Packaging is classified as one of back-end processes in the integrated circuits (ICs) manufacturing, highly capital-intensive and involves complex processes. Unlike the front-end process that fabricates wafers, the back-end process is rarely uniform. Because of the complexity of the process and increasing variety of products, the packaging foundry occasionally encounters complaints that can be categorized into classes depending on the loss. We apply rough set theory to discover important attributes leading to complaints and induce decision rules based on the data of a Taiwanese IC packaging foundry that ranks one of the largest in the world. The data contain 454 records and each record includes 11 condition attributes as well as one decision attribute characterizing the class. We first obtain important set of attributes that ensures high quality of classification, and then we generate rules for each class of complaints. The strongest rules obtained relate to two attributes, number ...