Machine learning tools are often touted as mechanisms to drive next-level benefits in freight movement, but current use cases are more feet-on-the-ground than head-in-the-cloud applications.
Natural language processing
The two software companies — Trucker Tools and Parade — have released new tools in recent weeks based on an offshoot of machine learning and artificial intelligence (AI) called natural language processing (NLP).
These tools essentially take information locked away in bulk emails and conversations and structure the information so it can be utilized more effectively by brokers to provide capacity to shippers. For instance, Trucker Tools earlier this autumn began performing such activities for its broker customers.
The company set out to tap into capacity locked away in email blasts sent daily from carriers to brokers showing what capacity is available by location at the start of the day. It’s a well-worn way for carriers to try to find loads for available capacity, but Trucker Tools, in its experience shadowing broker customers, found that the emails were hardly being used for a couple reasons.
First, it’s a time-intensive and laborious process to match the available capacity in a static email to loads available in the broker’s system. And second, even if there is a match, there is often too much latency between the time the email with available capacity is sent out and when the broker finds a load that matches. That results in wasted communication between the broker and carrier, only for the broker to learn the capacity is no longer available.
The second problem is magnified by the reality that the carrier’s email blast goes to multiple brokers, all scanning the same set of potentially outdated capacity. To combat the problem, Trucker Tools started applying NLP to the emails its broker customers receive to find date, location, and equipment type matches between available capacity and available loads in its Smart Capacity system, to tie those elements together “in a structural way,” said Murali Yellepeddy, chief technology officer at Trucker Tools.
“Now this capacity is structured and in the system where it can be used,” Yellepeddy said. “We take that email and turn it around, find matching loads, and send those to the carrier. Think about it from the carrier perspective — they send these emails into a black hole that’s being used less and less. That has a huge impact on quality of the calls coming into the carrier. When that email hits the broker, the data is absolutely true at that moment, so the likelihood of that match will be extremely high. So we’ve gone from the data being ignored to being available, and it can be pushed into the broker’s TMS [transportation management system], and also into the matching algorithm.”
By introducing unstructured information in an email into a structured environment, Yellepeddy said Trucker Tools is also able to hone its system to better serve its customers.
“Brokers and carriers are having these conversations outside a system, not through a TMS or any other system the broker uses,” he said. “That’s painful for the broker. They’ve not really recorded this conversation in any system where it can be used later, especially in terms of the amount of turnover some of these brokers have. That information is staying with the employee. Instead of trying to get everyone to use a system, we should look for avenues where we take advantage of communications they’re doing in natural form.”
If Trucker Tools finds any matching loads from the capacity gleaned from the carrier capacity emails, it automatically sends a response to the carrier, but only to the brokers’ in-network carriers. Those are carriers that are vetted by broker within the Trucker Tools system.
“The advantage we have is the opportunity to clean up this data,” Yellepeddy said. “The current approach breaks down when a carrier tries to use the return email three hours later. We get load data into the TMS in almost real time, and we’re cleaning up this data several times a day.”
Opportunities from NLP tied to Smart Capacity
Michelle Potter, senior director of strategic development for England Logistics, said the NLP capability tied to Smart Capacity creates opportunities that the brokerage business would have had to dedicate resources to uncover.
“We have all these different data sources,” she said. “The load boards gets stale very quickly. If you even take time to enter the emails into your system, you wonder, is it still valid by that time? You’re hoping the email went to the person who has the need for a piece of equipment. But unless you’ve hired someone who went through those emails, it’s only good for people coming in at 5 a.m. to cover loads at 6:30 am.
“Carriers are pushing out availability during the day. But they all have variations. Some will send a large list at the beginning of day. Some will send smaller portions throughout the day based on available capacity. Some carriers, you’ll get multiple emails. Some will send you from their regional offices based on regional accountability.
England automatically forwards availability emails to Trucker Tools or has carriers send them directly to the software provider. “However it gets there, Smart Capacity consumes that information and uses that as another means to verify the capacity that they’ve determined through other sources.”
The use of machine learning to determine capacity that might actually exist is a functional example of the potential of such systems.
“What’s really interesting to me is that you can use these big data sets and examine patterns,” Anshu Prasad, founder and CEO of LogisticsExchange, said Oct. 30 on a panel about machine learning and AI at the JOC Logistics Technology Conference in Las Vegas. “The machine can interrogate the data without using structured hypotheses. What are you doing with the analytics that will prompt a change in behavior? If you’ve been managing freight the past year, you probably don’t have the bandwidth to entertain new ideas. It’s taking problems off their plate; that’s valuable. And how can you bring new options to them that they wouldn’t have seen before?”
Prasad works with shippers on making procurement more effective through the use of machine learning and AI.
“Putting stuff out there in a more structured way, applying logic learned over time,” he said. “We’re working on the digital contracting end of the problem. As your forecast starts to come into focus, points of convergence tell you those points of freight are going to move. We’re starting to see some interesting nuggets that we can feed back to our customers.
“There’s always 100 things on the list. The challenge is to come back to the five or 10 things that matter right now. The beer budget customer has to decide what’s worth solving right now. We need to have that practical, actionable thing you’re worried about today as part of our solution. If we don’t, we don’t get the luxury of working with you on neater things down the road.
AI’s potential impact in logistics
Anthony Sutardja, CEO and co-founder of the freight transportation software provider Parade, said the primary reason AI is so potentially powerful in logistics is that “systems are not designed to scale relationships. There are still lots of phone calls, emails, and tribal knowledge.”
Sutardja said Parade, which works with several broker-oriented TMSs, including McLeod, MercuryGate, Transport Pro, and TMW, helps “make more out of their email and TMS data and generate more accurate predictions on what carriers are trying to do.”
Second, Parade aims to give brokers business intelligence to use that capacity data. “Not just where and when that capacity is, but what is my relationship to them,” Sutardja said. “All the things that capture the business relationship with the actual carrier and combining that data into a central dashboard.”
The last piece is leveraging that data into automated engagement with the carrier.
“The carrier sends an email that he’s got a truck going from LA to Seattle,” Sutardja said. “We’ll engage on behalf of the broker or transportation department to facilitate better opportunities to engage with the carrier or the carrier’s sales rep. These carrier reps get hundreds of emails a day about truck availability. They can manage some of these relationships, but not all of them at scale. The broker sales reps need to make 150 calls a day to cover their loads. So it’s about how to reduce that, or how to make those calls more productive. We use our predictive analysis to make sure the first three calls are to the ‘money carriers,’ the ones where you’re more likely to get hits.”
Satardja said there are benefits for the carrier to such an approach as well.
“On the carrier side, traditional tools take a shotgun approach,” he said. “The shipper/broker will spam out loads all day. It’s so transactional; they don’t bite, and they don’t have a high success rate on those types of transactions. So it’s understanding actual dispatcher behavior at the carrier. Hit them once a day, crafted in a way to see the opportunities, [then] they can dive in [and] give feedback, which feeds into our data. If they like a load, [through] the email engagement we start picking up the patterns.”
What’s inhibited the carrier-broker relationship is what Satardja calls “a system design problem. We measure the clicks and opens that the traditional broker hasn’t thought of in the way we have. We help them utilize that data to engage more automatically. They’re all struggling on procurement.”
Users of McLeod’s broker TMS, for instance, have access to a white-labeled version of Parade called Capacity Creator. Satardja said Parade is currently focused on the domestic third-party logistics (3PL) and freight brokerage market, but he sees opportunities to work with shippers with spot tendering needs.