Returns foresight: Explainable machine-learning models to predict and reduce product return propensity in omnichannel apparel retail

E-commerce retailers bear substantial additional costs arising from high product return rates due to lenient return policies and consumers’ impulsive purchasing. This study aims to accurately predict product return behavior before payment, supporting proactive return management and reducing potential losses. Based on the Graph Transformer, we proposed a novel return prediction model, Returnformer, which focuses on capturing user–product connections represented in topological structures of bipartite graphs. The Returnformer first integrates global topological embeddings into original node features to alleviate structural information loss caused by graph partitioning. It then employs a Graph Transformer to capture long-range user–item dependencies within local subgraphs. In addition, a graph-level attention mechanism is introduced to facilitate the propagation of global return patterns across different subgraphs. Experiments on a real-world e-commerce dataset show that the Returnformer outperforms four machine learning models in terms of prediction accuracy, demonstrating superior performance compared to the state-of-the-art models. The proposed model enables retailers to identify potential return risks prior to payment, thereby supporting timely and proactive preventive interventions.

Free Shipping Is Not Free: A Data-Driven Model to Design Free-Shipping Threshold Policies.

E-commerce has swiftly evolved to lessen the shop- ping burden of customers, it has at the same time seen an increase in fraud. Out of the numerous fraud types that online retailers face, the return of fake or tampered products has, in particular, become very expensive and destructive to the business. Traditional fraud prevention systems have their main focus on payment fraud, for example, the use of a fake credit or debit card. However, this project goes beyond that to detect the occurrence of fraudulent product returns. In fact, it talks about those customers who keep on returning fake products and thereby earning their money back through fraudulent means rather than in the area of payment fraud which has been less explored by the research community. Machine learning techniques viz. Random Forest, Logistic Regression, and XGBoost were utilized to scrutinize a fabricated dataset for detecting suspicious return behaviors. The ensemble model was 98% accurate in fact it was able to tell apart genuine from fraudulent returns. Besides, deep learning models were created and compared with conventional machine learning models, thus, showing even higher accuracy levels. The system’s distinct characteristic is the introduction of an email alert mechanism based on SMTP that informs the owners immediately when fraudulent activity is detected. By integrating explainable machine learning, deep learning, real-time alerts, gives robust, transparent, and actionable fraud detection to e- commerce platforms.