Cancer drugs are extremely ineffective, generally because current therapies do not address cellular heterogeneity. While hypothesis-driven research and functional genomics identify ever more novel putative therapeutic targets, the scientific community lacks rationales to attack the cellular heterogeneity in cancer, to select among the targets the most promising, and to design combination therapies. In particular, these all fail to provide successful adjuvant therapy settings after curative resection of the primary tumour before the onset of manifest metastasis.Here I propose a novel way to address cancer cell heterogeneity and to develop a rationale for the design of adjuvant therapies. The proposal rests upon the premises that (i) cancer initiation is causally associated with genetic changes, (ii) early, functionally relevant genetic changes -particularly involving DNA loss- have the highest probability to be shared among the progeny of a monoclonal, yet genetically unstable, cancer, and (iii) subsequent, cumulative genetic changes must either add to the fitness of the cell or at least be neutral to enable progression. The proposal is then built upon our observation that a subgroup of disseminated cancer cells (DCCs) displays normal karyotypes and DNA changes smaller than 10 Mb whilst primary tumours and more advanced DCCs harbour multiple additional chromosomal changes at the time of analysis. I suggest that although these karyotypically normal DCCs are the putative “loser cells” in cancer progression - since they are arrested in bone marrow - they are central to uncovering the early genetic changes of an individual cancer. With these cells we will identify for the first time the catalogue of initiating changes of sporadic cancers in a systematic way. We will then test the function of the early aberrations and perform functional viability screens to develop novel systemic therapies that target the Achilles’ heel of a given cancer: its shared critical alteration.