Real-Time Data is Transforming the Trucking Industry

MICHAEL KRIGSMAN: We’re discussing machine learning and data to make the trucking industry more efficient. Our guest is Dorothy Li, the chief technology officer of Convoy. Trucking is one of our most important backbone industries. It’s fascinating to hear how machine learning data and automation can transform this traditional business into a modern powerhouse using the latest technology. Hey, Dorothy. How are you today? 

DOROTHY LI: I’m doing great. Thanks for having me, Michael. 

MICHAEL KRIGSMAN: Dorothy, it’s great to speak with you. Tell us about Convoy and about your role as CTO. 

DOROTHY LI: Well, Convoy is in short is the first and most efficient digital freight network in the US. And what that really means is that Convoy started this movement in efficient freight back in 2015. And since that first day, we have really been driven by that desire to solve the toughest problems in freight and efficiency in this massive $800 billion industry. 

And my role as a CTO is to bring technology to transform this industry. And a lot of the problems in the freight network in freight today can be solved with technology with machine learning and with data. It’s really at its core to optimize how millions of truckloads are moved around the country. 

MICHAEL KRIGSMAN: So the focus is on bringing great efficiency to this enormous industry trucking, and you’re relying on data and technology and machine learning. Dorothy, as you look at the trucking industry, what are the core challenges that you’re trying to address? 

DOROTHY LI: I think there are, I would say maybe two core challenges that we’re trying to address. The number one core challenge is this idea of empty miles, or in other words, a wasted miles that are driven when a trucker is running without any loads to haul. And if you put us in some perspective, the industry annually has about 200 billion miles the trucker is running. Almost 35% of them are empty miles. That is 60 billion miles wasted. 

And if we can improve that efficiency, that we can– a lot of trees would be equivalent to a lot of trees getting planted. So that’s problem number one. And the second problem is around the trucking profession itself, right? We hear today about labor shortage, trucking shortage is really at the center of why we’re seeing a lot of the supply chain issues today. 

And to make that profession a much more sustainable, we also need to improve efficiencies for truckers. Reduce the time they spend waiting, make facilities a lot more convenient to truckers, and just in general make this a– make truckers lives better. 

MICHAEL KRIGSMAN: So I totally get the problem of empty miles, but how is empty miles a data problem? How do you solve that with data? 

DOROTHY LI: That’s a great question. Data is really at the heart of solving this problem. When you think about it, the challenge of empty miles is because we really don’t have a lot of visibility into– well, A, we don’t have a lot of visibility into how routes are optimized– can be optimized. And two, we also don’t have visibility in the traditional approach of where a truck is, how the trucker is spending his time. 

So if you can provide data points on knowing every point if a trailer is empty or not and if the trucker– where the trucker is on the road and where the route condition is. With all this data, we can really start to devise smarter ways of routing and also batching such that when one load is unloaded. He can pick up and create– load another trailer on the truck right away. 

MICHAEL KRIGSMAN: How does Convoy know the status of the truck? I’m sure that it’s not the case that the trucker arrives in Des Moines, Iowa picks up the phone and calls Dorothy and says, OK, Dorothy, I’ve unloaded. I’m ready to roll. What have you got for me? I’m sure that’s not happening. 

DOROTHY LI: That is not happening. 

[LAUGHS] 

There’s a couple of ways that we know where a truck is. One is with Convoy, every trucker uses an app and these apps as we know on a smartphone has geolocation tracking services. So with that, we know where he or she is. If a trucker has a Convoy trailer and these trailers are smart trailers, they are equipped with IoT sensors that have internet connectivity. That’s another way we know where the trailer is whether it’s empty. 

MICHAEL KRIGSMAN: So to solve the problem, it’s not enough to know that the trailer is empty. Obviously, that’s point one. But you have to know that there is another load within some reasonable distance that the trucker can pick up. So how does that work? 

DOROTHY LI: Right. We need to also have visibility into the inventory, which is where the low, where the demand is. That’s where we work with shippers. And in some cases, we also work with third party brokers as well to know where the demand is, what the facility the loads are and when a trucker can go pick it up. 

MICHAEL KRIGSMAN: Obviously, this then depends on you having a very broad network so that you have if not a completely comprehensive view sufficiently comprehensive so that you can operate this marketplace efficiently. 

