Using artificial intelligence for optimized last mile delivery

If you trace the evolution of the logistics industry, it is not very different from that of most other industries. Just like many others, it has always focused on cost-optimization. With accelerated digitization and with technology continuously shortening the turnaround time across all online services, customers have begun to expect (and even demand) higher level of delivery services. This change is particularly more visible in last mile delivery or the final leg of the delivery process (to deliver to the customers).

Last mile delivery is considered to be the most important element in logistics, as it directly impacts customer satisfaction. According to a McKinsey report, last mile delivery also plays a significant role in the overall logistics ecosystem (from the operations perspective), comprising as high as 50% of the overall shipment delivery cost.

How have customer expectations evolved with digitization?

With doorstep delivery becoming more a norm than an exception, customers expect much more now, as the market abounds with multiple types of delivery modes. With options like same-day delivery, home or workplace delivery, curbside delivery, and even delivery lockers with automated guided vehicles to pick up from a designated place, the customer is spoilt for choices. This has meant that customers now want faster and cheaper delivery, with more control over the experience. Don’t be surprised, if sometime soon, the experience of product delivery overtakes delivery price as the critical differentiator in determining customer experience.

The crux of the matter is that today, customers want delivery in a specific time window, as per their availability at home or office. Thus, operation planning managers at last mile delivery organizations need to optimize vehicles’ capacity and routes by considering the time window for each delivery. When this constraint of delivery time window gets added to the problems of driver shortage and limited vehicle capacity, it becomes very challenging to plan the logistics of last-mile delivery, which may result in late deliveries and lower customer satisfaction.

Given the delivery time window and the shortage of resources, meeting customer expectations results in higher expenses. Last mile delivery organizations have been struggling to reduce the driven distance to cut the overall transportation cost. Cutting transportation costs is the topmost priority for most logistics providers.

How can Artificial Intelligence/Machine Learning (AI/ML) improve logistics operations?

Technological enhancements and disruptions are already driving the top line of the logistics companies. Technologies like artificial intelligence and machine learning are playing a pivotal role in this era of digital disruption.

Logistics companies generate billions of gigabytes of structured and unstructured data every day. AI can harness this data to make real-life business and operational decisions, which earlier required human intelligence. As a result, with increasing investments being made in it every year, AI is getting C-level sponsorship more often.

What are the challenges that logistics companies face?

Logistics companies face various obstacles while trying to stay ahead of the curve. They face the shortage of local delivery resources for last-mile delivery and cannot adjust for the fluctuating demand. Also, their existing systems and processes are often designed to bring operational efficiencies by focusing on long-distance travel, and there is very little overlap between national and local transportation providers.

Planning operations at the local level is a different challenge as compared to long-distance planning. It involves more stops and several packages to drop in and around a small area. For example, finding the minimum distance from place A to B is simple. When we extend this problem where we need to visit multiple places (e.g., 500+), finding the best route can become more complicated. We can also consider other complexities such as time windows, customer preferences, multiple vehicles, and vehicle capacity & availability. To elaborate on this point, consider that there are 9.33e+157 (100 factorial) ways in which the 100 points can be covered without considering any constraints. The complexity grows further as we add constraints to the problem.

How can we overcome these obstacles?

Not too long ago, such problems were solved using statistical and rule-based solutions, wherein logistics-business-related rules and constraints were applied to statistical methods to find a solution. While these techniques offered a reasonable solution, they were slow and hence, not dynamic, and faced significant pressure with the customer’s growing demands. Further, the solution was often not optimal for the last mile delivery operations. It could not cater to last-minute changes or demand fluctuations from one point to another.

AI/ML algorithms provide better methods to solve such problems and achieve their business objectives. This technology evaluates the different possible solutions intelligently and extensively, generating better route density and bigger drop sizes. An AI-based solution for logistics should also have the capability to use historical data on drop sizes and routes to suggest a better route and pick-up plan. This enables the delivery organizations to plan and adjust against the demand fluctuations during peak seasons. Such a system can help prepare a better load plan for vehicles, based on the drop size and route plan, which can again be optimized through AI-based optimization. Over time, we can also leverage this technology to streamline logistics' business processes by bringing targeted suggestions and insights through historical data.

How does AI-enabled route optimization work in the favor of last mile delivery?

To reduce transportation costs while meeting customer expectations, more and more logistics companies have started exploring machine learning and artificial intelligence to optimize their delivery routes. An AI/ML-enabled route optimization solution can provide information about the optimal number of vehicles required and the shortest route to be taken to deliver the packages within the delivery time window. At the same time, the system can continue to learn from the already made deliveries every day and continue refining itself to meet maximum delivery windows while optimizing the transportation costs.

A route optimization solution, powered by artificial intelligence, can make logistics operations more efficient, resulting in cost reduction, improved customer experience, and better resource management. It can reduce the manual efforts required to adapt the routes and provide better visibility of the shipment to the planning managers as well as the customers.

How can Nagarro help?

Nagarro has worked on delivery and routing optimization projects many times for different clients. For us, the advent of AI/ML made us believe that we could leverage them to solve the challenges of delivery and routing optimization far more efficiently and quickly. Keeping this objective in mind, we developed an accelerator that leverages AI/ML to provide the optimal delivery route to be taken during last mile delivery. We have picked the CVRPTW problem (Vehicle routing problem, considering customers’ promised time window and the maximum capacity of each vehicle) and added new constraints around demand quantity and clusters of pickup/drop locations. We tried different approaches and, in the end, used a combination of heuristics algorithms and AI-based optimization to come up with a viable solution to the problem.

Our solution allows the route planner to configure input parameters such as:

• vehicles’ capacities
• drivers’ availability
• deliveries
  

It is designed to meet the maximum delivery time SLAs:

• while reducing the total distance driven (and hence the total cost)
• and increasing the route density and drop size by clubbing the deliveries based on location, proximity, delivery time window, and order specifications
  

It helps achieve better load utilization by optimizing the number of vehicles needed and utilizing their capacities. The solution integrates with dynamic map API from Mapbox (Google API can also be used) to provide integrated maps with real-time distance calculation. To summarize, this solution can act as a great starting point for the logistics companies to expand and optimize the last mile delivery network.

Try our AI-ML powered solution to optimize your last mile delivery. To learn more about how AI can make your supply chain management and logistics operations smarter, check out this webinar or reach out to us.