The macropore network of a hierarchically structured mesoporous hydrodemetalation catalyst is mathematically optimized to maximize the catalytic activity and robustness to deactivation over a given time on stream. A Random Spheres Model (RSM) accounting for catalyst deactivation in both the meso- and macropores is used to describe transport and reaction in a catalyst pellet. The useful lifetime can be nearly doubled at the pellet scale, and increased by 40% at the reactor scale, while using 29% less catalyst compared to a non-optimized purely mesoporous catalyst. An optimized hierarchical structure is also superior to an optimized purely mesoporous catalyst. Introducing an optimized distribution of macroporosities and broad pore sizes does not lead to a significantly higher catalytic activity or lifetime over an optimized structure with a single macroporosity and a single broad pore size. Finally, the predicted optimum broad pore network properties are compared with relevant data from the patent literature. (C) 2011 Elsevier Ltd. All rights reserved.