Mixed vs Siloed TNC Operations for Passenger and On-Demand Delivery Services in An Agent-Based Modeling Framework

Transportation Network Companies (TNCs) increasingly operate in both passenger mobility and on-demand delivery markets, prompting interest in whether a unified fleet serving both request types (mixed operations) performs better than maintaining separate fleets (siloed operations). This study uses the POLARIS agent-based simulation framework, applied to the six-county Austin metropolitan area, to compare mixed and siloed TNC operations. Scenarios are evaluated across multiple levers, including matching strategy (greedy, optimization-based, and preference-based), pooling configuration (passenger-passenger, delivery-delivery, and passenger-delivery pooling), and fleet sizes. Results show that without pooling, mixed operations provide modest but consistent efficiency gains, reducing vehicle miles traveled (VMT) per served request by 3–10% and lowering empty fleet miles by up to 5%, while offering substantial user-level benefits. Pickup wait times decrease by 10–15% for passengers and 25–35% for deliveries, reflecting improved spatial–temporal flexibility in assigning vehicles to heterogeneous demand. When pooling is enabled, however, outcomes become highly sensitive to the types of requests combined. Passenger–passenger and delivery–delivery pooling remain beneficial, with delivery wait times falling by over 50% in some scenarios. In contrast, passenger–delivery pooling introduces significant trade-offs: delivery wait times improve, but passenger wait times increase significantly in certain mixed-operation settings due to heterogeneous time windows and preparation delays. These findings clarify when mixed-use fleets enhance performance and underscore the importance of carefully designed pooling policies when integrating passenger and delivery services.