Driving Decisions with Data: How AI is Powering the Future of Auto Auctions with Dr. John Coles

In this episode we sit down with Dr. John Coles, VP of Data Science and Analytics at ACV Auctions along with guest host Jamye Carpenter, VP of Customer Success at Privacy4Cars. Together, along with Angela Simoes, they explore how data and AI are reshaping the automotive auction industry—from real-time pricing and trade-in automation to empowering dealers with actionable insights. Dr. Coles explains how machine learning is streamlining the car buying and selling experience, reducing friction, and helping dealers make smarter decisions. The conversation also touches on the complexities of EV valuation, the potential for industry-wide AI collaboration, and what the next generation of dealership leaders might do with these emerging tools.

https://ridecell.com/
https://www.linkedin.com/company/ridecell/
https://twitter.com/ridecell

Transcript

Angela Simoes:

Welcome everyone to another episode of the Mile Marker podcast. My name is Angela Simoes, your host, and I'm here with my special guest host, Jamie Carpenter, who is VP of Customer Success for Privacy4Cars.

Jamye Carpenter:

So great to be here.

Angela Simoes:

I think we may have to just do this all the time now together.

Jamye Carpenter:

I'm in. You tell me.

Angela Simoes:

Our guest today is John Coles, VP of Data Science and Analytics for ACV Auctions. Welcome, John.

Dr. John Coles:

Thank you so much for having me. It's a pleasure to be on.

Angela Simoes:

So I'm thinking most people don't think of a data science and analytics person when they think of auctions. Right. So tell me a little bit about your background and then your role and what you do, what all that entails.

Dr. John Coles:

Sure. So I didn't come into the automotive space intentionally.

Angela Simoes:

Most people never do. 

Dr. John Coles:

But I got familiar with the data space, really because I saw opportunities to improve logistics and people's lives growing up. So I grew up overseas, spent time actually responding to disasters through my undergrad and grad school experience and saw how data could be used to help people. But there were very hard problems when you write down data, when it's stored in its spreadsheet somewhere. It doesn't all come together neatly. So I worked in that space for a while, and then when ACV auctions was kicking off, I jumped on the rocket ship about six years ago and started learning about the automotive space.

Angela Simoes:

And so did they know they wanted to date a person?

Dr. John Coles:

Yeah,

Angela Simoes:

I mean that's kind of, I guess very, what's the word I'm looking for? Not I thoughtful, but very astute of them to know that that would be a real need in the industry. So that's great.

Dr. John Coles:

Yeah. Yeah. ACV auctions was founded on that idea of taking a physical asset and making a digital version of it. I joke that we're the original NFT creator, if you're familiar with that. Right? So generating those digital assets requires that we can map the physical to the digital world.

Angela Simoes:

Got it.

Dr. John Coles:

About condition reports. The part that really excited me too was seeing how people engaged with the auctions in real time. So actually real time market dynamics.

Jamye Carpenter:

And I think it would be helpful for you to know that ACV started out as an online auction.

Angela Simoes:

Oh, interesting. Okay. Very nontraditional. So data was baked in from the beginning. Exactly. It all makes sense now. I mean, the industry in general has been talking about data for so long, and how do you make sense of it? How do you monetize it? How do you bring it all together? So talk to us a little bit about maybe some of the biggest advances you have made at ACV when it comes to that, and then where you're seeing things going.

Dr. John Coles:

So ACV, going to Jamie's point, it was founded as a digital marketplace. Our acronym stood for actual cash value. Our focus was if we can find a buyer and a seller and a good, and they agree on the actual value of that vehicle. So one of the key areas that we've focused on is actually providing pricing guidance to our dealers and to their customers and consumers to remove friction in the marketplace.

Angela Simoes:

Got it.

Dr. John Coles:

The faster you can agree on a price, the faster you have a transaction. So that's been a key focus for the last six years.

Angela Simoes:

So automation or is AI, so what are some of the key things that have helped you get there?

Dr. John Coles:

So really machine learning has been the foundation of a lot of the pricing side. You need three things. You need high quality, consistent, and then longitudinal data. If you have those three components, then you can actually build the machine learning algorithms to repeatably, predict what's a price and tighten that range. And so that's really where we've been focused from a pricing side to add value to our dealers.

Angela Simoes:

And are you pulling in data from other sources, not just your platform, so you can get pricing like the market as a whole, not just your platform? Yeah,

Dr. John Coles:

Yeah. We do pull from third party data sources as well as our whole ecosystem, but the biggest driver for what a vehicle is worth is what is the asset and what will people pay for it,

Angela Simoes:

Right?

