The spatial distribution of solid components and porosity within a composite granuleits microstructure-is an important attribute as it carries information about the processing history of the granule and determines its end-use application properties, particularly the dissolution rate. In this work, the problem of rational design of granule structure is formulated, and two methods for its solution are proposed-stochastic design, which is based on random permutation of points within the structure using the simulated annealing algorithm, and variational design, which is based on direct simulation of granule formation from its constituent primary particles, followed by direct simulation of granule dissolution. The variational design method is demonstrated in a case study of the effect of primary particle size, radial distribution of components, and composition of a two component granule (active, excipient) on the dissolution profile. Selected granule structures designed computationally were also physically made by fluid-bed granulation, their structure analyzed by X-ray micro-tomography, and dissolution curves measured. It was confirmed that the designed structures are feasible to manufacture and that they meet the required dissolution profiles. (c) 2006 American Institute of Chemical Engineers AIChE J, 52: 3762-3774, 2006.