Internal benchmarking is the process of comparing the performance of one entity with the performance of comparison entities. Assessments of the various entities of the criminal justice system, such as police officers, judges, correctional facilities, and neighborhoods, often involve the construction of benchmarks with which to compare their relative performance. However, the typically made comparisons do not adequately account for the underlying differences in these entities. This article presents a general method, based on propensity scoring and doubly robust estimation, for constructing benchmarks for assessing the performance of entities within the criminal justice system while properly accounting for potentially confounding differences among the entities. The article demonstrates the method on an assessment of police performance in Cincinnati.