In spite of the large literature documenting discrimination in various contexts, we know little about the impact of providing additional objective information about candidates’ abilities on gender differences in labor market outcomes. We study this question by leveraging data on over 60,000 online interviews for soft- ware developers, combined with the quasi-random introduction of a device providing an objective measure of candidates’ coding and problem solving performance. Despite gender gaps in the objective measure being much smaller than those in subjective interview evaluations, the improved quality of information does not reduce the gender gap in evaluations. We combine these results with a theoretical model, which has testable predictions for how gender gaps in ratings vary with additional information. Our results are not explained by attrition, endogenous matching, positive selection or gender differences in performance. We provide suggestive evidence that for a given objective performance, women receive lower subjective ratings than men.