Similar to natural photosynthesis, Z-scheme photocatalytic water splitting relies on two different light absorbing components that are coupled by a redox active mediator that shuttles charge between them. Such a two-absorber system possesses several advantages over single absorber photocatalytic system, including higher theoretical solar-to-hydrogen conversion efficiency, relaxed band alignment requirements, and the potential for inherently safe operation whereby H2 and O2 evolution occur in separated compartments. However, a major disadvantage and challenge for Z-scheme photocatalysis is that the presence of a redox mediator introduces two undesirable back-reactions on top of parasitic H2 oxidation and O2 reduction reactions that can occur in a single absorber photocatalytic system. Previous research efforts have identified the use of semi-permeable oxide coatings as an attractive approach to suppress these thermodynamically favored redox reactions while still permitting the desired water splitting and mediator redox reactions to occur. Here, we present a combined experimental and computational approach based on model thin films that is used to (i) probe the performance limits of oxide-encapsulated photocatalysts, (ii) quantify the effects of coating defects on performance, and (iii) guide the rational design of coatings aimed at maximizing the solar-to-hydrogen conversion efficiency of a target photocatalytic system. This work specifically focusses on the development of silicon and titanium oxide coatings for Z-scheme water splitting based on a Fe(II)/Fe(III) mediator, showing that the best coatings can achieve selectivities > 90 % towards the H2 and O2 evolution reactions over undesired Fe(II)/Fe(III) back reactions. Another key finding from this work is that coating defects can have a significant influence on the performance of encapsulated electrodes, as revealed by scanning electrochemical microscopy (SECM) measurements that were used to locally quantify the parasitic back reaction rates around individual defects to determine their impact on the global selectivity of an encapsulated electrode.