Dr. John Coles:

So for us, a lot of learning about how to turn real-time market dynamics into that price is understanding how many people viewed a similar car yesterday and what were they willing to pay for that specific similar car. And then providing insight on if we launch it a certain time of day, certain type of auction, certain length, and then we send out alerts to buyers and sellers to drive them to that auction, then we're actually predicting what the real time value of the car is. Not just saying, Hey, here's what you can list it for now, work backwards. We’ve got to figure out what will it transact for at this moment.

Angela Simoes:

Got it. And moving forward from now in the next couple of years, where's your focus without giving away the secret sauce? We don't want to reveal any secrets here. I get that, but what's the direction there?

Dr. John Coles:

I think there's so much more opportunity in the automotive space as a whole to use data to help our teammates. If you think about some of the really challenging parts of the automotive industry, a dealer who's staying there 12 hours trying to move one more car, we've been really focused on how do we use artificial intelligence, machine learning, these different data capabilities to automate the easy stuff to make each teammate and each dealer partner more productive. And so some of those opportunities ahead, if you sell a car, can we auto buy and autofill that car for you? Yeah. If you purchase a car off the street, you need to know exactly what you need to recondition it for, how much you can retail it for what the best location for it is, or whether or not to wholesale it. Those are things that we're fully automating and starting to put those tools into our dealer's hands. But for the years ahead, it's really about continuing to solve the pain points for our dealer customers.

Angela Simoes:

And so great transition to my next question there, Mr. Cole. I was going to ask, what are some of those pain points or challenges that your dealer partners are talking about where they're coming to you and saying, I really need help solving this issue?

Dr. John Coles:

One of the biggest challenges that our dealer partners have run into that we're really providing new tooling for is accelerated curb purchases. How do you get in front of consumers as fast as possible when there's still meat left on the bone with that vehicle. And get them into your service lane, get them into your showroom and capture upstream that two to five-year-old vehicle that there aren't that many of. Relatively speaking historically, it's a pretty rare car right now, but how can you make that street purchase with enough insight that you beat maybe some of the really large non dealer independents to that car and provide them enough value to the consumer? You get the car and you get the customer. That's a pain point that our dealers have today. And so we're working on solving that with clear car products and putting pricing right inside of our dealer's hands and their consumers.

Angela Simoes:

And so for our listeners who don't understand how that whole process works, can you walk through, do a quick getting the car and some of the terms that you're using, like the curb? If you could just kind the curb purchase. Yeah. If you could just give a quick explanation of what that looks like, I think that could be helpful.

Dr. John Coles:

Sure. So consumers today, I mean, I get inundated with the email saying, Hey, a value for your car, are you ready to trade it in?

Angela Simoes:

Right. Okay.

Dr. John Coles:

That is the first touchpoint for many consumers. We're powering our dealers to do is get an email, a touchpoint, and on their website when they go in and they start looking for a car, let's say you open up your local dealership website, you see a car you're interested in, you select it and say, Hey, I want to learn more about it. You'll be asked, Hey, do you have a trade-in? We power that experience for over a thousand dealers today and at scale for even other e-commerce partners as well.

Angela Simoes:

Okay. So that's getting the car that is a trade-in. And then selling on the platform is pretty self explanatory. 

Dr. John Coles:

Yeah. I mean, once the consumer brings that car in, what we do is we provide the dealer and consumer with an agreed upon condition. So that removes a lot of that friction of if you go in and you're told one price, if it's different than when you walk in, that introduces friction. So we've taken the power of all the vehicles we inspect and actually put that data and pricing behind the number we give a consumer and the clarity for why, so that the consumer and dealer have the same, they're looking at the same page when they walk in. We don't push our dealers towards wholesale, so our goal is to get them inventory. They need inventory, that's the dealer pain point, wholesaling vehicles as efficiently and profitably as possible. That's one part of a dealer's business that is the auction space. But there's so much more to our dealer partners that we're enabling by helping them find and source and bring in that customer to their showroom.

Angela Simoes:

And what's usually the reason for, let's say I get the email right? Hey, do you want to sell your car? Do you have data on why customers don't? Is it that they don't like that price that they are given? Which is usually the case, right? We'd like to believe that our cars are worth more than what they actually are. So is that usually the case or is, I mean, what is the challenge in getting the car?

Dr. John Coles:

The biggest challenge is, “Hey, I think my car's worth more than that.” Or, look, I'm happy with the car that I have. I'm not looking for another car. So for that first pain point, the key thing that we're trying to do is give the highest number possible to the consumer that is real and connected with what the dealer can actually offer and get on a retail or wholesale transaction later. We're able to do that because we're looking at that specific vin, the zip location, the context of that car, and the condition elements that way, instead of just adjusting down and saying, hey, take retail or wholesale, go back 30%. We know, look, you may have a 2022 that's actually in high demand. Our algorithms automatically identify that, and it's a much smaller adjustment specifically for your car, and that means you'll get a higher price for our dealer partner than you'll get from any other platform. The vast majority of time is time. 

