In the sales workforce, some sellers may generate greater revenue than others. For instance, an impactful seller may engage their customer in a unique way, such as more personable communication or demonstrated understanding of the customer’s business. The firm could interview their top-performing sellers about best practices, so that it can learn and implement them widely, but primary data collection is costly and time consuming. To improve efficiency, we consider a modification on outcome-dependent sampling using posterior estimates of sellers’ revenue impacts from a Bayesian model. Specifically, we adopt posterior credible intervals to identify impactful sellers both on their estimated rank and the width of the interval, in order to improve the reliability of the sample; a possible extension includes matched case-control designs, where the most highly ranked sellers are compared with others. We demonstrate the application of our approach in simulated (synthetic) data and consider business and policy implications for outcome-dependent sampling to understand and improve sales practice.