FashionDSS: A Human-in-the-Loop Decision Support System for forecasting production efficiency and predicting item defects in the luxury fashion industry

In the luxury fashion industry, where quality directly impacts brand reputation and customer satisfaction, detecting product defects is crucial. This paper introduces FashionDSS: a Human-in-the-Loop Decision Support System designed to facilitate proactive quality management during the design phase. The system integrates three structured data sources, Product Registry, Bill of Materials (BOM) and Claim Records, into a unified predictive framework. Using machine learning models that are tailored to moderate-sized, information-rich datasets and validated through expert interaction, FashionDSS can perform two tasks: forecasting prototype production time (Task T1) and predicting potential defects in newly designed products (Task T2). Experimental results on real-world luxury fashion data demonstrate the feasibility of extracting meaningful predictive signals from structured complaint and production information despite data constraints. Although validation is currently limited to a single company’s dataset, the findings emphasise the potential benefits of combining predictive analytics with human expertise for design-oriented quality management.