How AI and Predictive Truckload Pricing Are Impacting Shipping | Episode 16

Hey, everybody. It's Patrick Escolas with another Banyan Tire Tracks Podcast. We're here at Connect 23. We're here with Dawn Salvucci-Favier of Greenscreens. She was our keynote speaker, and we're very happy to have you here.

Thank you so much for having me.

Well, thank you –

Excited to be here.

Yeah. Thank you, not only for speaking to everybody and also, for taking the time to talk to little, shiny head me.

It's a trend. There's a lot of shiny heads in this industry.

Yeah. It's all by choice, you know?

Yeah, yeah.

It's it's a really creative decision to do with my hair.

I might have to join the club.

Save a lot of money on shampoo, you know?

Yeah, exactly.

In a quick elevator pitch, because I know I talked with Matt and depending on how I release this, someone might have heard of Greenscreens. What's Greenscreens to you? When someone says, “What is that? Are you making green doors? Or what's that, you know?”

Yeah, we make video production. No, I'm just kidding. We don’t. We do predictive pricing for the truckload market. It's spot pricing. We use machine learning, data science, too. Use vast amounts of data, 20 billion in our network right now.

I was going to say, yeah, define vast. That's a lot. Yeah.

Yeah. We have 20 billion. That's about 13 million transactions is what we're using today in our network.

How far back does that 13 million spread?

It's really about 18 months’ worth of data.

Okay.

It is what that 20 billion is right now.

Yeah.

It's on a rolling basis, right? Every time we bring on a customer, they submit one to four years of their data to us. We use that data to train our model.

Got you.

What's used pretty consistently is 12 to 18 months of data is what.

Awesome. Before you were a CEO with Greenscreens, were you into AI and robots? Or, where you come from?

I do not have an AI background. I have a freight background. I've been in the freight industry for 33 years. Yes, I was six when I started.

Yeah. I like that. Yeah.

All right. That's what I like to tell people. But –

I believe it.

Thank you. 20 of those years was spent in the TMS space.

Okay.

I have been in product strategy, global strategy, all in TMS for various vendors. Most recently was the founder of the TMS company that I departed seven years ago. That's when I was introduced to Greenscreens. My background is tech, is freight technology specifically, but this is my first go round with AI. It was very exciting to me when I was presented with this opportunity.

And when you came into contact with Greenscreens, where were they then? Was it just an idea? Or was it like, “Hey, we can do some pretty cool stuff. We don't know what we're doing with it.” What did that look like?

It was really post-hypothesis, I'll say. Post-hypothesis, a little bit of product. We had three early adopters onboard, beta customers that were working with us to prove out the hypothesis to develop the algorithms and the user experience around it.

Got you.

That was in – company started in February, 2020. By the time I joined in the summer of 2020, we had already engaged with those three beta customers. NFI, Gulf Relay Logistics, RDS, a company called Cargo that is actually no longer in business anymore. We're among those customers. I came onboard in the summer with the primary goal of getting to MVP product, right?

Okay.

Product market fit, MVP product and really, the messaging, the pricing, the go-to market strategy.

Yeah. I mean, what was the – not just because it was cool and you could do it, where was the gap in the market that Greenscreens was supposed to – it's supposed to support, or add value?

Yeah. I mean, really the hypothesis was that for the freight brokerage market specifically and in 2020, the tools that were available were inadequate. Most of the tools that were on the market then and even still today are using historical averages. They're looking at what has happened in the past, right? Three days, seven days, 30 days. As we know with a market that moves as fast as our market does, past performance is not necessarily an indicator of future performance.

Okay.

There was really nothing in the market that was using AI, data science, machine learning to predict forward. Not what has happened, but what will happen.

What did that amount to as far as, if I was using that other source, or I was looking back, what does that mean for when I use that moving forward?

There's a couple different things there. Number one, what we are able to do with the AI and the machine learning is to take your data as a broker and train that model about the way your business specifically buys, right?

Okay.

Broker A and broker B are going to have very different buying power and buying behaviors in the market.

Right.

Size may not necessarily be a determinating factor either. Broker A, who may be a many hundred-million-dollar revenue business and broker B, who may be much smaller, in some markets, broker B might actually be able to buy better than broker A, because of their carrier mix, because of their expertise and equipment specialization, flatbed, reefer, whatever the case may be. It's not necessarily a size indicator, but it is what is your freight mix? What are your shipper relationships? What are your carrier relationships? What is your buying strategy?

Okay.

That's what we use your data to train the model of how you buy specifically. That's the number one difference, because the older tools were really a market rate, which is great, right? They're useful, they're great, but it is overall, an average. What is the market doing? But that doesn't mean –

It's relevant to you.

