An extensive vibration survey was conducted on Red Line trains of the Massachusetts Bay Transportation Authority. Vibration from 420 individual train passages was recorded in the tunnel under the Beacon Hill neighborhood in Boston. A statistical model was developed to characterize the distribution of vibration levels in the fleet. Detailed vibration time histories showed that specific train cars could be identified by their vibration signature, offering a method by which maintenance could be targeted on the worst vibration offenders in the fleet. The statistical fleet model was used to predict the effectiveness of a targeted maintenance program as measured by the potential reduction in the maximum and average fleet vibration levels. A targeted maintenance program was found to reduce the maximum Red Line vibration levels by 4.5 dBV (vibration level in decibels).