Breeding schemes typically involve first the choice of parents and then the selection of offspring within crosses. Genomic prediction is adapted both for predicting the cross mean and for ranking the genotypes within a cross. These steps correspond to the components of the predictive ability of genomic prediction. The cross mean is the sum of the breeding values of the parents if allelic effects are only additive, but in practice, some deviation may result from dominance or epistasis. To date, only a few studies have investigated cross mean predictive ability in heterozygous crops, and none have investigated the parameters that influence it.