Visually Interactive Models Help E-retailers Navigate Last Mile Complexity

The growth trajectory of e-commerce is increasing the complexity of online buying channels. E-retailers need to change the way they deliver to urban customers to successfully manage this complexity and remain competitive.

CAVE interactive visualization at MIT CTL

It is a daunting challenge. Companies must assemble a complicated jigsaw of fulfillment centers, collection/delivery points, and freight carriers that achieves the optimal balance between cost and service.

To help e-retailers develop these solutions, MIT CTL’s Megacity Logistics Lab (MLL) has developed a modeling framework that considers, in an integrated fashion, strategic design decisions regarding network configuration, delivery service offerings, the choice of transportation modes, and other factors. In addition, the advanced visualization technology provided by the MIT CTL CAVE (Computational and Visual Education) Lab, enables companies to interact with the model’s inputs and outputs intuitively.

Explicitly geared to large-scale, multi-service delivery networks in online markets, this powerful combination of analytical tools has been applied to a real-world e-retailing operation in Brazil with impressive results.

The changing face of last-mile logistics

E-commerce was already growing dramatically when the COVID-19 pandemic gave it a boost. Homebound consumers have switched to online channels in increasing numbers, adding product categories such as groceries to the list of items they like to buy virtually.

In addition to soaring demand, e-retailers face rising customer expectations concerning lead time, delivery location, and service flexibility. These requirements fragment deliveries to residential areas and congested business districts, increasing routing complexity and costs. As a result, the industry is reshaping the logistics of the last mile on various fronts. Here are some examples.

• E-retailers are offering a more extensive choice of delivery service options. Notable examples are increasingly popular same-day and instant delivery alternatives.
• As expedited delivery options gain popularity, configurations of logistics facility networks are changing to cater to the new service demands. Companies are investing in multi-echelon (multi-layered) distribution networks with facilities that vary in both type and size. The aim is to minimize the cost of providing faster, more responsive delivery services.
• Many e-retailers and parcel operators are creating hyper-local facilities that add a layer to their distribution networks. At the same time, they are expanding the range of product exchange locations to offer alternatives such as pick-up at collection and delivery points (CDPs). These locations are being automated to different degrees, ranging from fully-automated parcel lockers to attended pick-up points in stores.
• The mix of transportation modes deployed by e-retailers over the last mile is diversifying. For example, companies are employing cargo-bikes and three-wheel vehicles that are well-suited to navigating congested, highly regulated urban centers. The trend towards an on-demand economy leads e-retailers to use third-party operators — including crowd-sourced transportation options — to complete last-mile deliveries.

As e-retailers develop the last-mile delivery networks they need to support these changes, it is critically important to evaluate the various configurations available to them before investing in expensive infrastructure and logistics systems. Moreover, these evaluations must be scalable in the real world and accessible to strategic decision-makers. This is where the CAVE’s network visualization technology — which bridges the gap between analytical models and practical decision-making — makes a vital difference.

Multiple distribution options

Key to the MLL’s modeling framework is its integrated approach to last-mile delivery network design. The sheer number of facility and system combinations that e-retailers can deploy to create high-performing delivery operations can be overwhelming. Modeling exercises must take all of these possibilities into account in an integrated manner.

MIT CTL’s network modeling team has carried out numerical experiments to demonstrate how its new modeling approach together with the visually interactive user interfaces developed by CAVE meets the needs of today’s e-retailers. And to confirm that the method generates solutions that can be scaled to real-world applications, the tools were used to analyze a Brazilian E-retailer's last-mile delivery operation. Importantly, the results are not just theoretical — the company’s management team is using the findings to help shape their delivery strategy.

Here is a rough breakdown of the operation modeled by the team.

• One distribution hub with a capacity of 17,000 parcels per day.
• Six possible satellite facilities (SFs), with capacities ranging from 1,500 parcels and 6,000 parcels per day.
• Six local transshipment points (LTPs) with a capacity of 500 parcels per day.
• Three delivery services: standard (STD) with a delivery lead time of at least one day, express (EXP) with a delivery lead time of six hours, and instant (INST) with a lead time of four hours. This is a simplified version of the actual transportation modes used.
• Demand aggregated to orders per day per square kilometer in demand zones (zones receiving less than 10 orders per year were omitted from the study area).
• Approximately 15,000 orders served across the study area on a typical day.
• Two demand scenarios: base (standard home deliveries) and baseCDP (baseline with home and collection and delivery points, or CDP, deliveries).

Hubs represent the first echelon in the network and originate last-mile freight flows. These are usually large facilities where inventory is stored, and parcels are made ready for online order placement. Satellite facilities (SFs) constitute the second network echelon. These facilities receive shipments from hubs in large-capacity freight vehicles. At SFs, a fraction of parcels is sorted, consolidated, and loaded onto last-mile delivery vehicles. The remaining parcels are transferred to local transshipment points (LTPs), small, third-echelon facilities that serve localized last-mile delivery routes.

Using the CAVE’s interactive visualization capabilities, the team looked at many different network configurations. The analyses yielded many insights. For example, faster delivery services increase the average cost per parcel for the demand scenarios modeled. This finding is mainly due to an increase in last-mile distribution costs. The introduction of CDPs reduces the cost of standard deliveries. The model underscored the growing trend to establish more local facilities. In the high CDP scenarios, three SFs and six LTPs are activated. Three SFs and one LTP are activated in the base scenario. The models also confirmed that the mix of delivery services and product exchange options impacts the configuration of second-echelon facilities in the network.

Such insights informed the management team’s long-term investment decisions for restructuring the company’s urban distribution network and introducing new services. For instance, the company decided to utilize LTPs as a central element of its product exchange offerings. Cargo-bikes could feature in the E-retailer’s last-mile distribution footprint, to help meet its sustainable transportation goals.

Some general advice

In addition to these company-specific insights, these analyses also offer findings that are of value to any e-retailer. Here are some examples.

• E-retailers need to offer time-differentiated delivery services to meet diverse customer expectations. However, faster delivery services also tend to be less efficient and more costly. The modeling shows that establishing facilities such as LTPs near to customers and the use of outsourced and crowd-sourced transportation can offset these downsides.
• Vehicle capacity influences the degree to which loads can be consolidated and length of delivery routes, and hence the infrastructure required to support delivery fleets. For example, the model indicates that cargo-bikes and motorbikes are cost-effective when deliveries are local to a facility. However, for more distant, standard deliveries, the larger capacity of minivans makes these vehicles a better choice.
• The choice between in-sourced, outsourced, or crowd-sourced transportation modes depends on the density of the demand zone and its proximity to the freight facility. For example, the more distant the demand zone, the more cost-effective outsourced minivans become.
• CDPs achieve lower transportation costs as they consolidate multiple deliveries in a single stop. Introducing facilities of this type can alter the optimal configuration of an upstream delivery network and should be considered with other facility-related design decisions.

The best jigsaw wins the network design modeling challenge

It is worth emphasizing again that the integrated approach to modeling last-mile network design systematically outperforms exercises that consider the numerous factors involved in isolation. Network designs based on the ad hoc approach are less likely to provide cost-effective delivery solutions in today’s dynamic competitive environment.

New delivery concepts will no doubt emerge, but in the main, the existing options will be the ones that shape e-retailers’ last-mile delivery strategies. Those companies that develop and implement designs based on the most efficient combinations of these options will gain a competitive edge.

Further Information:

• See the CAVE Lab video that explains how interactive visualization takes network design and related decision-making to new levels.
• Four Last-Mile Routes to Successful Omnichannel Retailing by Matthias Winkenbach.