It’s tempting to write off university research as too ivory tower, especially in such a pragmatic business as freight management. But in our experience academic studies that are tied to real-world issues can pay back hugely.
We’ve sponsored a number of projects at MIT Center for Transportation & Logistics (MIT CTL) over the years. Below are a few of these projects:
Even though we did not sponsor this research, in today’s post, we’d like to focus on a MIT CTL project titled “The Effects of Truckload Freight Assessment Methods on Carrier Capacity and Pricing” carried out by Lukasz Kafarski and David Allen Caruso Jr. for their 2012 MIT Supply Chain Management Program master’s thesis. MIT CTL Executive Director, Dr. Chris Caplice, was the thesis advisor.
Don’t let the uninspiring title fool you; this piece of research has a lot of practical value for supply chain managers who are prepared to think about how they can apply it.
The authors worked with Niagara Bottling, LLC, a C.H. Robinson customer, that at the time had six production facilities across the US. They analyzed 12 months of shipment transactional data and five months of shipment tender data generated by the company’s freight operations in 41 high-volume lanes.
Kafarski and Caruso investigated how carrier assignment methods can impact available capacity and pricing. For example, they studied how lane aggregation – where companies combine similar lanes and award them as a single lane with a broader scope – can increase operational efficiency, and reduce variability as well as the complexity of freight networks. Here are some notable observations from the study.
Short Haul Aggregation: The best way to aggregate for moves less than 100 miles is not by zip code or zip code range, but by length of haul using rings in 10 mile increments. From a price standpoint, carriers were more concerned about distance than they were about direction when moves of less than 100 miles were involved. By using the ring aggregation model mentioned above, the shipper had achieved the same level of robustness (defined as the system’s ability to absorb demand fluctuations on specific lanes without impacting service levels or price) exhibited in longer haul lanes where other means of aggregation were used.
Long Haul Aggregation: We looked at this subject a few years ago in a thesis paper titled “The Impact of Bidding Aggregation Levels on Truckload Rates” by Julia Collins and Ryan Quinlan that we sponsored at MIT. This project and other research indicated that there were at least two tradeoffs in play. Lanes with higher volume tend to have lower rates so we suspected that aggregation is one way to capture higher volume. At the same time, we knew that more specific lanes tend to generate better rates, so if our zones got too big we might lose the aggregation effect.
Kafarski and Caruso highlighted that it is not just volume that matters but how predictable a lane is. When high-volume and low-volume lanes are combined, you reduced overall variability in the lane.
This observation should not be surprising; it is the same mechanism one sees when forecasting demand. A total forecast can be fairly accurate but starts to break down when the figures are more specific. Below is an example of a customer we recently did a procurement event for. The green line represents the variability of their largest point-to- point (PTP) lane in this aggregated lane. The purple line shows the variability of the whole lane made up of the one large lane and several smaller lanes all from the same origin and going to the same three-digit destination. As can be seen from the chart, there is significantly less variation in the larger, combined lane.
100 to 300 miles Increased Variability: This may be the most intriguing finding from the study. The researchers recognized that movements in this mileage range had the most variability in rates. Moves over 300 miles were typically handled by irregular route carriers who were not necessarily trying to return immediately to their original shipping point. However, moves between 100 and 300 miles were completed by a combination of players, and there was much more diversity in the amount of dead head involved in a move between carriers. These differences accounted for the variations in price.
This observation is so intriguing that we decided to determine whether it holds up for a larger group of shippers. We’re currently sponsoring research with an MIT CTL graduate student to explore this question. If the variability does indeed apply for the larger sample, it will have important implications for how and where a shipper should focus their efforts when trying to control transportation costs.
Results like these underline why academic research can deliver real value and enhance competitiveness. In addition to producing hard data and analytical insights, it can spark ideas that ultimately make your freight network more efficient. And that is a very practical, non-ivory tower, outcome.