Imagine that an empty UFO has plopped down to earth. It’s an impressive machine, but there are no cute E.T.s around to tell you how it works. Now imagine that you need to reverse-engineer it, build a new one that can reproduce the flight path, and then teach someone else how to fly it for lightyears to come.
That’s pretty much what was asked of Dr. Alex Rodrigues at the University of Tennessee’s Global Supply Chain Institute (GSCI). But instead of a project out of science fiction, he was asked to redesign two of the most important measurement tools of the transportation industry: the Cass Truckload Linehaul Index and the Cass Intermodal Price Index.
“I had extensive experience in terms of such research projects and working with the business community,” says Dr. Rodrigues. “But, here at UT, it is the first time leading this kind of project and being involved with real-life businesses and real-life data.”
The indexes are published by Cass Information Systems, Inc., the largest freight payment provider in the United States. Cass populates the indexes with data derived every month from the invoices they manage for over 350 companies. That’s about 36 million invoices that total $28 billion every year.
Both indexes have operated since 2011 and feature data stretching back to 2005, and Cass’s data is widely referenced not just in industry publications but also in broader financial media like The Wall Street Journal and Forbes. Useful for various types of forecasting under normal conditions, the indexes (and Cass’s accompanying analysis with Stifel) become vital when economic conditions are volatile and uncertain.
Each index provides a different kind of cost and activity estimate month-to-month. The Truckload Linehaul Index indicates how much it costs, per mile traveled, to transport goods in a semi-truck over a distance of at least 100 miles, and it excludes the notoriously volatile cost of fuel to give a more stable, accurate picture of trending prices. When cargo moves between multiple transports (e.g. from a train to a container ship to a truck), the Intermodal Price Index tracks that activity, inclusive of fuel costs.
The methodologies for both indexes had been developed and maintained by a third-party provider over the years, but Cass decided to take the indexes in-house.
“[Cass] reached out to us, we accepted the challenge and initiated a partnership,” says Dr. Rodrigues. “Here we had the ‘black box’ that was left from that third-party company and we have to replicate their methodology.” The UFO had landed.
Dr. Alex Rodrigues of UT used multiple linear regression modeling to revise two of the most important commercial indexes in the transportation industry. As the centerpiece of a long-term partnership between Cass and GSCI, the recalculation represented a substantial undertaking for Dr. Rodrigues and his collaborators at Cass. First, the complex task of recreating nearly 15 years of data with dizzying accuracy. This would require a comprehensive methodology aligned with the academic literature surrounding projection modeling and the transportation industry. And then, second, operationalizing that system to be simple and tangible enough for Cass to use every single month. How do you do that?
For Dr. Rodrigues, the early days of the project were spent conducting an extensive review of the academic literature and analyzing public information from the Department of Transportation, the Federal Reserve Bank, and other aggregated macroeconomic data sets.
This led to a key insight for Dr. Rodrigues. The data provided by Cass’s invoices alone “could not have built the model. It couldn’t explain, historically, the index,” he says. “It was an indication that the third-party company was using other macroeconomic variables in addition to the Cass internal variables.” What emerged were additional factors that Dr. Rodrigues attributed to “the art of modeling.”
He explained that the model needed to have “variables that are able to capture trends, seasonality, cycle- and recession-type of aspects of the time series. The literature review clearly indicated that if you’re trying to measure freight activity, these are things that are associated with these activities.”Finally, there was the question of specific modeling techniques. There was, according to Dr. Rodrigues, “this kind of overall, umbrella concern that the methodology could not be extremely complex.” Though he considered the application of dynamic modeling via neural networks, he ultimately arrived at a technique familiar to his undergraduate students at UT: multiple linear regression modeling.
As a relatively simpler model, it would be more amenable to the analytical needs of the team at Cass. “The pace that they need to update their index calculation, and also generate some type of valuable qualitative interpretation, insights, is very fast.” He says. “It’s very intense: it’s every month. So, the need for the automation and precision of the model was higher.”
For Dr. Rodrigues, though, the most important measure of any model, of course, is accuracy. So he and his colleagues at Cass were delighted when, upon combining all the variables in their model, they were able to replicate the indexes’ historical data with a margin of error equal to less than one percent. “If only I had this same type of precision with the stock market!” Dr. Rodrigues jokes.
Once the project was completed in November 2019, Cass’s more than 8,000 subscribers were able to take advantage of the refined indexes. Perhaps more than ever, as the world faces unprecedented economic turmoil, tools to plan capacity management are vital for every industry. As noted in Cass’s October 2019 report, “tracking the volume and velocity of those goods has proven to be one of the most reliable methods of predicting change because of the adequate amount of forewarning that exists.”
To be sure, Dr. Rodrigues says, adapting capacity in the transportation industry is especially difficult. “It’s capital intensive,” he says, “and decisions are in the medium- to long-term, for the adoption of new machines, technologies, processes, and information systems. They require investment or partnerships to constantly balance their resources with fluctuations in the transportation market requirements.”
“That increases, I would say, the importance of such indexes.”