To elevate the separation performance, two-dimensional liquid chromatography (2D-LC) uses two chromatographic columns with different stationary phases to diversify solute interactions with the resin, hence providing a second "dimension" to solute-specific separation. Developing methods for 2D-LC starts therefore with preliminary column selection. Selecting columns that yield (metaphorically) orthogonal dimensions is of utmost importance, but remains challenging. Although several metrics exist to quantify column orthogonality, currently there is no established methodology, and none of the existing methods accounts for the non-homogeneity of peak band broadening across each separation dimension. In this work, we propose a new approach to select columns a priori. This approach is based on critical resolution distribution statistics and implicitly accounts for local peak crowding and peak band broadening. Furthermore, we assess the importance of preliminary column selection during in-silico method development and multi-objective optimization of comprehensive 2D-LC. The comparison of the multi-objective Pareto fronts revealed that column pairs selected with our approach provide better separation quality and reduce analysis time compared to column selections via the most established metrics in the literature. Our results prove the importance of preliminary column selection for method development and optimization of 2D-LC systems, and they also show that choosing the right orthogonality metric (such as that proposed here) is crucial.