To aid the characterization of gene–phenotype relationships, de Angelis et al. developed and implemented statistical methods to optimize the design and analysis of multicentre, large-scale, broad-based phenotyping of mouse knockouts. The authors analysed data from over 27,000 mice with a total of 449 mutant alleles representing 320 unique genes, and found at least one deviant phenotype in 83% of mutant lines and more than one in 65% of lines. Notably, there were significantly more phenovariant lines among homozygotes than heterozygotes, and phenotypic effects were stronger in homozygotes. Finally, the authors identified new phenotypes for 87.9% of genes with previously unknown function and identified a large number of candidate disease-associated genes, demonstrating the value of broad-based phenotying on a large scale.