Human metastasis has been modeled mostly by xenotransplantation of cell lines in immunodeficient mice. Since this approach frequently uses cell lines derived from metastases, it ignores the significant role of cellular selection processes before and during metastatic progression and, in fact, models metastasis from metastasis and not metastasis from primary tumours. While the importance of the latter for the fate of patients is proven, the existence and clinical relevance of metastasis from metastasis is still unsettled. On the other hand, transgenic or gene knockout models of cancer offer novel experimental approaches to dissect the metastatic cascade from its very beginnings. Here, we briefly review the attempts to model metastatic progression and the strengths and limitations of the different experimental approaches and describe how transgenic mouse models recently helped to promote our understanding of systemic cancer progression.