Machine Learning for Transportation Execution and Visibility

Transportation Execution and VisibilityI am kicking off some new research on the Transportation Execution and Visibility Systems market in the next week or so. This will be an update on a study Steve Banker did a few years ago, which looked mostly at the execution space, and how the Uber for Freight model was disrupting the market. According to ARC’s definition, transportation execution (TE) systems & marketplaces allow shippers to connect to multiple carriers and then tender, track, and pay using the software or marketplace.  Many solutions include analytics. In addition to tendering, these solutions may garner module revenues or transaction fees from closely related transportation services such as dock scheduling or trailer management. ARC’s definition of TE solutions does NOT include load boards or systems with robust optimization capabilities (we cover the latter in our transportation management system study).

Transportation Execution and Visibility

One of the hot areas within this market is visibility. In this study, I am specifically looking at transportation visibility. And by that, I mean can you track assets in real-time across multiple modes to meet delivery schedules. According to my latest omni-channel research, there is a serious gap when it comes to visibility with real-time carrier updates of merchandise across an organization’s entire distribution network. Just how big is this gap? For inbound shipments, inventory is sent from a supplier to either the warehouse or to a brick-and-mortar store. Obtaining visibility into inbound shipments tends to be more problematic for many businesses, since this requires real-time updates from both internal and external (supplier) systems. According to survey respondents, 72.5 percent of respondents have real-time visibility of merchandise going from the supplier to the warehouse. However, only 45.9 percent have visibility into shipments going from the supplier to the store. Shipping to the store adds another layer of complexity, as the merchandise does not go through the central hub and often involves much lower quantities than those going to the main warehouse. The fact that nearly 30 percent of respondents do not have real-time visibility from the supplier to the warehouse shows that a problem exists as well.

Earlier this week, Steve Banker and I spoke to Jim Hayden, CTO, and Vicki Warker, CMO, from Savi, a provider of technology solutions for comprehensive asset tracking and supply chain visibility. The basic idea is that the company is able to capture structured and unstructured data from a variety of sources to give companies a better understanding of where their assets are and when they can expect them to arrive. This data includes data from carriers and third party aggregators, including crime data and port strike data. Surprisingly, weather-related data is not included as the company has found that it has very little impact on planning.

Location Data

As Jim Hayden explained, the company uses sensory data to track goods across multiple modes. The company sees itself as having a “container-centric” view of the world – rather than tracking drivers or trucks, they track the container through all modes. The solution uses satellite GPS for location, which is transmitted on a public wireless network. Satellite line of site gives accurate GPS location; if there is something blocking the line of sight, the company uses cell phone towers to triangulate the location of the container. For ocean cargo, the solution uses Automatic Identification System (AIS) data to determine the location. While outages in the AIS data occur in the middle of the ocean or when traveling through high risk areas, it is a good indicator of the vessel’s actual location. Savi receives 20 million vessel location updates every day, so it is a good source of data to use with machine learning for accurate arrival times.

Machine Learning

Hayden explained that when it comes to machine learning, no one model works for any mode. For that reason, the company has developed a big data solution for any time a shipment is created. Savi makes a new model for every leg for different carriers; they have tens of thousands of models deployed, with variations based on lanes, carriers, modes, time of day, and a variety of other criteria. An example of applying machine learning is for land-based shipments in North America, where drivers need to comply with hours of service regulations. While you may not know how long they have been driving when they pick up a shipment, Savi has a model to make a best estimate by lane and carrier on how much rest time will be needed in transit, and uses this information to determine the appropriate ETA.

Visibility across the entire distribution network is a critical component of the transportation execution and visibility systems market. Currently, too many companies are lagging in their efforts to accurately track shipments and provide a realistic ETA. This can lead to lost sales, higher total landed costs, and reduced customer satisfaction. However, as companies continue to improve their visibility efforts, these negative results can be avoided, and business can continue to grow.