DOROTHY LI: Yes, absolutely. We work with many of the Fortune 500 shippers and we operate really the nation’s largest carrier network with 300,000 carriers that we work with. And building– and we understand the density and conditions of lanes in different cities. And all of this you can imagine is creates a lot of data. And at the fundamental, at the core of it is a data and data analysis. 

MICHAEL KRIGSMAN: One of the terms that I’ve heard you use is elastic capacity. 

DOROTHY LI: Yes. 

MICHAEL KRIGSMAN: So what does that mean when it applies to the trucking industry? 

DOROTHY LI: Right. Elastic capacity really at the core of it or simply put is the ability to– is the ability for almost limitless capacity when demand surges, and the ability to then when demand shrinks to not to be locked into a contract that you might have signed years ago. And it really is meeting shippers where we need, where they need and the ability for them to be able to service their customers in any market condition. And that’s where the flexibility comes in is to be able to provide that capacity in any market condition, whether it’s a hot or a soft market and not be gouged and locked into a contract. 

MICHAEL KRIGSMAN: Then is that the same then as saying, we have short term contracts based on the market conditions as opposed to being locked in to long term contracts? 

DOROTHY LI: Yes, that’s one part of it. That’s a big part of that. Another part is to be able to tap into– in order to have this capacity, you need to be able to tap into a much larger network than the shipper himself may have with their own private fleet. 

MICHAEL KRIGSMAN: Why? 

DOROTHY LI: Many shippers do have their own private fleet, but they are a much smaller portion of the entire trucker fleet that is available nationwide. And so when demand surges, their own private fleet is not sufficient to serve demand. And by having this public carrier network, they can tap into the supply that Convoy has when demand surges. They’re not reliant just on their own private fleet. 

MICHAEL KRIGSMAN: So in other words, you have a sufficiently clear picture of shipping demand, which is coming from the shippers and a sufficiently clear picture of shipping supply. In other words, who’s available to take those shipments that you can do your data and machine learning magic and bring the two sides together very quickly. Is that a correct way to place in point? 

DOROTHY LI: Absolutely. That’s a very good way of putting it. 

MICHAEL KRIGSMAN: And so then the machine learning you do does what? 

DOROTHY LI: So the machine learning we do as on both sides for the truckers, we use machine learning to recommend the best loads for them and taking a lot of the truckers past preferences into past loads into consideration. And for the shippers, we provide real time pricing and bidding information. Again, using past information and real time market conditions. 

MICHAEL KRIGSMAN: So you’re creating this spot market really? 

DOROTHY LI: Yes. 

MICHAEL KRIGSMAN: Or I shouldn’t say you’re creating it because the spot– 

DOROTHY LI: The spot market exists. We are– the spot market exists, but it used to be that it takes a while, right, you also don’t because we don’t have a lot of the pricing information or the real time information. It takes a while to create a tender, which is a request to bid and the ability to respond. And that takes– actually, in the traditional market, it an take days. With Convoy, we’re able to reduce it to a matter of minutes. 

MICHAEL KRIGSMAN: So historically than the market between shippers and truckers was very inefficient to use the term that you’ve brought up so many times very inefficient because, OK, we know we have to ship a load and so– and that’s going to be in a few weeks from now or months from now or days from now. And well, we’re going to have to put out tender offers and let’s get bids. And it’s really hard to do this, so let’s have some long term contracts and we’ll really knock down the price as well. 

DOROTHY LI: Yes. 

MICHAEL KRIGSMAN: So that will be good for us as a shipper and the truckers will have to eat it. 

DOROTHY LI: Yes, that’s right. 

MICHAEL KRIGSMAN: And you guys are using the data and the machine learning to cut through all of this. 

DOROTHY LI: Exactly. And we’re able to also cut down that response time, right, so that you can have much faster response time. And so when we think about flexibility and not only its flexibility and capacity, it’s also this fast response time that allows you to react quickly to market conditions. 

MICHAEL KRIGSMAN: So you’re really a marketplace between shippers and truckers and you’re able to create a spot price market essentially. 

DOROTHY LI: Yes, that’s one of them. But we’re also more than that in that, we don’t just facilitate between shippers and drivers. We also have these smart trailers that allow us to also guarantee a service delivery to the shipper. So we don’t just in addition to facilitating, we can also give shipper– this day and age, one of the key things is a promise and to be able to meet the delivery estimates that the shipper wants. 

