In this work, we present theoretical and experimental results of bicomponent granulation. We formulate a population balance approach to describe a bicomponent population of particles that contain an ingredient of interest ("solute") mixed with an inert compound ("excipient"). We classify granulation kernels into three categories: (i) kernels such that dissimilar components have higher rate of agglomeration compared to similar components; (ii) kernels such that dissimilar components have lower rate of agglomeration; and (iii) kernels that make no distinction between the components. Case (iii), which we refer to as "ideal" aggregation, leads to blending of components into a distribution that is Gaussian in the mass fraction of the solute. The mean composition of granules approaches quickly the desired composition and the variance of the distribution decreases inversely with granule size. Kernels of type (i) lead to quick blending of components. Kernels of the type (ii) inhibit blending and lead to a population of granules that is relatively segregated. Eventually, however, these kernels also produce Gaussian distributions but at a much slower rate compared to kernels (i) and (iii). The results of these simulations are compared with experimental granulation studies using mixtures of two excipients. We find that depending on the surface properties of the excipient, these systems exhibit behavior similar to that of kernels of type (ii) and (iii).