A key problem in modelling trucks is in obtaining reliable information about actual truck movements. Surveys are difficult and expensive to conduct and often do not provide usable data. Some studies have used GPS data. Many large trucks now carry GPS units that provide a running record with latitude/longitude and time stamp. The American Transportation Research Institute (the research division of the American Trucking Association) sells this data to public agencies, at a fraction of the cost of a typical survey. Several studies have used this information to develop new truck trip models. This data can prove to be very useful in providing observed data on truck travel. However, the data must be viewed with a lot of caution. There are typically a lot of records (tens of millions) and they must be processed very carefully to correctly identify stops. The size of the database itself may preclude certain software options for processing the information. ATRI imposes a lot of confidentiality requirements that may complicate the use of this data for certain applications. For example, ATRI automatically geocodes the records to an agency’s TAZ system and this introduces certain inaccuracies. No identifying information about the truck or the operator is provided, making it difficult to stratify medium from heavy truck trips, or to develop expansion factors. Multi-day files may include the same truck several times. The time stamps are not uniform, but random, and the data contains many anomalies such as apparent movement between locations with no elapsed time. Therefore, users of this data must be especially careful in processing this information to convert it into usable, realistic trips or tours. If sufficient skill is employed, ATRI data can become a very useful, cost-effective way of identifying truck movements in support of model development.