1. There is a growing consensus that the distribution of species trait values in a community can greatly determine ecosystem processes and services delivery. Two distinct components of community trait composition are hypothesized to chiefly affect ecosystem processes: (i) the average trait value of the species, quantified by community-weighted mean trait values (CWM; related to the mass ratio hypothesis) and (ii) the degree to which trait values differ between species in a community, quantified by different indices of functional diversity (FD; related to non-additive community effects). The uncertainty on the relative effect of these two components is stimulating an increasing number of empirical studies testing their effects on ecosystem processes and services delivery. 2. We suggest, however, that the interdependence between CWM and FD poses a challenge on disentangling their relative importance. We present a framework that allows designing experiments to decouple and assess the effects of these two community functional components on ecosystem processes and services. To illustrate the framework, we focused on leaf litter decomposition, as this is an essential process related to important ecosystem services. Using simulations, we applied the framework for plant leaf litter traits (litter nitrogen and phenolic content) that are related to litter decomposition. 3. CWM and FD generally showed a hump-shaped relationship (i.e. at more extreme CWM values, communities can have only low FD values). Within this relationship, we showed that it is possible to select quasi-orthogonal combinations of CWM and FD that can be treated statistically. Within these orthogonal CWM and FD combinations, it is also possible to select species assemblages controlling for other community parameters, such as total biomass, total density and species richness. 4. Synthesis. The framework provides a novel approach for designing experiments to decouple the effects of CWM and FD of communities on ecosystem processes, which otherwise cannot be easily disentangled. To apply the framework and design proper experimental layouts, it is essential to have a priori knowledge of the key traits by which species affect ecosystem processes and service delivery.