MICHAEL KRIGSMAN: And how do you do that? 

DOROTHY LI: How do we do that a part of going back to data. We have these smart trailers that are equipped with these IoT sensors. So at any point on the road, we know where the driver is or we know when if it’s going to be delayed. And so as a result, we’re able to give a lot more transparency to the shipper about where my truck is. And if there is a delay, there’s a hurricane, there’s a weather condition somewhere we can proactively notify them. 

MICHAEL KRIGSMAN: And in contrast, what was going on in the trucking industry before you started? Can we say placing sensors through the trucking industry? 

DOROTHY LI: I think so. I think we are– and we’re also one of the first to start this program called flexible drop and hook. The traditional drop and hook program are not equipped with all these smart sensors and intelligence. And so because we have the transparency and real time information about where the truck is, we’re able to really reduce the time that a trucker is waiting at a facility and so can just drop a trailer, get a new trailer and off you go on the road. 

MICHAEL KRIGSMAN: In order to work, you are machine learning magic. 

DOROTHY LI: Yes. 

MICHAEL KRIGSMAN: You need lots of data as you were describing. 

DOROTHY LI: Absolutely. Exactly. 

MICHAEL KRIGSMAN: Can you give us a sense of the kinds of data that you collect as inputs and then we’ll talk about what you do with that data? 

DOROTHY LI: Right, absolutely. I’ll give you one example. As I said, we collect data throughout the shipment– throughout the shipment lifecycle. But we also collect data in our mobile app of reviews of what a trucker reviews of a facility. And it turns out that– this is a really interesting example of how we use the data and use it to improve the industry. And so what a lot of people don’t realize is that the people, right, in a big shipper organization that has the authority to make changes is sometimes disconnected from how the facility is run. 

And so little things, like, whether the facility has restrooms is really makes a difference for the experience of that trucker. With our mobile app, our trucker can add reviews, add these data points. And we surface these to these decision makers. And as a result, they can make changes and improve how the facility is run making more efficient and more convenient for truckers. So that’s just one example of the kinds of data we collect and how it’s improving the truck drivers life. 

MICHAEL KRIGSMAN: I see. So you’re collecting data around the locations of the facilities, the locations of the trucks themselves, congestion levels on the roads. What other kinds of data serve as inputs into your analytical processes? 

DOROTHY LI: That’s right. We also collect– for the driver, we also collect obviously the geolocation and the appointment when he’s started when he’s finished– when he’s finished the particular assignment. And on the marketplace itself, we have collected the historical pricing data. And with our machine learning model, now we can more accurately give more longitudinal forecasts of truck pricing, for example. 

We also collect data around how often a truck is being replaced. So that’s an interesting study we call a survival study. And so with that, we can inform on how often a truck is replaced and we can use that data to either create incentives, for example, for when it’s replacement time. We can proactively let the driver know and we can create incentives to replace them with more green vehicle options. 

MICHAEL KRIGSMAN: So you’re collecting all of this data and you’re storing the data. Before we talk about how you analyze the data, can you tell us a little bit about the nature of your infrastructure? 

DOROTHY LI: Yeah, so our production services run our relational databases. And to do ad hoc and historical analysis, we use a cloud data warehouse. And for a lot of the real time information, we also have an event based platform that does streaming. 

MICHAEL KRIGSMAN: Dorothy, regarding the real time data that you’re collecting, can you give us some examples of what you do with that data? How do you use it? And link it back to why this is beneficial to the truckers, and how does this all create efficiency? 

DOROTHY LI: Yeah, yeah. Absolutely. Well, one real example that helps a trucker is we collect real time geolocation data from active loads. And one of the ways that helps truckers is that it actually allows us to pay them better. And so let me expand on that. Before Convoy and before this technology improvement, truckers would turn in– how do they get paid? They were turning in these carbon copy slips. Or they would– very manual, they would record when I started this load, when I actually arrived. And if there’s a delay, they often wouldn’t get paid for days. 

But now we can bypass all those information because we know at any time where they are, whether there’s a delay, whether that delays caused by the trucker, whether the delay is caused by facility. So we can give them quicker pay, we can give them– we can pay them pretty much right on the spot. So that’s a very real example of how data is improving the truckers life. 

MICHAEL KRIGSMAN: And what’s the– 

DOROTHY LI: Yeah, go ahead. 

