The ELD – better trucking through data science

The ELD – better trucking through data science

We have entered the age of the electronic logging device (ELD), and the vast potential of these tiny digital recorders to enable companies to visualize, quantify, and address complex trucking problems is beginning to take shape.

For example, for-hire drivers only make money when they’re rolling with freight. And yet, even in today’s capacity-constrained market, drivers are often compelled to remain idle at pickup and delivery points for hours. Anecdotally, drivers tell worst-case-scenario stories of waiting for days at customer locations before proceeding with their cargo. At a recent industry event, I surveyed the audience of trucking professionals by asking the question: how many hours per day do your drivers spend on duty, but not driving? The most popular answer selected was three hours, followed by frustrated grumbles from the crowd.

In my research, ELD data give us context to understand this frustration at a deeper level. My own ELD data set, comprising about 1,500 drivers from 2016 to 2018, shows with statistical significance that drivers are spending more time on duty (not driving) now than they used to. We have also analyzed how particular days of the week, specific customers, and delivery sequences correlate with driver wait times and under-utilization. These insights will allow us to diagnose what is causing the wheels to slow underneath American truck drivers.

Visualizing inefficiencies

But the opportunity is not just in describing problems like these; ELD data also give us the ability to envision solutions.

ELD data are a digital, transferable record about a critical attribute of the rolling resource: how many hours the driver has left to legally work that day. Surprisingly, this figure is often absent from pickup and delivery decisions at warehouses and distribution centers (DCs). Although many companies do their best to synchronize inbound and outbound loads, many DCs are often working with incomplete information and, as a consequence, are not optimally sequencing their pickups and deliveries. Anecdotally, shippers and receivers tell me that when a driver waits for a day or more, it is often because the individual’s legal hours of service (HOS) for that day have run out as they approached the dock for their appointment, which forces rescheduling for the following day.

The current situation is not unlike trying to schedule runways for landing airplanes at busy airports without knowing how much fuel is left in each plane’s tank. In this sense, the ELD is not just a regulatory burden; it could also become a mission-critical fuel gauge.

Big data issues

However, to realize the full potential of ELD technology, we must appreciate the scale and complexity of the data sets it generates. For a moderate-sized trucking company such as the ones I’m working with, ELDs generate approximately 100,000 rows of data per week — a size and rate that quickly crashes most computer spreadsheet applications. Data sets generated by ELD-enabled fleets are indeed ‘big data,’ making real-time processing and synchronization a technical challenge, and parsing and understanding them a specialist’s task. Moreover, the data sets are complex: ELD internal clocks can start and re-start data recording (i.e., making new rows) in myriad ways depending on time clock, geographic position, and engine status. Moreover, the HOS rules that they are meant to monitor are multifaceted, with relevant 24-hour, 60-hour, and 70-hour cycles. At the MIT FreightLab, we’ve made early headway in cleaning and organizing such data sets, but there is much more data cleaning left to do.

There are serious, non-technical challenges to leveraging ELD data too. Actors such as trucking companies and DC schedulers fulfill different roles, and the industry needs to develop new types of alliances and partnerships to align incentives for the exchange of data. The legal complexities of sharing and using ELD data also must be addressed. In the event of disputes such as traffic accidents, HOS data can be used to assign liability. In addition, drivers are understandably suspicious of devices that monitor their working lives, and these misgivings must be allayed.

Worthwhile endeavor

These are significant challenges, but overcoming them will be well worth the effort as ELD-driven analyses shed new light on truck operations.

Despite deep-seated misgivings about ELD devices, they will unlock a treasure trove of data.

In the near term, ELD projects such as mine will help bolster drivers’ cases for better treatment and hence support driver hiring and retention programs, and enable more efficient over-the-road freight transportation through data science.

David Correll is a research scientist at the MIT Center for Transportation & Logistics. Contact him at: dcorrell@mit.edu.