Urban logistics has become increasingly fragmented due to on-demand and time-sensitive delivery. Urban distribution systems have become multi-tier and multi-modal, and increasingly include multiple deconsolidation, cross-dock, and inventory locations. This implies an increase in the use of urban space, both for storage and movement of goods. While models to support urban planners and companies have advanced substantially, the combination of space and time requirements have received little attention. We develop innovative strategies leveraging advanced analytics to cope with the inherent dynamics of the urban logistics system, including stochastic models designed to support decision makers to best use of existing networks of logistics facilities and delivery modes, and to cope with limited urban space while meeting the increasing and time-sensitive customer expectations. We take a ground-breaking approach to also include the welfare of delivery couriers explicitly into our modelling approach, recognizing the anxiety and stress that human logistics operators face in this challenging environment. Our strategies and models are evaluated based on the urban realities of Bordeaux (France) and Chengdu (China). This allows us to compare logistics practice in two medium-sized cities with various topology and business environment, and with relatively low urban density to that in a metropolis with extremely high density.