Applying modern machine learning techniques to traditional personal lines pricing data, we were able to provide predictive models which were more accurate (when measured against an independent test set) than a standard GLM model. Such techniques enable clients to exploit their data to the fullest.