A proper orthogonal decomposition (POD)-based method is proposed to reduce the dimensions of the design space for the shape optimisation of marine propellers. The effectiveness of the proposed approach is proven in the case of the INSEAN-E779A propeller, which blade shape is modified to maximise efficiency while reducing suction side cavitation. The 23-dimensions design space defined by the conventional shape representation is reduced by the POD method to 5, 12 and 15 dimensions, retaining up to the 98% of the geometric variance of the original space. A multi-objective optimisation algorithm drives the simulation-based design optimisation (SBDO) process in the new design spaces using BEM for the hydrodynamic predictions. Finally, optimal designs are verified using RANSE to assess the correlation between the performance improvements, the dimensionality reduction and the corresponding geometric variance. The effectiveness of the proposed POD-SBDO framework is discussed with respect to a design by optimisation process relying on the conventional parametric representation of the blade geometry.