A computational characterization technique is presented for assessing the spatial heterogeneity of two reactant phases in a three-phase chemically reactive composite. This technique estimates the reaction yield on multiple microstructure length scales based on the segregation of the two reactant phases and the expected reaction stoichiometry. The result of this technique is a metric, quantifying the effectiveness of phase mixing in a particular microstructure as a function of length scale. Assuming that the proportionate mixing of reactant phases on multiple length scales will enhance reaction kinetics and the overall level of reaction completion, this tool can subsequently be used as a figure-of-merit for optimizing microstructure via appropriate processing. To illustrate this point, an example is shown where a bimodal three-phase microstructure has a higher reaction yield at every length scale when compared with a monomodal three-phase microstructure with the same constituent loading.