In order to setup an academic experimental design in fatty liver production, we considered different scenarios in a simulation study, where the factors of variation were the number of successive batches, the number of measured animals and the number of sires and dams used to produce the animals. The three duck populations classically involved in fatty liver production (Muscovy duck, Pekin duck and mule duck) were considered. We aimed at determining a minimal number of animals to rear and measure, in order to obtain reliable estimates of the parameters of interest. They were the parental contributions to the mule duck performance and the purebred-crossbred genetic correlations between traits measured on the hybrid or on its parents. Our decisions were based on the mean square error of these parameters. Using 30 sires and 60 dams in each parental populations to produce 180 measured animals per batch, two successive batches were found sufficient to obtain reliable estimates of parental contributions. Genetic correlations required a much larger number of individuals.