Modern economies are characterized not only by the growing number of market participants but also by a steadily increasing range of product variations. Therefore, the higher number of products offered is associated with a more significant amount of product data that must be managed. To ensure efficient communication between market participants, meaning that information of different formats is collected, and no data is lost or duplicated, central information systems such as the product information management (PIM) systems are applied.
Since PIM systems are specially designed to support marketing activities in e-commerce by enriching product data, monitoring the results of an image classification procedure in this context is of importance, as incorrect information might be published. Rich and correct product information in online retailing can thus be regarded as a fundamental prerequisite for the success of a retailer as it influences the purchasing decisions of customers. Thus, the integration of AI-systems raises explicitly the question of how the success and added value of these procedures should be measured and monitored in the operative business. This thesis intends to analyze these questions and close the gap in research concerning PIM systems and the business-related handling of AI-caused misclassifications.
