This article examines issues raised by the application of a generic actuarial juvenile risk instrument (the Model Risk Assessment Instrument) to New York City, a context different from the one in which it was developed. It describes practical challenges arising from the constraints of locally available data and local sensibilities and highlights important differences between locally relevant recidivism predictors and generic tool predictors. The analysis shows that the generic tool is less predictive than a locally developed risk-assessment tool and also performs less well than unassisted clinical judgment. This is true even after the generic tool has been validated and optimized on local data. This is because the tool does not include key demographic variables relevant to the New York City context. (Published Abstract)