Patient derived tumor xenografts (PDXs) are important models for pre-clinical testing in cancer research and personalized medicine. PDXs often represent patient tumors with high similarity in terms of histology and driver mutations. However, certain limitations exist that warrant a detailed understanding of PDX heterogeneity and evolution. Hynds et al. demonstrate the relevance of primary tumor heterogeneity in PDX model establishment and explore multi-region sampling to determine the extent to which PDXs represent primary tumors.