Beyond theory: data evidence from Belgium’s cargo bike transition
Read the report
As cities face increasing delivery volumes and environmental pressures, optimising urban logistics has become a critical challenge. While cargo bikes and Light Electric Vehicles (LEVs) have emerged as potential solutions, the logistics industry has lacked comprehensive comparative data to inform transition decisions.
At Kale AI, we’ve been examining this question through a series of research initiatives. Our November 2023 study with Urbike examined cargo bike efficiency in Brussels, while our work on urban micro-region embeddings developed a framework for analysing vehicle performance across diverse urban environments.
Building on this foundation, we’ve collaborated with the Belgian Cycle Logistics Federation (BCLF) to analyse data from KGS Group, a Belgian logistics operator that integrated cargo bikes into their operations beginning in 2022. This presented an opportunity to analyse comparable delivery flows performed by both cargo bikes and vans under the same operational conditions.
Analysing a real-world transition
Our research centres on KGS Group’s delivery operations in Brussels. After beginning to test cargo bikes in 2022, KGS progressively expanded their use based on operational experience, creating a mixed fleet that serves various urban contexts.
This study represents an advancement in methodology compared to previous research. By analysing 345 routes and 32,547 deliveries across Brussels – all performed by the same company delivering identical types of parcels – we could observe performance differences under comparable conditions.
The scope of our analysis spans multiple urban contexts from dense city center to suburban areas, providing insight into where and why each vehicle type performs differently. The data collection combined GPS trackers on cargo bikes, delivery management system data, and PDA tracker information to reconstruct delivery routes and analyse patterns.
The service time factor
A notable finding from our research is that approximately 60% of a delivery driver’s day isn’t spent moving but rather on “service time” at delivery points – finding parking, walking to doors, and completing the actual handover.
The data shows that in Brussels’ city centre, cargo bikes complete this phase 30% faster than vans on average (with service times of approximately 2.5 minutes versus 3.5 minutes). This difference becomes more pronounced in the most challenging urban areas, where cargo bikes perform 75% faster (2.4 vs 4.2 minutes per delivery). For a route of 100 deliveries, this translates to approximately 1.5 hours saved in the city centre, and up to 3 hours in the most difficult areas.
The analysis of service times shows that the worst deliveries for vans are significantly longer, but vans also have more “bad deliveries” than cargo bikes (e.g. 6% vs 2% of deliveries are longer than 10 minutes in the centre).
Analysis of outlier cases revealed additional patterns. In the five most challenging areas of Brussels, the 95th percentile (worst 5%) of van deliveries took up to 14.5 minutes, while cargo bikes rarely exceeded 5.8 minutes in these same areas. This suggests that cargo bikes offer not just improved average performance, but more consistent service times as well.
The reliability advantage becomes evident during peak hours (14:00-18:00). While van service times in central Brussels increase from approximately 3 to nearly 6 minutes, cargo bikes maintain consistent 2–3 minute delivery times throughout the day.
We identified what could be described as an “efficiency boundary” across the city – a zone where the relative performance advantage shifts between vehicle types. Within this boundary in core urban areas, cargo bikes consistently deliver faster than vans. Beyond it, in less dense suburban areas, vans begin to regain their efficiency advantage. This highlights the complementarity of different vehicles across diverse urban areas.
Strategic implications for urban logistics
The implications of these findings extend beyond operational efficiency. Using high-resolution population density data, we observed that while Brussels’ city center comprises only 12% of the total area, it houses 25% of the population (326,000 people). When including the wider dense urban area, 75% of the population (950,000 people) lives in zones where cargo bike operations appear to be most efficient.
Population density in Brussels.
This population concentration suggests that targeted vehicle transitions in dense urban areas could achieve significant positive impacts. Cargo bikes produce substantially lower emissions than diesel vans, reduce noise pollution, decrease congestion, and improve street space utilisation.
Importantly, the data indicates complementary roles for different vehicle types. Cargo bikes demonstrate efficiency advantages in dense urban cores, while vans maintain advantages in suburban areas with greater distances between deliveries.
Our analysis also highlights the importance of hub location. When including stem time (travel between hub and first/last delivery), the efficiency of both vehicle types decreases with distance from the hub, though vans appear more significantly affected. Networks of micro-hubs are imperative to maximise the efficiency of cargo bikes across urban areas, making up for their smaller capacity while extending their operational range.
The digitalisation challenge
We visited 6 different operators in Belgium, studied their operations, and interacted with their systems. Throughout this process, we observed consistent challenges with data collection and digitalisation across logistics operators. Many companies continue to rely on manual processes for route planning, delivery tracking, and performance analysis, with limited data collection systems.
This observation aligns with what we identified in our article on the “Ten hurdles to overcome for better urban logistics.” As we noted, “Data gaps exist on both sides of the transition. Most logistics companies can’t properly measure the full costs of their van operations… Meanwhile, limited operational data exists about LEV performance in different contexts.”
The predominantly manual nature of many cargo bike operations creates inefficiencies as companies attempt to scale. Without systems to capture and analyse operational data, these operators face difficulties optimising their processes and scaling.
This technological gap represents an opportunity for efficiency improvements. Our analysis suggests that combining the urban advantages of cargo bikes with intelligent logistics systems can strongly enhance the economics of sustainable urban delivery. The data supports what physics and economics indicate – LEVs offer advantages for urban freight, particularly when supported by appropriate logistics systems.
Building intelligence for urban logistics
At Kale AI, we focus on developing AI systems for urban logistics that can help transform complex city deliveries into more efficient operations. The patterns observed in this research – from the significance of service time to the location-dependent efficiency differences – inform our work on tools for multi-hub orchestration, mixed fleet optimization, and route planning.
These findings build on our previous work on urban micro-region embeddings, where we developed machine learning models to analyse how different vehicles perform across urban contexts. By understanding the factors that influence vehicle performance in various environments, we can design systems that leverage the strengths of each vehicle type.
We see the future of urban logistics involving not just vehicle transitions, but intelligent systems that coordinate complementary networks. Strategic micro-hub placement, vehicle allocation based on urban context, and continuous optimisation through machine learning represent important elements in developing more sustainable and efficient city delivery systems.
Read the reportThe full report “Urban Deliveries: Data Evidence from Belgium’s Cargo Bike Transition” is available here. This research was conducted in partnership with the Belgian Cycle Logistics Federation, with support from multiple Belgian logistics operators. Kale AI thanks all participants for their valuable contributions to this study.
For those interested in our research approach, we recommend exploring our November 2023 study with Urbike, which established a framework for evaluating cargo bike performance, and our paper on Urban micro-region embeddings, which details our machine learning methodology for urban logistics analysis.