Many mutations, including those that cause disease, only have a detrimental effect in a subset of individuals. The reasons for this are usually unknown, but may include additional genetic variation and environmental risk factors1. However, phenotypic discordance remains even in the absence of genetic variation, for example between monozygotic twins2, and incomplete penetrance of mutations is frequent in isogenic model organisms in homogeneous environments3,4. Here we propose a model for incomplete penetrance based on genetic interaction networks5,6. UsingCaenorhabditis elegansas a model system, we identify two compensation mechanisms that vary among individuals and influence mutation outcome. First, feedback induction of an ancestral gene duplicate differs across individuals, with high expression masking the effects of a mutation. This supports the hypothesis that redundancy is maintained in genomes to buffer stochastic developmental failure7. Second, during normal embryonic development we find that there is substantial variation in the induction of molecular chaperones such as Hsp90 (DAF-21). Chaperones act as promiscuous buffers of genetic variation8,9, and embryos with stronger induction of Hsp90 are less likely to be affected by an inherited mutation. Simultaneously quantifying the variation in these two independent responses allows the phenotypic outcome of a mutation to be more accurately predicted in individuals. Our model and methodology provide a framework for dissecting the causes of incomplete penetrance. Further, the results establish that inter-individual variation in both specific and more general buffering systems combine to determine the outcome inherited mutations in each individual.