While restorative justice (RJ) theory and practice have grown quickly in recent years, the science of restorative justice has lagged far behind. This is not unusual since science takes time to follow up on the long-term effects of different kinds of justice. One of the most important emerging questions about RJ concerns its consistency and variability. Even within the various categories of RJ programs, it is clear that actual practices vary widely. The design of different RJ programs may turn out to be less important than how they are implemented. RJ could benefit from experiments designed to reveal how differences in implementation may cause differences in outcomes. The primary focus of this analysis is repeat offending. The central question is whether qualitative differences from case to case in the conduct of a standardized RJ process can produce different effects on repeat offending. Some variable aspects of the delivery of a RJ conference with a focus on the police facilitator are examined. RISE data, including observational measures taken during the conferences as well as interview measures taken several weeks after the conferences, are included. The analysis shows that respect for police predicts repeat offending for some types of offenses but not others. It shows that repeat offending after a RJ process varies widely across different RJ facilitators; and suggests that experience of RJ leaders has little direct relationship with respect for police after RJ conferences. Results also show that the level of RJ leadership experience may be inversely related to repeat offending rates, controlling for prior offending rates, and for drink-driving cases. This also suggests that experience and practice do not affect the impact of RJ on juvenile cases, at least within a 2 year time frame with a RJ conference in the middle. There may be little difference between a system of experienced “professional” or “specialist” RJ leaders running all conferences, at least for juvenile recidivism. 1 table, 14 figures, 1 note, 25 references