You have probably heard in recent months from a friend, a co-worker, the news, or a source on social media about the impact artificial intelligence is making on our day-to-day lives. Spot’s technology arm, Red Technologies, has been integrating A.I. into Spot’s products to automate processes that will provide better, competitive services for our shippers and carriers.
Three key members from the Red Technologies team were featured in the inaugural episode of Spot’s new podcast, “More Than a Broker,” to discuss how their team is leveraging A.I. in their current projects. You can listen to the first episode of More Than a Broker in the player below or on Spotify, Apple Podcasts, and other major podcast platforms.
- A.I. is a broad field that essentially encompasses any algorithm that uses intelligence. In contrast, machine learning uses algorithms that, instead of being told exactly how to perform a set of tasks, use advanced mathematics to learn relationships in data to optimize tasks.
- A.I. applications will include advanced, faster quoting for more competitive rates, as well as providing better carrier sourcing for carriers and shippers.
- Feedback from carriers and shippers is paramount to the success of developing A.I. models – the better the data, the better the model used.
What they said
“Of course, customers want a competitive rate, but I think even more importantly, they want to make sure that the things that they’re shipping are in good hands. On the carrier side, we want to make sure that we’re going with the best, the best, and that the really good carriers aren’t losing out to some of the things we’re seeing a lot like double brokering.” – Trevor Crupi, Red Technologies
Ben Garvin: “Welcome to Ahead of the Curve, where industry experts discuss the advanced technology driving innovation in the logistics industry. I’m Ben Garvin, director of Red Technologies. With me today, I have Tyler Hampton, DevOps Manager, and Trevor Crupi, our Data Scientist. On today’s show, we’re discussing how Red Technologies is embracing A.I. and machine learning to provide our customers with better rates.
Tyler, would you give us a quick run-through regarding the project, how it started, and really the timeline moving into the A.I. and the machine learning technology?”
Tyler Hampton: “Thank you, Ben, for the introduction. The project for us started about a year ago. We came to a point where we were growing just like a lot of other brokers in the industry, and we had a lot of data in a lot of different places, and we needed to bring that all together.
We had a lot of ideas that we wanted to do around data engineering, data science, and business intelligence, and we really didn’t have the data infrastructure to handle that internally. What we did is we started moving our data to a more cloud-forward environment. We leverage Azure, the cloud provider. We also leverage tools that are hosted within Azure such as Databricks, Azure Machine Learning, etc.
Once we were able to bring in all of our data into a data lake environment, Trevor and our other data folks were able to do more data-heavy work, more expensive queries that can bring the data that we hold here to the fingertips of our users, whether that be Sales, Carrier Sales, Operations folks, etc.
From a timeline perspective, it’s been about a year now, and we’re about 50-75% of the way there. We’re working every day to make sure the data is right for our users – customers and shippers alike.”
Ben Garvin: “Thank you, Tyler. Staying with that topic, did you guys do this all in-house or did you bring partners in? How did you leverage that project?”
Tyler Hampton: “We leverage a partner that we work with currently; they have three to four folks on the team and then we have three folks here on our data team that work on the project as well. At the beginning, it was a few people and we’ve grown to seven people currently.”
Ben Garvin: “Moving over to you, Trevor. Could you give us from the business approach side an overview of A.I. and machine learning and what that looks like? Once again, just keeping it high level. I know you could get really into the weeds and technical and lose me quickly, but moving forward, just the A.I. and machine learning from the business approach.”
Trevor Crupi: “A.I. itself is a super broad field and it essentially encompasses any algorithm that uses intelligence. That could be anything from some hard-coded rule set, it could be a linear regression, or it can be one of these massive, large language models like ChatGPT or Bard from Google.
Machine learning is just a subset of artificial intelligence. Honestly, I think it’s what a lot of people mean when they talk about artificial intelligence. It’s just this very broad class of algorithms that are taught how to perform a certain task by learning from a set of data.
The real beauty of these machine learning algorithms is that instead of being told exactly how to perform a set of tasks, these ML algorithms use some advanced mathematics to actually learn relationships in your data in order to optimize the tasks. I think that has huge implications for businesses because instead of relying on extremely simple data insights – plotting X versus Y – you can make use of these sophisticated machine learning techniques to pull patterns from your data that could otherwise be hidden extremely deep within your data sets.”
Ben Garvin: “So, Tyler setting the foundation with getting to the A.I. technology and the machine learning technology piece of it. Now that we have that in place, talk some highlights around what we’ve done for the projects, and how that benefits Spot and also the customers.”
Trevor Crupi: “I think logistics is a really unique industry, and that is sort of this mix between new school and old school where a lot of companies have been very diligent for years about collecting data and harvesting data but have never really taken the jump into hiring data science and kind of utilizing these algorithms. There’s a lot of data scientists like me out there who are really excited to take on these challenges.
I think the first big challenge that we’ve tackled with Spot’s massive mountain of data that we have has been improving our rates on the spot market. It’s a super interesting challenge from a technical perspective because you kind of have to find this perfect balance between coding a shipper too high because you might price yourself out of it and not stay competitive. If you go too low, then as a broker, you’d be losing money on every load.