MICHAEL KRIGSMAN: No, please. 

DOROTHY LI: Another example with real time data is we do real time marketplace bidding and pricing data. That allows us gives us unique visibility into the actual state of the market and not just a backward look at previously back transactions. And so back to the spot and flexibility, it really gives us both transparency to the shipper as well as give that flexibility of real time pricing. 

MICHAEL KRIGSMAN: So you have an insight into this market on both sides of the market the seller, meaning the trucker, as well as the buyer of the trucking services which is the shipper. You have this unique insight– 

DOROTHY LI: That’s right. 

MICHAEL KRIGSMAN: –all the time as to what’s going on. 

DOROTHY LI: That’s right. 

MICHAEL KRIGSMAN: What is the shipper get out of all of this? You’ve described the benefits to truckers, but what about the shippers? 

DOROTHY LI: Well, I think in simple terms, the shipper gets one, more flexibility as we mentioned earlier with elastic capacity. Hugely important when you think about in this market where we can have demand surge, we can have demand goes down. And that flexibility is really is at the heart of what drives the efficiency here. And the second is lower cost and higher quality supply, right? Give all that transparency with pricing. We’re able to serve our customers and, in this case, our shippers at a lower cost and with higher quality supply. 

And the third is we do improve– help them improve the quality in some ways their service to truckers, right? And so going back to what we said earlier with facility reviews, it allows shippers to improve their facilities and become a shipper of choice for truckers. 

MICHAEL KRIGSMAN: So you feed your data from the market in terms of pricing, for example, as well as feedback from the truckers to the shippers so that they can make more informed decisions and make themselves more attractive to the truckers, to their suppliers. 

DOROTHY LI: Yes, absolutely. 

MICHAEL KRIGSMAN: So there’s a very important supply chain aspect here as well because trucking is hard to come by right now. And from the shippers point of view, anything we can do to make us an attractive customer is going to be really beneficial. 

DOROTHY LI: Absolutely. 

MICHAEL KRIGSMAN: How do you do all of this? I know you mentioned machine learning, but how do the pieces come together? 

DOROTHY LI: We start with– again, we start with the data. And Convoy collects as thousands of data points, not only through the end-to-end life cycles, but also the features that we use on our website and our mobile app. We have some of the world’s best data scientists coming from companies like Amazon, Zillow, Microsoft, and the best institutions and organizations in the world. 

And we develop these machine learning models that are able to both predict real time and help on one side help truckers recommend loads and give them the best loads that are appropriate for them. On the other side, giving real time pricing information that reflects real market conditions. 

MICHAEL KRIGSMAN: How do you solve the empty miles problem using your machine learning data? How does that help? 

DOROTHY LI: Right. One of the key ways we solve the empty miles problems is through a program called Automated Reload. And what that really means is that we’re able to batch shipments together with a day that we have such that the trucker can then when they are on one load– when they take one load from say New York to Milwaukee on the way back, we have already batched his returning, his round trip. And so he’s never coming back with that empty trailer. So batching or Automated Reload is one effective way that allows us to reduce empty miles. 

And then the other one is really through automation. When you think about the time that a trucker is used waiting at a facility waiting for schedule appointments, automation really helps. You reduce that wait time and gets the trucker going. 

MICHAEL KRIGSMAN: What kind of automation are you referring to? 

DOROTHY LI: I’m referring to, for example, just scheduling appointments. It’s one of the key problems or the things that we used to do with paper and pencil, or phone calls and emails and there’s a lot of time wasted just figuring out what is the optimal time. And with Convoy, you can just book your appointments on your website on your mobile app and we– it just reduces a lot of time. 

MICHAEL KRIGSMAN: So the whole thing is bringing automation and prediction to this very old traditional industry that I guess was right for change. 

DOROTHY LI: It’s a backbone. That’s right. And is the backbone of our economy. 

MICHAEL KRIGSMAN: From a machine learning standpoint, do you run experiments or A/B tests, for example? 

DOROTHY LI: Yes, absolutely. The experimentation is really central to almost every product decision that we make here at Convoy. And our data scientists and our researchers are embedded in every product development team. An example of the kind of experiment we make is, for example, I mentioned earlier that on our mobile app, truckers can bid in real time and they can search for loads and we run experiments around, what is the best kind of ways to recommend a low to them? 

