Despite identical genetic constitution, a cancer cell population can exhibit phenotypic variations termed as nongenetic/ non-mutational heterogeneity. Such heterogeneity - a ubiquitous nature of biological systems - has been implicated in metastasis, therapy resistance and tumour relapse. Here, we review the evidence for existence, sources and implications of non-genetic heterogeneity in multiple cancer types. Stochasticity/noise in transcription, protein conformation and/or external microenvironment can underlie such heterogeneity. Moreover, the existence of multiple possible cell states (phenotypes) as a consequence of the emergent dynamics of gene regulatory networks may enable reversible cell-state transitions (phenotypic plasticity) that can facilitate adaptive drug resistance and higher metastatic fitness. Finally, we highlight how computational and mathematical models can drive a better understanding of non-genetic heterogeneity and how a systemslevel approach integrating mathematical modeling and in (vitro/in vivo) experiments can map the diverse phenotypic repertoire and identify therapeutic vulnerabilities of an otherwise clonal cell population.