Curb to corridor: How Arizona models last-mile delivery behavior

Transportation planning for retail used to mean mall-access turning lane studies and adequate passenger vehicle parking. Today, last-mile delivery behavior shifts that activity. Instead of  hundreds of individual vehicles traveling to one shopping center, one stop-and-go commercial vehicle visits hundreds of shoppers. 

Door-to-door delivery is growing, but last-mile transportation metrics remain largely hidden. Without the right analysis, that delivery van appears as a single Origin-Destination trip from the warehouse to its half-hour lunch break, failing to capture its true multiple-quick-stops activity.

Altitude’s door-to-door vocation classification reveals this emerging travel pattern, helping planners adapt to changing road and curbside use. A recent case study in Arizona illustrates how this door-to-door vehicle vocation data solves a real modeling gap: last-mile delivery forecasting. 

What “last mile” means for planners

In a simple representation of today’s complex supply chain, there are three stages of activity. The “first mile” moves raw goods from their origin to manufacturing facilities; the “second mile” ships finished goods to storage and distribution warehouses; and the “last mile” gets product into consumer’s hands, either via retail stores or at the doorstep.

Last mile can be further broken into multiple types of transportation, most commonly business-to-business sales, business-to-consumer sales and an emerging group of consumer-to-consumer sales via online shopping platforms. 

Each of these commercial transportation activities has different origin, routing and delivery patterns, and they can only be captured by an analytics platform that recognizes and can categorize last-mile behavior.

Why last-mile data matters

As commercial vehicle patterns shift, planners risk falling behind on how to respond to and prepare for changing travel demand. With a more accurate model, planners can address multiple commercial travel issues beyond merely congestion and infrastructure wear and tear.

For example, a last-mile model supports more accurate vehicle miles traveled data (VMT), specifically in densely populated areas where there is a higher demand for parcel delivery. A better understanding of VMT and time-of-day patterns informs congestion management strategies and policies. 

Sustainability analyses rely on accurate VMT for emissions modeling. A last-mile travel demand model offers a better estimation of the environmental impact of these types of trips.

The data can be used for site management, including demand for parking locations or pickup and drop off requirements. There are also safety implications for these larger vehicles (vans or light to medium duty trucks) making trips and stops within residential areas. Planners must consider those vehicles differently than passenger cars. 

The data can also be used for a new generation of mobility solutions such as considering micro hubs or local distribution centers, especially within dense areas. 

Finally, the data is critical for truck trip estimates for urban congestion mitigation policies, which can include off-peak delivery regulations. In Arizona, a team of planners decided to create a model to capture the nuances of this growing segment of commercial transportation to better understand the shift in travel activity behavior.

Building a new framework: The MAG region analysis

A recent partnership between the Maricopa Association of Governments (MAG) and a professional services firm leveraged Altitude’s data to focus on door-to-door delivery. The team developed a dedicated framework for a new model to forecast e-commerce and last-mile delivery truck trip behaviors. 

The project had several goals:

• Create a separate modeling framework for door-to-door and delivery-oriented travel
• Use anonymized commercial vehicle telematic insights to identify trip chaining and high-frequency stop behavior
• Use stop analytics to support characterisation of curb activity and delivery dwell patterns
• Create a framework to establish a foundation for enhanced e-commerce demand forecasting

Using a two-sided modeling approach, planners focused not just on the delivery end of the trip, but the origin as well. Identifying travel patterns for both demand and supply required two different modeling approaches: 

Demand side — forecasting parcel volume by traffic analysis zone (TAZ) using land use and demographics

Supply side — routing parcels from depot to destination using service area partitioning and network schema

Altitude’s role in the project was to provide data for calibrating and validating the models. Our commercial transportation analytics uses real operational data from more than six million connected commercial vehicles. Because we are able to classify those vehicle movements by vehicle vocation, or use, the data can isolate last-mile activity from other types of commercial travel, including long-haul, regional and hub-and-spoke.  

What the data revealed: Key last-mile lessons

During the Arizona project, planners found several key takeaways for building this type of model. 

1. Stop analytics data is primary. Trip origin is usually fixed from a depot, distribution center or warehouse, but those trucks then make up to 120 stops per day on a tour. Capturing stops is more useful than trip generation or trip pattern data. 
2. The time window plays a major role. With last-mile, there is a consumer expectation for delivery and therefore an operational constraint on the courier services. Including these time constraints offers a better estimation and forecast. 
3. Vehicle classification is critical. Identifying the vehicle type is not sufficient, since many types of vehicles are used in last-mile activity. Instead, the model must be able to apply a behavioral filter or vocation attribute that can separate and extract different types of behavior for different vehicle classes. 
4. Urban context matters. Dense urban areas show different types of and different amounts of parcel delivery than rural or transitional areas. 
5. Be aware of computational limitations. A last-mile model capturing stop behavior can grow exponentially very fast, sometimes too large to compute quickly and easily.

Existing travel demand models built on passenger and freight vehicle trips cannot fully capture the growing shift to last-mile delivery. Modelers and planners need different types of modeling and tools to capture this travel behavior, and Altitude offers access to a more detailed and nuanced level of information and operational detail. 

By leveraging Altitude’s commercial vehicle data, planners in Arizona were able to identify last-mile delivery trips, and quantify the stops associated with them, then extract the operational characteristics of those types of trips. The analysis reveals a new framework that establishes a foundation for enhanced e-commerce demand forecasting. 

Watch our webinar, “From Data to Decisions: The Arizona Regional Webinar” to see more insights about this last-mile delivery behavior analysis.