Phenotyping single cells based on the products they secrete or consume is a key bottleneck in many biotechnology applications, such as combinatorial metabolic engineering for the overproduction of secreted metabolites. Here we present a flexible high-throughput approach that uses microfluidics to compartmentalize individual cells for growth and analysis in monodisperse nanoliter aqueous droplets surrounded by an immiscible fluorinated oil phase. We use this system to identify xylose-overconsuming Saccharomyces cerevisiae cells from a population containing one such cell per 104 cells and to screen a genomic library to identify multiple copies of the xylose isomerase gene as a genomic change contributing to high xylose consumption, a trait important for lignocellulosic feedstock utilization. We also enriched L-lactate-producing Escherichia coli clones 5,800[times] from a population containing one L-lactate producer per 104D-lactate producers. Our approach has broad applications for single-cell analyses, such as in strain selection for the overproduction of fuels, chemicals and pharmaceuticals.