With the ever-increasing aerodynamic and thermal loads, the endwalls of modern gas turbines have become critical areas that are susceptible to manufacturing and operational uncertainties, making them highly prone to thermal failures. Therefore, it is of vital importance to quantify the impacts of input uncertainties on the aero-thermal performance of endwalls. Firstly, based on the Kriging surrogate, an efficient uncertainty quantification (UQ) method suitable for expensive CFD problems is proposed. Using this method, the impacts of slot geometry deviations (slot width, endwall misalignment) and mainstream condition fluctuations (turbulence intensity, inlet flow angle) on the aero-thermal performance of endwalls are quantified. Results show that the actual performance of endwalls has a high probability of deviating from its nominal value. The maximum deviations of aerodynamic loss, area-averaged film cooling effectiveness, and area-averaged Nusselt number reach 0.33%, 45%, and 5.0%, respectively. The critical regions that are most sensitive to the input uncertainties are also identified. Secondly, a global sensitivity analysis method is also performed to pick out the significant uncertain parameters and explore the relationship between input uncertainties and performance output. The inlet flow angle is proved to be the most significant parameter among the four input uncertain parameters. Besides, a positive incidence angle is found to be detrimental to both the aerodynamic performance and the thermal management of endwalls. At last, the influence mechanisms of the inlet flow angle on endwall aero-thermal performance are clarified by a fundamental analysis of flow and thermal fields.