This paper compares predictions from a 3D Reynolds-averaged Navier–Stokes code and a statistical representation of measurements from a cooled 1-1/2 stage high-pressure transonic turbine to quantify predictive process sensitivity. A multivariable regression technique was applied to both the inlet temperature measurements obtained at the inlet rake, the wall temperature, and heat transfer measurements obtained via heat-flux gauges on the blade airfoil surfaces. By using the statistically modeled temperature profiles to generate the inlet boundary conditions for the computational fluid dynamics analysis, the sensitivity of blade heat transfer predictions due to the variation in the inlet temperature profile and uncertainty in wall temperature measurements and surface roughness is calculated. All predictions are performed with and without cooling. Heat transfer predictions match reasonably well with the statistical representation of the data, both with and without cooling. Predictive precision for this study is driven primarily by inlet profile uncertainty followed by surface roughness and gauge position uncertainty.