There’s this really interesting information asymmetry about how we rate specific markets and rating those specific markets depends on so many factors – number of miles, types of trucks, month of the year, etc. We’ve been tackling this problem by using machine learning because if you throw a machine learning algorithm at this problem, they can find these really deep insights in these really high-dimensional data sets.
From a customer perspective, that means getting way more competitive rates while still being extremely confident that we can get your loads from point A to point B really safely.”
Ben Garvin: “How’s it going? If you put it on a grading scale: B+? A? What are your thoughts around how the project’s currently going?”
Tyler Hampton: “I would give it a B. I think it’s been a really exciting project for us to start, but I think moving it into a more intelligent, quoting environment from our perspective is going to give us a little bit of a more competitive edge in the future.
A lot of customers want real-time rates, and with this environment and with this model, we’re able to give that to them. From an internal perspective, we also need to make sure that we’re giving rates that our sales folks can be confident in and that we know that we can provide service for our customers as well.”
Ben Garvin: “You touched on the control piece, and the one thing that I’ve picked up, I know that some users believe, ‘Oh my gosh, they’re taking this away from us now. They’re going to push this into, you know, this model.’ Does it go through the pricing piece of it to stay with that topic? You really need, and pushing hard for the users, that feedback to really tweak and improve across the board. Trevor, would you mind touching on that piece?”
Trevor Crupi: “Getting feedback from Sales is probably one of the most important things that we’ve started doing because Account Managers, they know the market. They know the changes in the market, they know what needs to be done in order to win a load.
I would say that the point isn’t to replace Account Managers. The point is to allow them to do their jobs. They don’t have to spend so much time quoting loads and worrying about pricing. They can focus on getting your stuff from point A to point B.
Ben Garvin: “It’s that speed and accuracy, right? You’re going to give them the better rates and also give it quicker as well.”
Trevor Crupi: “I think that the main place that we’ve seen a lot of success with this is on quotes that come in so quickly that they don’t have time to go through and rate all of them. Now you can quote a thousand things in an hour, and you can be confident about your quote.”
Ben Garvin: “If you fast forward six months, a year, two years down the road, what is the A.I., the machine learning doing for Spot and then also the future for our customers and our prospects, too?”
Trevor Crupi: “I think the most obvious next step is probably to use some of our insights into pricing the spot market and take that to the contractual side. The contractual market rates tend to be a little bit more stable since you’re talking averaging over these really long periods of time. You have to also be extremely cognizant of some of the larger-scale seasonal fluctuations that you have to see. A lot of those seasonal fluctuations, the best way to handle those, is through machine learning, A.I., statistical inferencing, those sorts of techniques.
I think another big area that we’re really excited to tackle is the carrier sourcing side. This is a big piece that affects pretty much all aspects of our business. Of course, customers want a competitive rate, but I think even more importantly, they want to make sure that the things that they’re shipping are in good hands. On the carrier side, we want to make sure that we’re going with the best, the best, and that the really good carriers aren’t losing out to some of the things we’re seeing a lot like double brokering.
This is a similarly tough problem to tackle, but I think it’s vital not just to the success of Spot, but also to the success of the customers and the carriers that we work with.”
Ben Garvin: “Tyler, would you throw anything in there with just the future of the benefits around A.I. machine learning?”
Tyler Hampton: “I think one of the most important pieces that we’re going to cover next is going to be that carrier selection piece. We have a lot of location data, and we have that data internally. There’s a lot of market data as well that we can use to really help our carrier sales team create those relationships and really bring that ease and peace of mind to our customers to make sure their product is getting from A to B safely. We can leverage a ton of A.I. and machine learning to really find that carrier that’s going to match that line perfectly and provide really good service.”
Trevor Crupi: “I think Tyler mentioned a word that I think is really important here, which is the point isn’t to go out and immediately grab a carrier – the point is to build these relationships. We want to make sure we’re building relationships with good carriers that we really care about and really care about us.”
Ben Garvin: “I’m excited about just enhancing the relationship with our customers. When you think about before COVID, people looked at the source of consumption or supply chain or supply chain end-to-end, the visibility there, and we’re a sliver in that. What we can bring for our customers to help them enhance real-time visibility, overall efficiency, bottlenecks, and drive down cost with that piece, I think it’s huge. I’m excited about that for our customers and for Spot.
My last question. Trevor, I’m going to put you on the spot here. Would you rather have better data or a better model?”
Trevor Crupi: “This is the age-old question in data science. The true correct answer is you want to have both. Context is extremely important here where if you have great data but a bad model, it’s not going to look good. If you have bad data but a good model, it’s going to be equally as bad.
I think, especially in logistics, one of the things you have to remember is that when you create a model for pricing, for example, your model is going to be as good as your brokers. If your brokers tend to do certain things, your model is going to pick up on those biases and it’s going to tend to do certain things. I think in the logistics industry, specifically, having good data is really important. I would probably put having good data over having a good model.”