Integrated Optimization of Product Assortment, Inventory, an‎d Pricing for an Omnichannel Retailer Considering Customer Loyalty an‎d Customer Time Sensitivity

4/30/2026 12:00:00 AM

In the era of omnichannel retailing, making coordinated decisions on pricing, inventory, product assortment, an‎d delivery time in the presence of dynamic customer behavior has become a fundamental challenge. Prior studies have typically examined these decisions in isolation an‎d treated customer loyalty as a static an‎d qualitative construct, leaving its role underrepresented in mathematical decision-making models. This research introduces an integrated, data-driven framework in which customer loyalty is modeled as a dynamic, quantitative, an‎d predictable variable placed at the core of operational decisions. In the first stage, customer loyalty is predicted through a hybrid approach that combines graph convolutional networks for capturing network-based interaction patterns an‎d a transformer architecture for learning temporal dependencies in purchasing behavior. Comparative results against a baseline model based on RFM indicators an‎d customer lifetime value demonstrate that the proposed model provides higher accuracy in identifying high-value an‎d at-risk customers. Furthermore, incorporating predicted loyalty into a recommender system increases offer acceptance rates by up to 35 percent an‎d significantly enhances customer lifetime value. In the second stage, a robust multi-objective optimization model is developed to simultaneously maximize profit, increase customer loyalty, an‎d minimize carbon emissions. Solving the model generates a set of non-dominated policies on the Pareto front, an‎d sensitivity analysis reveals that cross-channel coordination has the strongest positive impact on customer loyalty. To address environmental dynamics an‎d competition, a dynamic game is formulated an‎d integrated with proximal policy optimization–based reinforcement learning to derive adaptive optimal policies. Extensive simulations show that the proposed framework outperforms benchmark methods such as genetic algorithms an‎d static proximal policy optimization, achieving 9 to 15 percent improvement in profitability an‎d loyalty prediction accuracy. Overall, the findings indicate that combining advanced machine learning techniques with robust multi-objective optimization provides an effective tool for data-driven decision support in omnichannel retailing an‎d enables simultaneous improvements in economic performance, loyalty management, an‎d environmental sustainability.

خرده‌فروشي همه‌كاناله , وفاداري , ، هماهنگي بين كانال‌ها، , كاهش انتشار دي‌اكسيد كربن , يادگيري تقويتي

Omnichannel retailing , loyalty , Cross-channel coordination , Carbon dioxide emission reduction , Reinforcement learning

Shima Roosta

Seyed Jafar Sadjadi, A hmad Makui