Start Date
12-13-2015
Description
Product returns are a major challenge in e-commerce that severely affect the economic and ecological sustainability of the industry. While many static one-size-fits-all approaches to limit product returns have been proposed, there is a gap in the literature regarding strategies based on individual consumption patterns. We introduce a decision support system for the prediction of product returns, including a new approach for large-scale feature extraction. This system can be used as the basis for a returns strategy that allows online retailers to intervene before problematic transactions even take place. Using a dataset containing 1,149,262 purchases obtained from a major German online retailer, we demonstrate that our decision support system can identify consumption patterns associated with a high product return rate at sufficient accuracy for such a strategy to be feasible. We also show that the system outperforms a wide selection of state-of-the art classification and dimensionality reduction algorithms.
Recommended Citation
Urbanke, Patrick; Kranz, Johann; and Kolbe, Lutz, "Predicting Product Returns in E-Commerce: The Contribution of Mahalanobis Feature Extraction" (2015). ICIS 2015 Proceedings. 2.
https://aisel.aisnet.org/icis2015/proceedings/DecisionAnalytics/2
Predicting Product Returns in E-Commerce: The Contribution of Mahalanobis Feature Extraction
Product returns are a major challenge in e-commerce that severely affect the economic and ecological sustainability of the industry. While many static one-size-fits-all approaches to limit product returns have been proposed, there is a gap in the literature regarding strategies based on individual consumption patterns. We introduce a decision support system for the prediction of product returns, including a new approach for large-scale feature extraction. This system can be used as the basis for a returns strategy that allows online retailers to intervene before problematic transactions even take place. Using a dataset containing 1,149,262 purchases obtained from a major German online retailer, we demonstrate that our decision support system can identify consumption patterns associated with a high product return rate at sufficient accuracy for such a strategy to be feasible. We also show that the system outperforms a wide selection of state-of-the art classification and dimensionality reduction algorithms.