Microbiomes contribute directly or indirectly to host health and fitness. Thus far, investigations into these emergent traits, referred to here as microbiome-associated phenotypes (MAPs), have been primarily qualitative and taxonomy-driven rather than quantitative and trait-based. We present the MAPs-first approach, a theoretical and experimental roadmap that involves quantitative profiling of MAPs across genetically variable hosts and subsequent identification of the underlying mechanisms. We outline strategies for developing ‘modular microbiomes’ — synthetic microbial consortia that are engineered in concert with the host genotype to confer different but mutually compatible MAPs to a single host or host population. By integrating host and microbial traits, these strategies will facilitate targeted engineering of microbiomes to the benefit of agriculture, human/animal health and biotechnology.