This paper addresses several issues that we encountered when implementing a randomized experiment to evaluate a promising intervention for serious juvenile delinquents, multi-systemic therapy (MST). The issues overlap with issues others experienced when trying to conduct randomized experiments in clinical practice, but are complicated because MST takes place in a legal context. The complex juvenile justice system makes it difficult to come up with a simple system of randomization. Problems encountered are the complexity of the referral process in this type of intervention, legal issues and the ethical concerns raised by clinicians. Our experiences might be useful to others planning to carry out a randomized efficacy study within the juvenile justice system. Firstly, other researchers might feel supported if they realize that others experience similar problems, and, secondly, we provide some concrete suggestions concerning the following: selecting the procedures and determining the moment of randomization, dealing with and overcoming the resistance of institutions involved, maintaining the cooperation of both the institution and the sample, and keeping turnover of research staff to a minimum.