Micromechanical models of 3-3 piezocomposites have previously relied on an idealised unit cell containing a single pore. This oversimplification has contributed to the poor agreement between experimental and model results. With increased computational power and more efficient modelling techniques, larger models can be created to represent a network structure with numerous randomly placed pores, filled with a secondary passive phase of air or polymer. Such models more closely represent a porous microstructure and enable the prediction of hydrostaticfigures of merit. A series of large three-dimensional porous piezocomposite structures were modelled, with predicted values in good agreement with existing experimental results. In addition, modelling a large number of random porous microstructures revealed the variability in composite properties, as is found in practise.