Motor vehicle accident databases provide valuable safety information about the real world crash experience for millions of motor vehicles. The associated research often requires specific information about vehicle characteristics derived from the Vehicle Identification Number (VIN). The complicated process of large-scale VIN decoding is made easier with software that identifies specific make, model and other characteristics. Data entry errors and truncated VINs (i.e., less than 17 characters), however, pose challenges to reliable vehicle identification. A study of VIN coding requirements and data entry patterns indicates that reliable accident vehicle identification can be accomplished by supplementing VIN decoding software to systematically minimize common transcription errors and invalid characters. This process can yield reliable vehicle identification and maximize utilization of vehicle records for motor vehicle crash analysis. This paper discusses the processes involved in VIN decoding for millions of accident-involved vehicles and demonstrates, using VINDICATOR software, the application of supplemental VIN processing.