Level Up Your Sneaker Game : Applying machine learning techniques to support data-driven investment decisions in the sneaker resale market

Abstract

Sneaker resale has become a worldwide phenomenon. The resale market is growing, expected to potentially reach up to $30 bn by 2030. More and more people want to take part in making fortunes out of shifting high valued Nike SB sneakers, rare Air Jordans or eccentric Yeezys. Notably, traditional customer roles are changing: consumers are no longer only buying sneakers for wearing them themselves, but are also engaging in resale activities. Additionally, new market participants are entering the game with the sole aim of making profits as large as possible from buying and then reselling brand new shoes. The purpose of this thesis is to provide insight into how machine learning methods can support data-driven investment decisions in the sneaker resale market. Two different reseller personas will be introduced, together with a description of scenarios and questions these might encounter. Using data from StockX.com, the leading marketplace for sneaker resale, various machine learning techniques will be applied to arrive at founded investment decision for these two personas. To meet the needs of the different personas, this thesis makes use of both simpler methods, such as linear and logistic regression as well as KNN and regression trees, and more complex methods such as Random Forest and XGBoost models. The authors chose a practical approach with the analysis of different scenarios, aiming to allow sneakerheads, who engage in and are hence interested in information on resale markets, to profit from the insights. The research shows that both simple and complex methods can be useful in these decisions, reaching high accuracy values as well as oftentimes good predictions. It also shows that the sneaker resale price is influenced by a myriad of factors, and that especially celebrity collaborations seem to have high influence on resale value of sneakers.