This thesis proposes a research framework that leverages high-resolution traffic and urban infrastructure data to improve analytical approximation methods used to inform strategic decisions in designing last-mile distribution systems. In particular, this thesis explores the effects of the road-network on the circuity of local trips, and introduces data-driven extensions to improve predictive performance of route distance approximation methods by increasing the resolution of the underlying urban road-network. Overall, these circuity-based extensions significantly increase the real-world validity of routing approximations compared to classical methods, and entail relevant implications in the configuration of logistics networks within urban markets. The framework presented in this thesis entails three inter-dependent levels of analysis: individual trip, consolidated route and last-mile network levels. In Chapter 2, we introduce a method to quantify and analyze the network circuity of local trips leveraging contemporary traffic datasets. Using the city of Sao Paulo as the primary illustrative example and a combination of supervised and unsupervised machine learning methods, significant heterogeneities in local network circuity are observed, explained by dimensional and topological properties of the road-network. Results from Sao Paulo are compared to seven additional large and medium-sized urban areas in Latin America and the United States. At a coarse-grained level of analysis, we observe similar correlations between road-network properties and local circuity across these cities. In Chapter 3, this thesis proposes a data-driven extension to continuum approximation-based methods used to predict urban route distances. This extension efficiently incorporates the circuity of the underlying road-network into the approximation method to improve distance predictions in more realistic settings. The proposed extension significantly outperforms classic methods, which build on the assumption of travel according to the rectilinear distance metric within urban areas. By only marginally increasing the data collection effort, results of the proposed extension yield error reductions between 20-30% in mean absolute percentage error compared to classical approximation methods and are within 10 - 20% compared to near-optimal solutions obtained with a local search heuristic. Further, by providing a real-world validation of classic continuum approximation-based methods, we explore how contemporary mapping technologies and novel sources of geo-spatial and traffic data can be efficiently leveraged to improve the predictive performance of these methods. Finally, building on the augmented route distance approximation, in Chapter 4 we explore the effect of road-network circuity on the design and planning of urban last-mile distribution systems. These improved routing approximations are used within an integer linear programming model to solve large-scale, real-world instances of the two-echelon capacitated location routing problem. Using the parcel delivery operation of Brazil's largest e-commerce platform in the city of Sao Paulo as the primary example to illustrate the impact and relevance of this work, we demonstrate how explicitly accounting for local variations in road-network circuity can yield relevant implications for fleet capacity planning, the location of urban distribution facilities, and the definition of facility-specific service areas. Results indicate that failing to account for local circuity would underestimate the necessary fleet size by 20% and would increase the total last-mile network cost by approximately 8%.