A Data Mining application to quality control: a case study in the fashion industry

Nowadays, quality issues represent one of the most time-consuming activities that companies have to manage and they are becoming even more relevant than in the past due to the increased attention paid to reducing lead time to reach the requested service level. Even though high-quality standards can be targeted with 100% control on the produced items, it will require non-competitive time to deliver final products. Statistical analysis requires skilled resources to practically guide companies in the identification of sampling rules specific for a single item or item category as well as related to different materials in order to carry out targeted controls. In addition, easy-to-update datasets and rules are needed in the current dynamic contexts, discouraging statistical analysis that even requires skilled resources. Data mining techniques can be leveraged to analyse large amounts of data efficiently and effectively, preferring methodologies that intuitively provide results even for the non-experts in the analytics domain. According to this, a data-driven approach based on data mining could support companies in setting the sampling rules that better fit the items to be produced, in order to optimize the trade-off between the targeted compliance and the time to check it. Finally, a case study will demonstrate how the proposed approach suits dynamic and low-tech industries like the fashion one. A step-by-step implementation of the proposed procedure based on Association Rule Mining, indeed, has been perceived by the final users as both intuitive and meaningful, supporting them in discovering non-trivial relationships between data.