Jamye Carpenter:

Is that the same for EVs, or is that more the same across the board, or do you see that your dealerships that will take a trade of an EV and resell it are still getting that higher percentage?

Dr. John Coles:

We do it car by car, so I won't say specifically exactly how the whole group behaves. I will say for EVs, the values have tended to fluctuate a lot more than some of the ICE and more traditional valuations.

Angela Simoes:

I should have sold my EV a year ago.

Dr. John Coles:

There's a lot of that going around. But yeah, I mean, we do price EVs. We hold ourself to the same accuracy standards that we do across the board, but there is more volatility in the EV space right now. I think in five, 10 years we'll see a little more stability. But that's where we have to look at the individual vin, not just the whole class of EVs, right? Are we talking about a Rivian? Those are still holding their values pretty well, and so it has to be that specific make, year everything. We don't just want to say, oh, it's an EV, we're going to cut it across the board. No, which, what are we working with?

Jamye Carpenter:

That makes sense. And I guess I was really more referring to just when I said EVs in general, just that we know that the price has dropped so much. So is there a little bump in what you would see? But yes, I get specific vehicle makes a huge difference.

Angela Simoes:

I would say specific vehicles, but then are you also looking at things like battery life? Are you able to pull that kind of data to factor that into the pricing?

Dr. John Coles:

We factor in all of the elements that we pull about the car. I won't dive too much into that.

Angela Simoes:

Oh, I hit the secret sauce. Put the cap back on.

Dr. John Coles:

That's right. There's a lot of moving parts in that space right now. And I think getting to that fundamental question of battery stability and battery range health, health is kind of that macro category, but fundamental question is, is it going to fall apart and is it going to get me as far as it used to and for how long? So we do account for those things.

Angela Simoes:

Okay. I will move on.

Jamye Carpenter:

If we're moving on. I'll say that John and I co-chair the AI committee for ARA. And John has just been so great. I'm so thankful for him because he's way smarter than I am.

Angela Simoes:

I was going to say, I don't know if I believe that. I don't think so.

Jamye Carpenter:

And he has just, you've really taught me a lot too, and so I appreciate that. Thank you. That's given me confidence that's given me the ability to write an AI book with Ridecell. But I know that we've talked about condition report writing and imaging. We're about to start talking about arbitrations, but I don't know if I've ever asked you like what do you think is the next big thing in AI?

Dr. John Coles:

I think the next big thing in AI will be taking all of the information that's hitting teammates inside of a dealership and consolidating it into clear summaries and recommendations. I think one of the challenges we've seen in the data and insight explosion in automotive over the last 20 years is it goes from one laptop to two screens to three to four, and you've got little mission control systems inside a dealership with split screens just to figure out, what am I doing with this car? Who's this customer? What car are they trading in? What's the new car? Did anything change in the market? Well, here's three reports. I think turning that information deluge into a very clear, concise recommendation and action steps will improve the quality of life and understanding between the dealer and consumer. So there's just so much information out there, and it is getting better and better. We're in the digital space first 10 years ago, but there's so much information that you can access. You shouldn't access all of it, especially for newer teammates coming into the automotive space. There's so many tools, so many systems. If you don't have a history in it, you're just going to get lost.

Angela Simoes:

Right. Well, and then you could better utilize the vehicle asset in whatever way is going to be most profitable for that dealer.

Dr. John Coles:

Exactly.

Angela Simoes:

Right. So does it make sense to sell it to a consumer or does it make sense to try to put it in a fleet, or do you salvage it? I don't know.

Jamye Carpenter:

Right.

Angela Simoes:

Or do you sell it? I mean, we sell it to auction. So we just had a guest on that was talking about the market for salvage vehicles in the UAE. I mean, holy cow, man. They're buying anything with four wheels and a steering wheel, he said. So maybe the dealer just does that.

Jamye Carpenter:

They need to buy our EVs.

Angela Simoes:

Maybe. But to your point, there's so much information. And how is, especially a new team, or even someone that's familiar with the traditional ways of getting rid of the car, how do you make that best decision? Right. Yeah.

