Gene expression variability between genetically identical single cells, previously viewed as random noise, can drive functional biological behaviors. These phenotypes, remarkably, can even be dictated by transient differences in a handful of key genes. Currently, we know little of how gene expression variability may affect reprogramming somatic cells into induced pluripotent stem cells (iPSCs) by expression of the four “Yamanaka” factors of OCT4, KLF4, SOX2, and MYC (OKSM). Even in homogenous conditions in isogenic cells, only a very rare (<1%) subset of cells admit reprogramming, and the characteristics of this subset remain unknown. The prevailing view is that this subset is determined by random chance, mediated by the probabilistic binding and activity of the OKSM factors early in reprogramming. However, recent demonstrations that related cells share the same reprogramming outcome implies the existence of pre-existing, intrinsic differences that drive reprogramming fates and that are heritable across cell division. Here, we leverage a retrospective clone tracing method to identify and characterize individual human fibroblast cells “primed” to become iPSCs in the initial population before exposure to OKSM. These primed fibroblasts showed gene expression markers of increased cell cycle speed and decreased fibroblast activation, both of which we demonstrate directly affect reprogramming outcomes. Furthermore, knockdown of a fibroblast activation factor identified by our analysis led to increased reprogramming efficiency, identifying it as a barrier to reprogramming. Changing the frequency of reprogramming by inhibiting the activity of epigenetic modifiers led to an enlarging of the pool of cells that were primed for reprogramming, indicating that the primed state is not fixed but context-dependent. Our results show that even homogeneous cell populations can exhibit heritable molecular variability that can dictate whether individual rare cells will reprogram or not. Furthermore, we provide evidence in support of a model of reprogramming that is predominantly deterministic instead of stochastic.