Every day, shipping containers carry imports and exports such as fruit, car parts, clothing, and furniture into and out of the United States. A product’s journey from ship to consumer consists of many logistical steps, one of which is a process called drayage (a 19th century term related to the sturdy, horse-drawn wagons called “drays” that carried such goods). Orchestrating efficient dray moves, or the assignments of trucks transporting containers between ports and inland destinations, is a prevalent operational challenge.
“This problem exists in various forms in the transportation business,” explained Zahir Balaporia, Decision Scientist for MatchBack Systems. “But in the world of ocean containers, there are certain nuances to the problem that make it harder to solve, which the industry has historically struggled with.”
The classic case is when an import is delivered to a customer by a truck that then returns to the port empty, while another truck takes an empty container to an exporter to pick up an export and bring it back to the same port. If the availability time windows of the import and the export overlap and the geography makes sense, these two independent moves using two trucks could be planned and executed as a tour using only one truck. At its simplest, this pairing of an import with an export to share a container move is called a street turn (or matchback).
While the concept is simple, it gets very complicated very quickly when considering thousands of imports and exports simultaneously, at multiple ports, with partially overlapping geography and time windows. Combining that with equipment sizing constraints (20-foot or 40-foot container, etc.), freight characteristics (hazardous material, food grade, etc.), and multiple trucking companies moving the same freight at different costs and capacity constraints, makes the problem explode in terms of combinatorial complexity.
Reducing this complexity into executable plans is the solution MatchBack Systems is delivering for the industry. The solution reduces transportation costs and improves equipment utilization for customers, while simultaneously improving revenue per truck per day for the trucking companies moving the freight.
“We’re one of the first in the market to strategically automate street turns with a holistic approach, ensuring we optimize for the best overall solution,” said Tammy Leurquin, COO of MatchBack Systems. Leurquin oversees the technology portfolio supporting MatchBack Systems’ mission, which includes mathematical optimization and decision science projects led by Balaporia.
When evaluating optimization solutions to generate efficient tours through street turns, Leurquin and Balaporia recognized their success would depend upon two things: what they run and how they run it. In other words, selecting a modeling framework and optimization solver was just part of the puzzle. Having the tools and infrastructure to efficiently operationalize their solutions was a significant part. They needed the ability to get optimization models up and running quickly, release significant capabilities frequently, flexibly adapt to change, and readily integrate with existing engineering architectures and workflows.
“If we’d done this the traditional way, we’d be standing up servers, maintaining them, writing the code for orchestration — all of that,” explained Balaporia. Instead, Balaporia was able to collaborate with his software engineering counterpart to readily integrate Nextmv’s platform consisting of deployment infrastructure, optimization integrations, exposed model configuration, and model monitoring. Everything was all ready to go and minimized the requirements of the engineering team to support a robust optimization solution.
Balaporia and team have also been able to iterate and improve upon their optimization offerings. Within Nextmv, they’ve been able to stand up and swap out different modeling and solver combinations. They’ve also explored the model testing and experimentation tools to help with plan exploration and even policy development. In one instance, Balaporia explored the relationship between cost savings and how quickly to match imports with exports. For example, if an import is arriving in two weeks and an export is available in two weeks, do you make the match and move on? Or do you wait to see if a more advantageous match becomes available?
Balaporia experimented with the different scenarios to see how the cost savings played out under different time frames and circumstances. If waiting until 2 days before only saves you $4, it makes sense to dispatch the load earlier. But if waiting saves you $400,000, there’s more incentive to wait. “We were able to tell a more quantitative story around why waiting could make more sense,” Balaporia said.
Created to facilitate collaboration, the Nextmv platform has enabled the MatchBack Systems team and Nextmv to work together to scale and iterate on model development as well as quickly onboard new teammates. “The math is not really the hard part anymore, it’s all the other stuff around it that is: speed to deployment and ability to interface with the model efficiently,” said Balaporia.
“Optimization will continue to be on our roadmap — it’s an ever-living, breathing thing,” explained Leurquin. “Nextmv is an essential tool that enables us to leverage optimization alongside other approaches to deliver more unique value to our customers, to help them save money, eliminate inefficiencies, reduce CO2 emissions, and meet sustainability goals. Nextmv plays a part in our ability to accomplish this in rapid fashion with a lot of confidence.”
