Parcel Categorization for Freight Return Modelling: Implementing the Multiple Discrete-Continuous Extreme Value (MDCEV) Model

The increasing volume of e-commerce returns poses significant challenges for urban transportation systems. However, little attention is paid to e-commerce returns in transportation research. This study integrates return parcel volumes into freight demand modelling using the agent-based framework logiTopp. By categorizing parcel demand (e.g. fashion and electronics) through a Multiple Discrete-Continuous Extreme Value model and linking these categories to return probabilities, we estimate category-specific parcel order and return volumes for Karlsruhe, Germany. The results indicate that roughly one in five parcels is returned, corresponding to about 27,500 return parcels per week.This framework provides a foundation for assessing the impacts of return flows on urban logistics and for evaluating operational strategies such as parcel lockers, autonomous collection systems, or integrated pickup by delivery tours. The findings highlight the need for empirical data collection on return parcel drop-off behavior to enhance the modelling framework and enable a more comprehensive assessment of its transportation impacts.