Document-type classification errors in bibliometric databases: Insights from the engineering/manufacturing field

Document types (DTs) - e.g., research articles, reviews, conference proceedings, letters, etc. - are not only used to classify scientific publications, but also to routinely guide inclusion-exclusion decisions in bibliometric assessments, often without adequate consideration of the quality of underlying content. This study examines DT-classification errors in Scopus and Web of Science (WoS), focusing on engineering/manufacturing publications. These errors - which may directly affect publication/citation counts, citation-impact indicators, and consequently academic evaluations and careers - are analyzed in a corpus of about 10,000 documents, using a recent semi-automated method. The results indicate that these errors, while occurring in several percentage points, are far from negligible. Furthermore, statistical analyses reveal systematic differences among publishers (e.g., Springer, Elsevier, Taylor & Francis, etc.), with some contributing more to errors, probably due to editorial styles or inconsistent metadata. This study provides insights for researchers, evaluators and database managers, highlighting the need for publisher-specific guidelines to enhance classification accuracy and reduce errors.