Designing Size Inclusive Fashion Assortments

信息管理与工程学院102室
上海财经大学（第三教学楼西侧）

上海市杨浦区武东路100号

Mehmet Gumus is Professor of Operations Management and Desautels Chair in Supply Chain Management and Business Analytics at the Desautels Faculty of Management at McGill University. He joined McGill in 2007 from the University of California at Berkeley where he completed his Ph.D. in Industrial Engineering and Operations Research and M.A. in Economics. In his research, Mehmet Gumus focuses on supply chain management, dynamic pricing, and risk management. His papers are accepted for publication in Management Science, Operations Research, Manufacturing and Service Operations Management, Marketing Science and Production and Operations Management. He serves as AE for Production and Operations Management, and IIE Transactions. In 2017, he co-developed a new specialized self-funded Masters program in Analytics (MMA) and since its inauguration, he is managing MMA program as the Academic Director. In 2015, he co-founded Logiciel Plannica inc. to develop and implement an integrated Supply Chain Analytics Solution that seamlessly integrates with variety of Database and Enterprise Resource Planning (ERP) systems.

Fashion retailers adjust their product assortments frequently to enhance profitability and customer satisfaction. Advanced analytics-driven systems supporting this process focus primarily on revenue improvement and have led to significant reductions in markdowns and increases in sales. However, this revenue-focused approach often results in the under-representation of plus-sized options which raise fairness concerns and negatively affects brand perception. This paper is motivated by a collaboration with a major European online fashion retailer and aims to tackle this issue by adjusting option planning to balance revenue and size inclusivity.  We consider a choice model that allows for size substitution and develop a measure of pairwise fairness among customer size groups. We then integrate this measure into the assortment optimization problem of an online fashion retailer. We estimate the size substitution coefficients and calibrate our model using a real dataset provided by our industrial partner. Simulation results show that our approximation algorithm can reduce unfairness by 30% with only 1.5% sacrifice in revenue. The results demonstrate the practicality of our algorithm as the retailer typically randomizes over a few assortments.