When deriving the sea ice thickness using a satellite radar altimeter, a correction is needed to convert the radar freeboard to the ice freeboard. Here, a robust correction method for the CryoSat-2 radar freeboard derived using the Threshold First-Maximum Retracker Algorithm with a 50% threshold (TFMRA50) is proposed. Snow depth was used as a predictor for the correction, similar to the traditional wave speed correction, but the coefficients were empirically determined by performing a direct comparison of the radar freeboard from CryoSat-2 and the ice freeboard from airborne observations. Consequently, this new empirical correction treats the physical effects (e.g., changes in wave propagation speed and the distribution of scattering at the snow and ice surfaces, etc.) and the retracker bias as a whole, which have been difficult to separate in the retrieval process. In this paper, we demonstrate that the retrieval accuracy of snow and ice variables and the consistency of the two independent retrieval methods are improved when the new correction is applied. The result of this study emphasizes the importance of compatibility between the retracker and the freeboard correction method.