Freight Matching’s Data Challenge

Thoughts from John Larkin

John Larkin recently recorded his thoughts on the state of digital freight matching and the opportunity that DFM Data Corp is working to seize. As a caveat, it is not a verbatim transcript. It’s been paraphrased, cleaned up and slightly reorganized to help his insights flow smoothly.

JL: The first problem with their data is that it’s old. I don’t mean ancient, but maybe a week old. I’m afraid the market changes so rapidly that what happened a week ago doesn’t necessarily reflect what might happen tomorrow or the next day. Summing up that first issue, no one has all the data. You have your own data to the extent you’re smart enough to keep it and you have data from these two sources which everybody has access to. But nobody has all the data and you need all the data to make better predictions about what pricing might be tomorrow or the next day or the day after that. As a result of not having all the data or it being outdated, is that you can make incorrect projections that are sub-optimal and mislead people.

Okay, Now the other issue is that there is a lot of double, triple and quadruple counting. Someone may offer a load to multiple brokers and that load will get posted on four or five different DFM working sites. Somebody will move that load at 10 o’clock in the morning and the other 4 or 5 Brokers will be working to find a carrier for that load at the desired price for two or three more hours. Then once they find a Carrier, it has already moved.

Michael [Darden]’s vision is to make all the data available to everybody and to make that data anonymized. So you can’t steal someone else’s customer, but you can still eliminate all the duplicates, triplicates and quadruplicates. You have all the data and eliminate all the double and triple counting.

That will give you a clean data set that’s current. Then, the way you differentiate yourself as the DFM is not by having the best data or the cleanest data, because everybody has that. You differentiate yourself by what you do with that data in terms of applying analytics and forecasting techniques to add value for your customers
so that you can help them optimize their supply chains.

Q: Can you give an example of how a DFM might look ahead? How you might forecast using this data, say if you see prices going up or down or a specific lane changing? Maybe you see that a price has changed to make a specific lane more economical than another lane.

JL: Well, the clean data allows you to make more accurate predictions of the future because you have an ap Old data set and it’s clean. So the conclusions you draw from that data set should be more accurate. For example, you are predicting that the rate to move a load of freight from Chattanooga to Chicago is $1.95 today, $1.80 tomorrow and $1.70 on Wednesday. If its not a service-sensitive load, you might just wait till Wednesday to move it.

That could impact even the mode. It might make sense on certain days of the week to move the freight via inter modal. It might make sense to move freight other days to be a full truckload. It might make sense on other days to move the load with a team expedited carrier. That all really depends on what’s happened historically in that lane in that direction.

There are so many data points here because you’ve got 50 loads that move today in one direction in the lane and you have 75 loads that move in the other direction in that lane. But there are thousands of lanes, origin-destination pairs, and all of tomorrow’s data and the day after’s data. So it’s a massive data collection, organization and analysis problem.

Michael [Darden]’s vision is to clean up the data so you can differentiate yourself based on the quality of your analysis and the quality of your recommendations to your customers, many of which may be delivered to the customer without any human interaction.

Now that the digital freight matching system has great data, it can make those recommendations to the customer without human intervention. In those cases where something doesn’t look right or smell right, you can kick it out to a human being to investigate whether or not something went awry.

But there’s potentially a lot less human intervention. If you went to a traditional brokerage firm next week, let’s say, it would not be uncommon to see a bunch of young people … let’s say forget about Covid-19 for a minute … a bunch of young people sitting in cubicles with six screens in front of them and a headset on. They’re kind of working based on experience and their their raw intelligence. What the DFM data can do is allow a DFM to work instantaneously and mechanically through their technology. So in theory, you ought to be able to handle a lot more loads per person than you can today.

When something goes wrong, say a truck breaks down, it needs to be re powered. Usually there’s going to be a human involved there. For the routine stuff, a lot could be mechanized … but it can’t be mechanized if the data isn’t any good.

Q: Can you talk a little as I understand it, the digital freight market is is fairly new emerging at this point. Can you talk a little about how you see it growing, especially with this sort of technology.

JL: Well, historically, the traditional brokers were doing it with three-by-five cards and coffee cans. Then over time, they’ve automated many of the functions.
Let’s say there are 90 functions involved. 90 steps to a brokerage process. Initially, maybe 10 of those were automated, and then it was 20. Then it’s 30. Today it’s maybe 60 or 65 in some cases … maybe 70 in the most automated case.

But there’s still a number of functions that have to be handled manually and a lot of that is due to the fact that you have to call people that are not plugged in to the technology … mostly small carriers. You may have to call eight or nine or 10 of them on average before you find somebody can handle the load. So part of what is needed is not only to install the technology at the broker or the DFM, but also with the drivers/owner-operator/independent contractors/trucking companies.

Scroll to Top