You can think of similar to almost like product recommendations on Amazon. There’s different parameters, different truckers may care about different things, some I care about proximity to home, some I care about time, or we take their past loads into consideration. And that’s the kinds of experiments that we make. 

MICHAEL KRIGSMAN: So you’re using machine learning then to optimize the loads, but again, in both directions so the truckers taking the right kind of load going out, but also has a load coming back to avoid the empty miles that you were describing earlier. 

DOROTHY LI: Absolutely. That’s right. 

MICHAEL KRIGSMAN: You are really changing this traditional industry. How have truckers responded to this kind of automation? 

DOROTHY LI: I think I would say that truckers really– we’ve seen overwhelming positive responses from truckers because they see the real benefits of technology, right, not for technology’s sake. So going back to the little things around, do I have– is there a restroom in this facility? Two, can I get paid quickly if there’s a delay in the facility? So those things. Or, of course, reducing the empty miles means that the trucker is earning more, too. So truckers have overwhelmingly been positive about these changes. 

MICHAEL KRIGSMAN: So you’re bringing direct and very tangible results to the truckers very quickly, I would imagine. 

DOROTHY LI: Absolutely. And within Convoy, not only do we talk about sustainability for the environment. We also talk about sustainability for this profession. Many of these truckers are– it’s a hard job. And to sustain it, we need to make their lives better. We need to do these things– all these little things that add up and make this profession, which is really the number one job in 29 states a lot more sustainable. 

MICHAEL KRIGSMAN: What about on the shipping side, have they had to change or adapt in response to this introduction of technology? 

DOROTHY LI: I think that since Convoy’s started, we’re seeing more and more shippers adopt technology. Digital transformation, it’s you guys have seen this happening across different industries, also happening in this very traditional industries as well. So many shippers have started adopting APIs, for example, and shippers also work with a lot of TMS providers that work with us on API integration as well. So we really are seeing shippers starting to adopt technology and digital transformation. 

MICHAEL KRIGSMAN: What’s a TMS provider? 

DOROTHY LI: A TMS provider is a Transportation Management System that they do things, like, they actually are the ones that would do things like scheduling appointments and manage the lifecycle shipments for a shipper. 

MICHAEL KRIGSMAN: So shippers are getting much more comfortable with orchestrating this entire process using technology as you’re describing? 

DOROTHY LI: Yes, they are. 

MICHAEL KRIGSMAN: Dorothy, where is all this going? 

DOROTHY LI: I think we are at a really pivotal moment in freight. When you think about when Convoy’s started smartphones or started to be in the majority of trucks, but machine learning was just getting started. Now, automation and machine learning capabilities are really being deployed at scale. And companies of every size are trying to optimize their supply chains to improve their bottom line and improve their ability to service. And also sustainability is starting to become really central to many of our shippers. 

And so this unique mix of technology transformation and the need and desire to improve climate really creates a very real opportunity to transform this decades old business of moving freight from this old approach to a new and innovative approaches. So I think it’s a very exciting time to be transformed this industry. And Convoy’s really leading that transformation with our technology approach. 

MICHAEL KRIGSMAN: Dorothy, what advice do you have for truckers, for shippers, for people who are using data to drive efficiencies and to change their part of the world? 

DOROTHY LI: When I think about the challenge in trucking today, it isn’t necessarily figuring out what to do. And maybe you can extend it to more than trucking in many industries as well. It’s really figuring out how to do it, how to bring these ideas to life? You know, what code, what data, what models do we use? And that’s why the role of the CTO is really more important than ever. And that’s why data infrastructure, the ability to conduct experiments quickly and be nimble is super important. So perhaps my advice is invest in your data and analytics. That is really the cake for that icing, which is machine learning. 

MICHAEL KRIGSMAN: I have to say it’s fascinating to hear how you’re collecting the data using machine learning to analyze that data, and it ends up resulting in better lives for truckers makes their lives easier. 

DOROTHY LI: Yes, absolutely. And this is one of the things that gets me going every day. 

MICHAEL KRIGSMAN: Dorothy Li, CTO of Convoy. Thanks so much for taking the time to talk with us. I really, really appreciate it. 

DOROTHY LI: Thank you for having me. 

MICHAEL KRIGSMAN: A huge thank you, a heartfelt thank you to Redis, for making our conversation possible. Literally, without them, we wouldn’t be here. So thank you, Redis.