Multi-Stop Route Planning in Distribution

The most common theoretical framework for solving a multi-stop delivery problem from the same location (s) is known as a Vehicle Routing Problem (VRP).

One of the specifications of all VRP’s is that they are stochastic (random) in nature, so every time the user runs it, the route plan is going to be randomly compiled. The majority of distribution firms are using routing software or do manual routing using what’s called a “static routes” system. Companies first create “static” routes and then try to squeeze in additional orders or manually reroute those orders with decreased demand on a daily or weekly basis.

This system was created long before the internet and PCs as well. Theoretically, building an optimal route plan for a medium-size distribution company with 1000 deliveries (nodes) is hard. It should be done by solving a Capacitate Vehicle Routing Problem with Time-Windows. This is a time-complex problem. It also requires substantial resources and computational power. While static routes are easier to compile, medium to large companies lose hundreds of thousands and even millions of dollars for underused capacity and overdriven miles.

Pros of static multi-stop route planning

• Drivers assigned to “static” routes so can more easily get to a location
• Drivers deal with the same customers. So, they know service times and have other customer-specific information
• Drivers know rush hours and parking specifications for locations

Cons of static multi-stop route planning

• Lower truck utilization
• More miles are driven
• Harder for planners to plan
• Less dynamic capabilities

There are a number of obstacles that make the “static routes” approach hard to implement. They also amplify the shortcomings of the approach, making it even more ineffective:

• Demand is not static for the same customers
• Orders from new customers never served before
• No orders from existing customers

Static routes create more inefficiencies during times of fluctuating demand or when the company is at a higher growth stage. It also can create inefficiencies when a company adds more depots or comes out from the M&E process.

Less Platform and stochastic multi-stop route planning

We have created daily route plans using Less® Platform’s weekly sample data and compared it to our customer’s historical delivery data, which used static routes (Figure 1).

Figure 1

We have also implemented tests for other periods and customers. The result is that the Less® Platforms stochastic optimization produced 10-30% less mileage compared to historical data.

Mixed model

The main issue regarding planning in distribution operations is how to find a balanced model when the real-life benefits of the static routing will sustain the status quo, but stochastic optimization will lead to better operational results. We have engineered a model that combines additional constraints to our CVRPTW algorithms.

• Assigning specific customers to specific drivers
• Using advanced sequencing algorithms
• Appling supervised Machine Learning (ML) algorithm to train on driver-specific data

While those constraints reduce results further from the optimality by about 3-7%, they still allow the combination of both worlds and secure hundreds of thousands of saved miles.

Less® Platform can turn your data into loads within 10 minutes and compare it with your historical results! You can also test all the necessary features for your logistic model and consult with our experts on
your challenges.

Author
Vardan Markosyan is the CEO at Less® Platform
MBA from the University of Chicago Booth School of Business
Ph.D. in Economics from the Institute of Economy of NAS RA
He spent decades of research and consultancy on business process optimization and system design