There is little evidence on the effectiveness of interventions with offending children and juveniles, either in Europe or North America. We present the evaluation findings of an innovative intervention and relate these to the existing evidence-base for young offenders and to the more extensive literature on older offenders. Methods Using an established risk predictor (LSI-R), we measured the criminogenic risks and needs of the intervention group and a non-intervention group of child and juvenile offenders at six-monthly intervals. Local police charges data were also collected for both groups as an indicator of offending behaviour. Results Over the first thirty months of the project, a statistically significant drop was seen in the LSI-R scores of the project participants, with favourable effect sizes. Effect sizes improved markedly with longer-term intervention. There were also strong indications that the level of offending behaviour had decreased during the intervention. The comparison group showed no change in risks, needs or offending rate. Conclusions We suggest this study provides an important contribution to the evidence-base of what works with child and juvenile offenders and we suggest that an increased focus on understanding effective intervention with very young offenders is required.