High-throughput genomic technologies have revealed a remarkably complex portrait of intra-tumor heterogeneity in cancer and have shown that tumors evolve through a reiterative process of genetic diversification and clonal selection. This discovery has challenged the classical paradigm of clonal dominance and brought attention to subclonal tumor cell populations that contribute to the cancer phenotype. Dynamic evolutionary models may explain how these populations grow within the ecosystem of tissues, including linear, branching, neutral and punctuated patterns. Recent evidence in breast cancer favors branching and punctuated evolution driven by genome instability as well as non-genetic sources of heterogeneity such as epigenetic variation, hierarchal tumor cell organization and subclonal cell-cell interactions. Resolution of the full mutational landscape of tumors could help reconstruct their phylogenetic trees and trace the subclonal origins of therapeutic resistance, relapsed disease and distant metastases, the major causes of cancer-related mortality. Real-time assessment of the tumor subclonal architecture, however, remains limited by the high rate of errors produced by most genome-wide sequencing methods as well as the practical difficulties associated with serial tumor genotyping in patients. This review focuses on novel approaches to mitigate these challenges using bulk tumor, liquid biopsies, single-cell analysis and deep sequencing techniques. The origins of intra-tumor heterogeneity and the clinical, diagnostic and therapeutic consequences in breast cancer are also explored.