Sea ice thickness plays an important role in the climate system, yet it remains one of the least constrained variables in climate models due to limited spatiotemporal coverage of observations. Satellite missions such as ICESat-2 and CryoSat-2 have provided long records of basin-scale freeboard observations in both hemispheres, enabling estimates of sea ice thickness. Here, we assess uncertainties in freeboard retrievals from these two missions and evaluate their potential for improving the sea ice freeboard representation in the state estimates produced by the project for Estimating the Circulation and Climate of the Ocean (ECCO). From comparisons of ICESat-2, CryoSat-2 and a preliminary ECCO version 5 forward model simulation, we find general consistency between the two observational products but substantial differences between modeled and observed freeboard across time-mean, seasonal, and monthly timescales. Biases in total freeboard (ice freeboard plus snow depth) are primarily attributed to the misrepresentation of the ice freeboard in the model. Time-mean model-observation misfits are typically more than 4 times larger than estimated data errors over the majority of the polar regions, while monthly misfits exceed data errors by a factor of ∼2. These results indicate that assimilating both total and ice freeboard observations could substantially improve the sea ice representation in ECCO state estimates across all timescales.