This research identifies the differences in predicted outcomes across race/ethnicity and sex groups for promotion of U.S. Navy line officers commissioned as O-1s (ensigns) in 2001-2018 to O-5s (commanders) in 2001-2018. Using administrative military personnel data, we train two machine learning models: one predicts retention, the other predicts promotion. We calculate the effect of each feature on an individual’s likelihood of retention or promotion and compare across demographics to identify differences in which and how much features matter.