Ecological systems often exhibit a positive but saturating diversity–function curve. Variation in the mechanisms generating this relationship can alter the slope and variance of the curve, with implications for the optimal management of biodiversity for ecosystem services. In biological control, prevalence of selection effects supports augmentation of the most effective natural enemy, but complementarity effects support augmentation of natural enemy diversity. Optimization of biological control strategies from the results of diversity–function studies is limited because few consider changes in function with relative or absolute changes in abundance, and many confound the relative importance of richness and density through experimental designs (additive and substitutive). By manipulating species richness across an abundance gradient we show that effects of species richness are density dependent and indicate how this may be incorporated into experimental designs or models predicting resource consumption in diverse communities. Furthermore, the underlying mechanisms causing an observed diversity–function response, and its associated variation, changed across the richness–abundance gradient. Finally, species-rich assemblages provided higher levels of minimum function than species-poor assemblages, without any compromise on the maximum function possible.