To you. You may not be able to buy at that rate. That's number one.

Okay.

The other thing, though, is because we're able to come up with these very prescriptive models that are specific to how you buy it, it comes down to a measure of accuracy.

Okay.

How accurate is that rate that you assume you're going to be able to buy at, compared to what you actually book at? We've done a lot of analysis with our customers that the difference between a 7% margin of error, which right now our customers, on average, across all 125 of our customers, we are at about a 7% average.

6.7% if you believe Matt, so yeah.

Okay, there you go. About 7%, 6.7%, yes, margin of error, versus some of our other customers have benchmarked other sources at an 18% to 22% margin of error. We have done the analysis that says, if you're assuming that you're going to buy freight at a price and then you're pricing that to your customer on a markup basis, the difference between a 7% margin of error and an 18% margin of error could be 43% margin.

Really?

Yes.

That's crazy.

Isn't it?

It is. I mean, that's reason enough to look in a Greenscreens there. What are some of the reasons that you might – because obviously, you're going to tell me all the reasons why someone should. What are the biggest obstacles, or hurdles you have to get someone to either, to buy into the idea of AI, or specifically Greenscreens itself within the truckload world?

Yeah. I think the biggest challenge, and I talked a little bit about it in the keynote today and it's not just with AI. I think it's with any digital transformation project is change management.

Okay.

It's really needing to have that buy-in at the top and middle level of the organization and people who are able to drive that change in mindset throughout the organization. My friend, Bart De Muynck, who used to be with Gartner, then P44, he recently did an article in Forbes, where he talked about the fact that the biggest reason for failed digital transformations is a lack of change management, and that we all need to really start thinking differently about technology and trusting the data. Not just tribal knowledge and gut feel, but trusting what the data tells you.

I think, you had also said, I like this line, “AI will not replace you, but a human using AI might.”

That's right. That’s right.

And so, if it's out there, you should be using it, or your competition is. Is that what that's talking about?

Well, I think we're getting there. As I said in my presentation earlier today, I think we are still in, let's say, the early majority phase of adoption of AI technologies. Now, certain parts of our industry, like warehousing, for example, has led us for a long time in robotics and things like that. But if we talk about freight specifically, I think we're really in the early majority. I think it is quickly becoming a haves versus have-nots from a technology perspective.

Okay.

We were created as a company really on the hypothesis that competitiveness is huge. Back in 2020, 2021, when we first came to be, it was the very big digital freight brokerages that had a lot of VC money. They were investing a ton in business intelligence and pricing intelligence and connectivity to shippers, which immediately put the majority of the other 18,000 brokers in the market at a competitive disadvantage, because they didn't have the capital, or the expertise to invest in developing that in-house, right? We were really bringing that level of technology, connectivity, data intelligence to the masses, to level that competitive playing field.

Now, today, competitiveness is still important, but given the market conditions today, very different. When we first came into the market, it's –

Well, yes.

Yeah. It’s every at-bat counts for a broker. Every opportunity to quote counts. You don't necessarily need to be the broker who comes in with the lowest price, but often, you just need to be the first one in with a reasonable price. How do we enable our customers to maximize –

It’s a great point.

- their bidding opportunities and do so more quickly?

Yeah. Like I said, I think that's a great point, is that it doesn't always have to be the cheapest, but if you get it there, get it to them quickly so they can make a faster decision, that might be the key there. Something else you had topped on, which I thought was really interesting, the idea of using technology, obviously, isn't anything new, but AI and robots get thrown in there. When was the first robot you started using that? I think you had that and I thought that was really interesting.

Yeah. The first, I think it was called Unimate. It was the first robot. It was invented in 1954 and it was first deployed in a GM plant in 1961.

Really, even though it looks a lot different than what we're doing now with it, none of this is new. It's just with a different medium.

Absolutely. Absolutely. But the adoption curve has really accelerated, again, talking to a warehouse automation, that goes back probably over the last 20 years, we've seen a lot of that. I think for more AI, machine learning-based technology, so forget about robotics, even forget about some of the generative AI, right?

Okay.

Like ChatGPT and the automated phone systems, right? That dreaded, you know.

I was a BDR. I hated that. I hated it so much.

Right, right. I mean, that stuff has been around for a while. But when we talk specifically about some of the deep learning and the machine learning and neural networks, that is all relatively up and coming, right? And early adoption on that.

With that, you talked on a lot of things. One of the things that – it's kind of an alphabet soup. You've got AI, you've got BI and all these. Which of those – what does it mean and what different things are used within Greenscreens to get it? Because I assume it's not just one thing.

