Exploring the Dynamics of Online Auctions for Remanufactured Electronics: Evidence from Explainable Machine Learning

Abstract

We conducted a study to understand how different factors influence the final auction prices of remanufactured electronics sold on eBay UK. Using the SHapely Additive exPlanation (SHAP) algorithm, we analysed the interactions among key variables across 1,039 remanufactured electronics auctioned on eBay UK. This study enables us to visually and quantitatively assess how these variables collectively impact the outcomes of auctions. Our findings reveal significant insights into seller strategies and bidder behaviours. We observe that less experienced sellers tend to set lower starting prices and opt for shorter auction durations to quickly attract bids and accelerate their sales history. In contrast, more experienced sellers leverage their established reputations to set higher starting prices, which not only attract initial bids but also increase final auction prices. In addition, our study highlights the impact of bid count and starting price on final prices. High bid counts can escalate the final price regardless of the starting price, especially when the latter is perceived as attractive. Conversely, a high starting price might limit bidding activity, potentially reducing the final price despite a higher initial bid count. Moreover, we identify that a combination of high starting prices and strong seller reputations synergistically maximizes auction outcomes. Strategically, sellers can optimize auction results by adjusting starting prices in anticipation of expected bid counts and enhancing item presentation through optimal photo quantities, especially during short auctions. This strategy reduces information asymmetry and builds buyer confidence, thereby fostering more competitive bidding.

Keywords: remanufactured electronics, Explainable Machine Learning, shapley additive explanations