A highly advocated approach to enhance pest control by indigenous predators is to add or maintain appropriate non-crop plant habitats in agrosystems. Although the addition of non-crop plant habitats can enhance the number of predators in crop by increasing their food resources or shelter, the effect is often insufficient to reduce pest abundance or damage. A number of explanations were identified in previous studies; the ability of such habitats to enhance predators, in particular, is affected by the spatial organisation of the habitats at the land-scape level. Here, we explore how intra-plot spatial patterns of non-crop habitats affect the per-capita predator effect on pest control. We use a spatially explicit and individual-based model to simulate the foraging movements of an earwig-like predator in a banana field. Predator movements within a day were based on a simple non-specific behavioural assumption: movement is a correlated random walk affected by habitats and edges. Population dynamic processes occurring at larger time or spatial scales, such as reproduction and immigration, were not considered. In this model, non-crop habitats added to plots were considered favourable to predators: movements were slower and more sinuous in non-crop habitat than in unfavourable habitats. The intra-plot spatial patterns of the non-crop habitat were built and characterised using landscape ecology concepts and metrics. We found that the per-capita predator effect was strongly affected by a spatial dilution of predators, induced by non-crop habitat addition, but this negative effect could be partially or fully mitigated by the spatial organisation of the non-crop habitat. At the banana plant level, a long edge length between the crop and non-crop habitat can compensate for this dilution effect by reducing the duration of the periods between predator visits to the banana plant. At the plot level, the best plots (i.e., those in which all banana plants were often visited by predators) were those with non-crop strips in the banana plant rows. Overall, the results support the idea that the spatial organisation of non-crop habitats at the plot level, characterised by the metric edge length in particular, can be managed to minimise the negative impact of the dilution effect.