We present the results of an intervention targeting discrimination towards Venezuelan migrants in accessing public services in Ecuador. The multistage intervention was run in collaboration with the World Bank and the Ecuadorian government and aimed to change the attitudes and behavior of civil servants in three Ecuadorian ministries: Health, Education and Labor. In the first stage, a diagnostic tool is applied to a small sample of 6,000 civil servants. Building on the lessons learned in the diagnosis, an online training course is designed from scratch, and offered to the 6,000 public officers before the intervention is scaled up to 30,000 civil servants in Stage 2 and to the whole 300,000 in Stage 3. The diagnostic tool strongly builds on behavioral methods and techniques (e.g., vignette experiments, relative trust scales, implicit beliefs) to characterize the prevalence and intensity of different discriminatory biases. Our results strongly suggests that discrimination blindness is common among civil servants, and that migrants’ discrimination is frequent and substantial. Discrimination is hardly based on beliefs, and consistent with both preference-based and systemic factors. We document the lack of intersectional discrimination, as Venezuelan women have a similar probability of accessing services. The three modules of the intervention successfully change attitudes and behavior. We causally link training to specific behavior changes using a mixed differences-in-differences approach.