What to Read Next
It is estimated that by 2025, a quarter of the world’s population will reside in its 600 largest cities, and these urban centers will collectively account for about 62% of global GDP.
Today, there are already 28 so-called megacities on the planet — defined as cities with more 10 million or more inhabitants — and the United Nations projects that the number will reach 41 by 2030, accounting for a population of some 453 million people.
These urban centers represent a vast market for a wide range of products. But a major hurdle to unlocking the market’s full potential is the last mile — the final segment of supply chains where products are delivered to urban customers.
Limited infrastructure, traffic congestion, and arcane city planning and regulations are some of the problems that disrupt last-mile operations. The fragmentation of demand, spurred by the dramatic growth in e-commerce, adds another layer of complexity.
In addition, most megacities are located in emerging countries where small retail outlets — or “nanostores” — dominate the landscape. A nanostore may be a mom-and-pop grocery, or even a simple kiosk, and, due to their highly limited size, their stocks need to be replenished frequently. In Mexico City, for example, some 60% of the city’s nanostores maintain only one to two days of inventory.
In spite of such challenges, there are many reasons to believe we are on the crest of substantial progress with even the most challenging of last-mile deliveries. Innovative models such as smart locker systems, the use of electric vehicles, and on-demand fleet services such as UberRUSH are being explored. Autonomous delivery vehicles, while still years from wide-scale implementation, hold game-changing promise.
But new delivery technologies will not solve this problem alone. Even more fundamental is the way companies mine and model their data. Enterprises already routinely collect a wealth of data on vehicle movements and product sales, which can be combined with other data sources to improve the design and management of urban delivery services. In the longer term, advances in distribution network modeling, combined with innovative simulation and visualization technology, will dramatically transform the way companies view, design, and manage the last mile in densely populated urban centers.
Insights from New Models and Old
In spite of technology’s alluring promise to remake the last mile, managers should resist thinking too radically right away. The best starting point toward even the most dramatic improvements is making better use of the data that companies already own. Managers typically evaluate delivery network performance using a limited number of basic Key Performance Indicators, such as total time and distance traveled, and the percentage of goods delivered. This is useful information, but only provides an aggregate view of last-mile operations. Companies may not know it, but they have much better information at their fingertips.
GPS data derived from smartphones carried aboard vehicles yields a treasure trove of locational data. When combined with other sources — transactional data, census and geo-spatial data, and information on driver activities — it is possible to build highly detailed models of urban delivery operations. Such analytics can provide management with last-mile insights both on the strategic and day-to-day decision-making levels.
Meanwhile, some well-established players such as Walmart and Anheuser-Busch InBev, as well as a number of aspiring companies including the Brazilian e-commerce platform B2W Digital and their Indian counterpart Flipkart, are engaging in research collaborations to build high-resolution models of their urban last-mile delivery operations.
These collaborations are yielding vital information about last-mile operations. For example, managers are learning how much time and money it takes to serve specific customers in certain parts of cities, where vehicles run into congestion, and how much time it takes for truck drivers to find parking spaces. The efficiency of different delivery models, load configurations, and vehicle technologies is being analyzed. Project teams are also forecasting the likelihood of service failures due to disruptions or unexpected demand spikes.
Our research at the MIT Megacity Logistics Lab has helped to identify other customer-specific insights. In Mexico City, for instance, a retailer expected the vehicle crew on one route to sort bottles in its back room before completing the delivery, adding some 45 minutes of non-value-adding time to the transaction. Management was unaware of the practice until an analysis of GPS traces and order data revealed that the excessive service times were a recurring pattern. In other cases, managers did not know that some drivers were failing to adhere to official routes owing to safety concerns or customer payment issues: The drivers knew that certain customers would not have enough cash to pay their orders at the early-morning delivery time for which a delivery was scheduled. They altered the official route plan to account for the issue. Only when GPS data analysis revealed systematic discrepancies between planned and actual routes was the unauthorized change identified.
Such anomalies might appear trivial, but when multiplied countless times across major cities add significant cost to supply chains and bring down service levels.
Moreover, this type of intelligence enables managers to make more informed decisions. They could deploy different types of vehicles that are more suited to the dynamics of specific routes. Rewarding drivers according to their logistics performance rather than their payment collection rates could eliminate multiple visits to customers.
When rolled up, new data models can help companies develop entirely new last-mile supply chain strategies.
Traditionally, companies have served large population centers with distribution centers located on the outskirts of the city, where space was more plentiful and land less expensive. But this practice is no longer flexible enough to meet the varying needs of many urban markets. A multi-tier system that adds another layer of distribution facilities is required.
One such layer might be a fleet of mobile warehouses parked at strategic locations throughout the city. Another model begins with larger trucks designed for rapid offloading to smaller, more agile vehicles at transshipment points within the city.
Rethinking urban distribution should also involve city governments. In Santiago de Chile, the municipal government, in conjunction with the national Ministry of Transport, is using a last-mile model to improve the availability of dedicated parking bays for freight vehicles supplying the inner city.
Visualizing the Future
Looking ahead, last-mile models will — and must — become even more granular, and new tools will be required to accomplish this.
For example, using open-source technology invented at MIT as well as commercial augmented-reality solutions, companies are now able to interact with delivery-simulation models both visually and on a haptic level. The elements of a last-mile delivery network can be represented by physical or virtual objects. These objects could be a miniature of a distribution center or flags indicating the geographic boundaries of marketing measures, infrastructure improvements, or freight regulations. Using a visual display of a city map, managers can move, add, or remove objects/elements within the map to change the network configuration and market environment. As they make changes, the system immediately displays the cost implications on the map for any point within the city.
Such visualizations can be powerful management and training tools, especially when applied in a cross-disciplinary environment. Logistics functions can be joined by sales, marketing, and finance to learn and make decisions together. Deploying interactive simulation and visualization tools in this way enables faster, better-informed decision making.
When one looks only at the demographics, it is easy to believe the black box of last-mile delivery performance will only grow harder to unlock, but advances in delivery technology and, more so, in the ways companies can mine, model, and visualize their data gives stronger argument to believe the black box can be cracked altogether. It’s a process any company can begin today.