Worker attrition is a costly and operationally disruptive challenge throughout the world. Although large bodies of research have documented drivers of attrition and its operational consequences, managers still lack an integrated approach to understanding attrition and making decisions to address it on a forward-going basis. To fill this need we build and then estimate a structural model that captures both the firm’s decision to terminate a workers’ employment (involuntary attrition) and uses an optimal stopping problem process to model a workers’ decision to leave the firm (voluntary attrition). We then estimate the parameters of the model and conduct counterfactual analyses on the population of 3,680 agents serving one client over five years for an Indian business process management (BPM) company. Our model reveals a number of interesting findings as we see that agents focus little on the future and are relatively insensitive to salary. These factors only increase with time spent at the firm. We find that supervisors have a strong impact on whether employees stay as they reshape the way that agents make their decisions. If all employees were managed by the best supervisor in our data then voluntary attrition would reduce by over 30%. Altogether our paper contributes to the burgeoning literature on people operations, as well as to managerial practice.