BACKGROUND: Central cancer registries are required to collect industry and occupation (I/O) information when available, but the data reported are often incomplete. METHODS: We audited the completeness of I/O data in the New Hampshire State Cancer Registry (NHSCR) database for diagnosis year 2005, and reviewed medical records for a convenience sample of 474 of these cases. We compared I/O data quality before and after a statewide registrar training session on occupationally related cancers. RESULTS: The original 2005 data contained both I/O data in 11.5% of cases, and lacked any I/O data in 74.5%. Corresponding figures for cases selected for audit were 15.2% and 77.2%, which improved to 54.2% and 11.8% after medical record review. After registrar training, 47% of reports contained both I/O data, and only 14.4% of cases lacked any I/O data. CONCLUSIONS: Statewide training to highlight the importance of I/O data is an effective method to improve I/O data quality.