development.