In this paper we examine the influence of the granularity of a rating system on the resulting regulatory capital requirements with respect to the framework released by the Basel Committee on Banking Supervision (Basel II). We define granularity as the number of cohorts or rating classes used to measure default rates and the relative size of these cohorts, respectively. We show that due to the convex relation between capital requirements and estimated probabilities-of-default (PD) capital requirements decline with the fineness of the rating system and, moreover, that for a given number of rating classes there exists an 'optimal' granularity in terms of minimal capital requirements. We provide several numerical and empirical results to assess potential effects for rating systems of intermediate granularity (as commonly used by banks) and 'naive' schemes to construct the relative cohort sizes. As a benchmark result we obtain a 12 bp (45 bp) reduction in capital requirements when switching from five to ten rating classes and using the optimal ('naive') relative cohort sizes. These findings are confirmed by an analysis based on empirical data about more than 11,000 Austrian firms provided by Creditreform.