The addition of iron (Fe) can in certain cases have a strong positive effect on the activity of cobalt and nickel oxide nanoparticles in the electrocatalytic oxygen evolution reaction (OER). The reported optimal Fe dopant concentrations are, however, inconsistent, and the origin of the increased activity due to Fe dopants in mixed oxides has not been identified so far. Here, we combine density functional theory calculations, scanning tunneling microscopy, and OER activity measurements on atomically defined Fe-doped Co oxyhydroxide nanoparticles supported on a gold surface to establish the link between the activity and the Fe distribution and concentration within the oxyhydroxide phase. We find that addition of Fe results in distinct effects depending on its location on edge or basal plane sites of the oxyhydroxide nanoparticles, resulting in a nonlinear OER activity as a function of Fe content. Fe atom substitution itself does not lead to intrinsically more active OER sites than the best Co sites. Instead, the sensitivity to Fe promoter content is explained by the strong preference for Fe to locate on the most active edge sites of oxyhydroxide nanoparticles, which for low Fe concentrations stabilizes the particles but in higher concentrations leads to a shell structure with less active Fe on all edge positions. The optimal Fe content thereby becomes dependent on nanoparticle size. Our findings demonstrate that synthesis strategies that adjust not only the Fe concentration in mixed oxides but also its distribution within a catalyst nanoparticle can lead to enhanced OER performance.