Recent advancements in Earth science and geodetic applications increasingly demand geoid models with sub-centimetre internal accuracy. Previous studies have demonstrated that the internal accuracy of a geoid model depends mostly on the spatial distribution of gravity measurements rather than their individual accuracy. Regional geoid determination typically relies on a combination of land-level gravity data from terrestrial, marine, and airborne platforms with synthetic observations from global gravity field models. Spatial gaps in the discrete gravity observations remain a major limiting factor in achieving the required level of accuracy. This study aims to determine the maximum acceptable gravity data gap that can be tolerated while still maintaining sub-centimetre internal geoid accuracy, relative to the precision of the gravity measurements themselves. A pre-analysis using the Stokes-Helmert geoid error estimation approach will be performed, employing established formulations for formal error propagation from gravity observations, global gravity models, and digital elevation models. The results will be applied to assess existing land and marine gravity data coverage, focusing initially on a sample case from Canada. The main deliverables of this study include quantitative recommendations for acceptable gravity data gaps necessary to achieve sub-centimetre geoid accuracy, as well as a map characterizing the spatial distribution and density of available gravity data.