We are purely AI-based, but there's different flavors. We are AI machine learning specifically. If you look at the category of AI, under AI, you've got deep learning, you've got machine learning, you've got robotics, you've got generative AI, which again, is like the ChatGPTs, or the natural language processing, those types of things. The specific branch of AI that we use in Greenscreens is a category called machine learning. In machine learning, there's also other subcategories, like neural networks and deep learning.

Got you.

We are purely machine learning-based today and we are employing deep learning, as well as some more traditional machine learning techniques there.

Without getting too in depth, what defines machine learning versus the others?

Machine learning is really about using data to learn the way that a human would learn, right? We'll again take the process of a freight broker pricing a load. He or she is going to gather as much data as they can, their historical data, what the market is doing, what capacity looks like, right? We heard Brent talk about OTRI as one of the indicators that he looks at and things like that.

You may have somebody who, let's say, moderate level of experience, they're going to go through some mental gymnastics and they're going to gather up as much data as they can and they are going to come up with what they think they're going to be able to buy at. Now the more experienced that –

Man, does not sound fun. Yeah.

It's hard. The more experience that individual has, the better and faster their decision-making process becomes.

I'm starting to see where you're going here.

You’re starting to see, right? With machine learning, it's essentially simulates the intelligence of a human. It's using data, but a lot more data than a human could ever possibly absorb in milliseconds. While it might take that broker, I don't know, 20 minutes, an hour, maybe more, to come up and price a lane, or a few lanes, the AI is doing it in milliseconds, and –

Between the screen refreshes. Yeah.

The more data it gets and the more time it has to evaluate that data, the smarter it gets.

That's awesome. Obviously, this is going to have a big impact. What do you see as a short-term impact of this on the marketplace and three, five years from now, what does that look like?

Yeah. I think in our space in pricing specifically, I think as I said, this is becoming a have versus a have not. I think that you're going to see a lot of companies that have embraced this more dynamic, predictive pricing are going to start outperforming those who don't have it. Those that are still relying on a post-and-pray mentality, or relying on history. This is what happened two weeks ago, but that's not necessarily relevant, especially again, in a very fast-moving market. I think it is very quickly going to become a have versus a have not in the area of pricing.

In other areas of the supply chain, I think we're probably still five to 10 years away from mainstream wide adoption of the AI technology in general, again, outside of warehouse automation and –

Is that just because there's gaps in the tech, or not the best applications yet?

I think it's an openness, and again, a change in mindset. I think when we first came to market in 2020, one of the biggest challenges, or two big challenges we had was A, people being unwilling to share their data and understanding that the aggregated data has more power for everybody. Those silos have been breaking down.

Okay.

I think it's, again, that change management of something new and something maybe that they don't understand. When we're selling our products into the market, I would say, more than half of our sale is about education, and educating the market on what this is and why it's different and why it's more important.

It's new. It makes sense that education is a big piece of that, because you got to let them know what's out there. Before you know, this is why you need it, too. With that, how would someone, because as we talked about, Greenscreens is this great piece, probably use it with other – how would you best leverage it if I'm a broker, I’m saying, “I want that,” how do I best make that work for me?

Yeah. I mean, look, I think certainly within the TMS and we're integrated with Banyan, as well as several other TMSs.

Great plug.

Yeah. Of course. Yay.

If the sign in Jersey wasn't enough. Yeah.

Yes. I think for us though, our goal is we want our users to be operating within their existing work in the systems that they use every day. We have a great user experience. We're very proud of our user experience.

But no Greenscreens TMS coming around the –

No Greenscreens TMS. Absolutely not. Absolutely not. There's enough. Why in God's name would anybody create another TMS.

Banyan's already perfect. Oh, no.

Right? Exactly. But no. So, we're very proud of our user experience. For those that aren't on a TMS that we're not integrated with, we do have a full, very robust experience. For us, it's really about keeping the users in the systems that they work in every day and allowing them to operate within that framework.

That's awesome. I think I'm really excited about the partnership and just about with more of the robust offerings that Banyan, though in the past, LTL, were really trying to expand. I think this, plus some of the other load board integrations is going to be a fantastic package to hand right now for 3PL brokers. We'll tap on it once. I know, this was all for brokers to listen to. No shippers should have gotten in here. I know we had a lot of people asking here, saying, “This sounds good. I'm a shipper and I want that data.” What are you guys thinking with that?

Yeah. Today, we do not offer our solution to shippers. Look, we had a gentleman in the room yesterday in one of our talks here. He said, “Look, I'm a shipper. I get it. I know I pay a premium on my freight when I give it to a broker and I'm willing to pay for that. But I would love to have this.” Look, our commitment to our broker customers is we're not going to give the shippers the same tool that we give you, because that in a way is exposing your –

It's their data.

