Abstract In this study, an effective parametric 3D geometry model of a propeller was established with the aid of reverse engineering. The goal is to reduce the free parameters while automating the modelling of the propeller. The process of building the parametric model begins by generating an initial point cloud by defining the feature matrix associated with the propeller blade profile shape. Subsequently, the initial point cloud is deformed and refined by the deformation feature matrix and resampling. Finally, a 3D geometry model of the propeller is generated by surface reconstruction. The model can be built automatically by interactively modifying the feature matrices. Two numerical analyses illustrate the performance of the parametric 3D geometry model. Specifically, two propellers are constructed using the proposed model to estimate the shape error between the reconstructed propellers and the original offset of the propellers. These propellers are selected as research objects to determine the hydrodynamic performance error between the propeller constructed by the proposed model and a benchmark propeller. According to the results of the numerical study, the parametric 3D geometry model can precisely reconstruct the aforementioned geometry within a valid error range. The hydrodynamic error analysis demonstrates that the geometric inaccuracy from the reconstructed model has less impact on the propeller performance. This indicates that the model described in this study is generalised and robust. Moreover, some uncommon propeller CAD models were generated in batches using the parametric 3D geometry model.