Uncertainty quantification (UQ) is essential for evaluating the impact of manufacturing-induced geometric variability on the aerodynamic performance of jet engine compressor blades. Conventional UQ approaches mainly include eigenmode-based methods, which use principal component analysis to extract geometric eigenmodes from measured shape data, and design parameter-based methods, which assess the influence of predefined design parameters based on a shape model. While design parameter-based methods offer interpretability for designers due to the use of design parameters, these methods require a shape model and may cause modeling error from actual manufactured blades. In contrast, eigenmode-based methods maintain high fidelity with measured data but lack clear links to design parameters, limiting their interpretability for designers. To bridge the gap between these approaches, we propose a data-driven methodology that connects aerodynamic performance metrics and design parameters, both directly computed from measured blade geometries, and provide interpretable sensitivity analysis with respect to design parameters. This framework enables the use of performance metrics computed from real blade geometries while retaining interpretability through design parameters. Furthermore, we apply a generalized sensitivity index that accounts for correlations among input variables—often neglected in conventional sensitivity analyses—to evaluate the influence of each correlated design parameter. Our results show that certain design parameters have a dominant effect on performance uncertainty. This is consistent with expert qualitative assessments, which suggest the reliability of the proposed method. This approach provides a practical and interpretable framework for sensitivity analysis of the uncertainty by integrating the advantages of both eigenmode-based and design parameter-based UQ methods.