- your differentiation, your secret sauce. We are working on a shipper solution.

Okay.

As I keep saying –

We won't hold you to a timeline. Don't worry. Don’t worry.

Yeah. No. I'm going to be a little bit cagey, but I'm hoping, fingers crossed that by this time next year, we should have a solution for shippers. It's going to be a little bit different.

I would imagine so.

It's going to be focused – today, our solution is focused on optimizing that broker to carrier buy rate. The shipper solution, obviously, would be more focused on that broker to shipper sell rate, first and foremost. Also, I think, for our solution to be meaningful for shippers, we also need to look more at long-term forecasting, as well as short-term, which we do today, because so much more –

Yeah. The great differentiating perspective.

- of the shipper’s rate is moving on a contract basis. I'll give you an example. We had a conversation very early in the days of Greenscreens with a very large retailer, orange logo, home improvement store, who at the time, they said, they moved 30 million dollars in spot freight every year, which, frankly, is more than some of our brokers.

I'm going to say, that at a certain point, that shouldn't be spot, right?

Well, that was less than 5% of their total freight spend.

Oh, so maybe that – Okay.

Yeah.

Damn.

That was just the stuff that was hitting the spot market. They had a whole lot, 95% of their freight was moving on contracts.

Right.

At the time, the conversation was like, “We love what you're doing for a spot, but it's such a small percentage of our spend that maybe when you guys have something for a longer term, we would be willing to have more conversations.”

I need some small customers like that. When that's small, right?

Yeah. Right. Right. Exactly. Exactly.

Oh, man. I might afford a really nice tote to pay, you know.

But that's, I think for us, that's key is having both the long and the short-term predictions. As I said earlier, R&D, the R part of R&D with data science and machine learning often takes a lot longer than the D part.

It's a real big capital R. Yeah.

It is a really big capital R. We've been working on this and we're getting close, right? We've had some early prototypes that we've seen and we're hoping that by this time next year, we will have a solution in the market.

Again, great answer. No, we won't show this. I hear it from now and be like, “What happened, Dawn? No.”

What happened? Why isn't it here?

No. But, you know, anyone that's listening, we have about a 50-50 break in our users of 3PL versus shipper. Obviously, they want to know what's relevant to me and when's that coming. I got a lot of information from you on here, and I'd like to give my platform here anything you have to say, whether that's from Dawn, just in the masses, or as Dawn CEO of Greenscreens.

Oh, boy.

Have actually, here's a soapbox. What do you got?

Yeah. I mean, look, it's something I touched on in my keynote today is just technology in general and I know to not appear tone deaf, I'm going to acknowledge that the market is tough right now for a lot of companies. We know there are a lot of companies cutting costs, barely surviving in this market. It is going to get better. By the way, it's going to take a little longer than we wanted to. More importantly, I think I cited a study this morning in my keynote by KPMG that they did this – how are they released this past year and it was looking at the companies that came out of the Great Recession of 2009 –

Okay.

- where they had actually been able to grow their revenue and grow their business through the recession when their competitors and their peers in the market were actually losing money, some going out of business. What they found that some of the key characteristics of all of those companies that won had in common was an adoption of technology that was going to help them. Continued investment in technology that was going to help them to be more efficient, to save money, to improve their margins, or cut costs, right?

Do more with less people. Yeah.

Yeah. I think more importantly, given our position is using the data to understand if your commercial strategy is in-line with what the market is actually doing. I think that is super key to what we do. What they said in their article, and I love to say it is, no, don't guess, right? Don't guess if your strategy is going to work.

It also sounds, just to pay you back and that sounds like, you really have to self-reflect to make sure that what your mission is, or your view now, or from three years ago when you first put it together is still accurate right now.

Yeah. I think the natural thing that a lot of people want to do when times get tough is, how can I diversify? How can I broaden my market reach? What we have found in our business and –

It’s not always.

- what seems to be successful is double down and focus on what you do extraordinarily well, and differentiate yourself through service, through relationships, through things like that and let the technology help you figure out the rest of it.

Dawn, thank you so much for the time. I appreciate learning everything from Greenscreens, especially the CEO. Like it says, like she said, find what you do well and really make sure you keep doing that great. For everything, looking at the price and with the AI and the truckload, I mean, Greenscreens and Banya is a great partnership. Thank you very much for the time today.

Thank you.

Thanks for being here and talking for our conference.

Of course.

And to anyone listening, thanks for tuning in to another Tire Tracks Podcast, and we'll see you with the next episode. Thank you. Bye.