Managing agricultural land to maximize the supply of natural pest control can help reduce pesticide use. Tools that are able to represent the relationship between landscape structure, field management and natural pest control can help in deciding which management practices should be used and where. However, the reliability and the predictive power of generic models of natural pest control is largely unknown. We applied an existing generic model of natural pest control potential based on landscape structure to nine sites in five European countries and tested the resulting values against field measurements of natural pest control. Subsequently, we added information on local level factors to test the possibility of improving model performance and predictive power. The results showed that there is generally little or no evidence of correlation between modeled and field-measured values of natural pest control. Moreover, we found high variability in the results, depending on the associations of crops, pests and biocontrol agents considered (e.g. Oilseed rape-Pollen beetle-Parasitoids) and on the different case studies. Factors at the local level, such as conservation tillage, had an overall positive effect on natural pest control, and their inclusion in the models typically increased their predictive power. Our results underline the importance of developing predictive models of natural pest control which are tailored towards specific associations between crops, pests and biocontrol agents, consider local level factors and are trained using field measurements. They would serve as important tools within farmers' decision making, ultimately supporting the shift toward a low-pesticide agriculture.