There is an increasing awareness that vegetation diversity can affect herbivore and natural enemy abundance and that plants can play a major role in directly manipulating natural enemy abundance for protection against herbivore attacks. Using data from cassava fields, we aimed at (i) testing the capacity of the predatory mite Typhlodromalus aripo to control the herbivorous mite Mononychellus tanajoa in a chemical exclusion trial; and (ii) testing, based on the differential preference by T. aripo for cassava cultivars, how combinations of two morphologically different cassava cultivars with differential suitability to the predator can improve its population densities on the non-favourable cultivar, thereby reducing M. tanajoa densities with subsequent increases in cassava yield. The study was conducted in a cassava field in Benin, West Africa. The experiments confirmed that T. aripo effectively suppresses M. tanajoa populations on both cultivars and showed, in the no-predator-exclusion experiments, that cultivar combinations have significant effects on M. tanajoa and T. aripo densities. Indeed, T. aripo load on the non-preferred cultivar was lowest in subplots where the proportion of T. aripo-preferred cultivar was also low, while, and as expected, M. tanajoa load on the non-preferred cultivar showed decreasing trends with increasing T. aripo densities. The possible mechanisms by which cultivar mixing could increase predator load on the non-favourable cultivar were discussed. Our data showed that appropriate cultivar combinations effectively compensate for morphologically related differences in natural enemy abundance on a normally predator-deficient cultivar, resulting in lower pest densities on the non-favourable cultivar. In practical terms, this strategy could, in part, enhance adoption of cultivars that do not support sufficient levels of natural enemies for pest control.