Detecting incorrect product names in online sources for product master data

The global trade item number (GTIN) is traditionally used to identify trade items and look up corresponding information within industrial supply chains. Recently, consumers have also started using GTINs to access additional product information with mobile barcode scanning applications. Providers of these applications use different sources to provide product names for scanned GTINs. In this paper we analyze data from eight publicly available sources for a set of GTINs scanned by users of a mobile barcode scanning application. Our aim is to measure the correctness of product names in online sources and to quantify the problem of product data quality. We use a combination of string matching and supervised learning to estimate the number of incorrect product names. Our results show that approximately 2 % of all product names are incorrect. The applied method is useful for brand owners to monitor the data quality for their products and enables efficient data integration for application providers.