1. Cross-fostering is the transfer of offspring between their natal environment and a new social environment. This method allows researchers to disentangle the genetic and interacting environmental effects that influence phenotypes, and is popular in both wild and laboratory studies. Here, we discuss three factors that might bias cross-fostering and influence ecological and evolutionary conclusions if not accommodated. 2. First, cross-fostering tends to be spatially and temporally non-random because heterogeneous breeding conditions can result in clustered breeding attempts. Second, cross-fostering will often change the brood composition because the exchanged broods are unlikely to be precisely matched in age, size, and composition. Third, some methods can introduce bias by using a systematically structured subset of the population, leading to a systematically structured data-set. 3. We use a 12-year case study of wild house sparrows (Passer domesticus) to demonstrate how to identify these biases with statistical modelling and how to adjust the cross-fostering protocol according to the identified biases. 4. In our dataset, cross-fostered nestlings were more likely to survive than non-cross-fostered nestlings, but post-fledging and overall survival were not affected. Survival differed between cross-fostering treatments, partially due to temporally non-random breeding conditions and non-random offspring selection, demonstrating two of the three forms of bias in data from a wild population. 5. In all cases, we suggest using statistical models to examine whether cross-fostering opportunities and offspring fitness are affected by non-random breeding, changes to the brood composition, and biased methodology. We provide guidelines for optimising a cross-fostering design and reducing inherent bias.