We introduce a driver characterization methodology employing two scoring methods, one identifies similar driving patterns using clustering algorithms (Behavior Group) and the other identifies potential losses via the evaluation of several quantile regression models (Risk Score). We discuss the interpretation the Behavior Group and the Risk Score, so that an insurance firm can select and price drivers from their driving behavior and potential risk. The Behavior Group and the Risk Score can be updated continuously as data flows are provided. With this methodology we are capable of synthesizing telematic information and obtain similar model performance than using all telematic variables, while reducing computational costs and improving pricing strategies.