Identifying car-sharing quality determinants: a data-driven approach to improve engineering design

Purpose - This study aims at identifying the quality determinants of car-sharing services, analyzing unstructured User-Generated Contents (UGCs) and, more specifically, online reviews generated by users of the same car-sharing service. Moreover, this paper discusses the implication of the proposed data-driven approach on engineering design. Methodology - A large dataset of car-sharing users' online reviews was analyzed by means of the Structural Topic Model (STM), i.e. a variant of Latent Dirichlet Allocation (LDA) technique which discovers underlying topics in a collection of documents also using document-level covariate information. Findings - This paper reports an analysis of UGCs related to different car-sharing services. The analysis unveils 20 determinants of car-sharing quality: customer service (physical office); accident & damages management; registration process; charges & fees; parking areas; app reliability; end trip issues; car condition; convenience; use rates; car proximity; car availability; efficacy; sharing benefits; customer service responsiveness; intermodal transportation; car start-up issues; customer service courtesy; billing and membership; car reservation. Originality – This paper proposes a novel approach to identify quality determinants by analyzing UGCs. The study of the quality determinants of a car-sharing service is a scarcely discussed field of research although the car-sharing sector is an increasingly important part of the transport economy.