Dr. John Coles:

And I think that's where knowing some of the common pitfalls in the current AI boom around chatbots is important. Is like what I anticipate over the next two to five years, we're going to have chatbots and AI summaries of everything in the automotive space. And every single one of them will have their own bent, their own objective. And so you'll have four different chat agents all telling you different things to do at that car. And that's where I anticipate, without guidance, without a focus on what are the real pain points of a dealer, how do we solve that? How do we maximize profit and take care of that dealer, make their teammates more efficient and their dealership more profitable? If you don't have that as the North Star and you have wholesale volume going to X, Y, Z auction as the north star, your chatbot's going to tell you to wholesale all your cars. And that's the bias that we have to be so careful to not introduce when we're providing tools.

Angela Simoes:

So once again, great transition. Lead into my next question because I wanted to ask, and this came up in another episode that we've done this week about training models.

Jamye Carpenter:

Yes. So that is where I learned this. He is the guru of what you're about to ask.

Angela Simoes:

Okay, excellent.

Jamye Carpenter:

Because I was going to ask it as well. 

Angela Simoes:

All the auction houses have data, maybe not as much as you guys. They're not online. Only dealers have data. There's data from third party sources that you can pull. And to your point about having AI that's not biased, is there an opportunity to train a large language model that could serve as a base layer or standard using data from all these different sources? And then yes, you're going to want a competitive advantage. So maybe you keep a specific set of data out of it, or you find another way to be competitive compared to all the other auction houses or online auctions. But it just seems like there would be some benefit to having a common layer or a common model that everybody can sort of build off of.

Jamye Carpenter:

Because it's already been trained. All of that data helps it to be trained accurately.

Angela Simoes:

Right. Does that make sense?

Dr. John Coles:

It does. I think that is where the industry could go. I think some of the challenges we'll run into, and I'm looking right at my friend at Privacy4Cars, is what is private information? At what level? What can be combined, what can't be combined? Am I violating an agreement if I put my data into a place that can be combined with something Jamie puts her data into? So I think that, go ahead.

Angela Simoes:

That’s easy to solve. Come on!

Jamye Carpenter:

That's why we've got privacy for cars. She solved it. And when I was at AutoCheck, it was the same thing. Who owns the data? Does the fleet company own the data? Does the auction on the data, is the data still owned by the previous owner before it gets in the hands of the new owner? And so there's all of these layers to what data you can provide, because privacy, number one and number two, nobody can agree on who owns the data.

Dr. John Coles:

And I think even within the models that we create on the machine learning, the LLM side, I am very thoughtful and I work with our head of legal, our CEO, others, are we following all of the contract agreements, all of our terms of service, we have to make sure that we are very clear on all those things just within our own modeling environment. I love that vision, and I think that's where we could go, but it's going to take time to know how to de-identify data.

Jamye Carpenter:

That is where you should use privacy for cars, folks,

Dr. John Coles:

Shameless plug too.

Angela Simoes:

I mean, there could be a consortium of sorts put together, or maybe it's an extension of your AI committee, something like that. Anyway, it just seemed like that would make sense so that you don't have four different chatbots recommending four different things. Right.

Dr. John Coles:

I don't really see a short-term path to not having four different chatbots recommending different things, but the space will innovate. I think right now, large language models, these chat bots, these automated capabilities, everybody's trying to help their dealers help accomplish their goal. We'll see a lot of first passes. I'm pretty adamant that we don't test them on our dealers and we test them internally first.

Angela Simoes:

Yes. That's smart.

Dr. John Coles:

That's a key thing in terms of building and rolling these tools out, is we call it “dog fooding” in the tech space. Are you dog fooding the technology or are you just like, we made a new chat bot? Hopefully nobody breaks it.

Angela Simoes:

Yeah. Good luck.

Jamye Carpenter:

Good luck. Fingers Crossed.

Angela Simoes:

What are you most excited about?

Dr. John Coles:

In the automotive space?

Angela Simoes:

Yeah,

Dr. John Coles:

I was like, I'm excited to see my kids. I'm excited there's a new restaurant downtown.

Dr. John Coles:

No. In the automotive space. I'm excited to see how the next generation of AGMs, GMs and dealer principles adapt and integrate new technologies in ways that make their lives more productive, their teammate's lives more productive, and make it a better overall experience for the car buying consumer. And that vision of removing pain points, I think is key. And with a shift in who's running some of the dealerships, how they're organized, it's been a shift that was happening and got accelerated through covid. I'm excited to see how the next generation can meaningfully integrate technology in ways that make their and their teammates and their customers lives better. Going back to my vision, right. Data to help people. 

Angela Simoes:

So that's what I'm most excited about. And it'll happen. And you're right. That's definitely something to be excited about because you almost don't even know what they're going to do, but they're going to do something really cool,

Dr. John Coles:

Right? They're going to do something cool. 

Angela Simoes:

Agreed. Well, this has been awesome. Thank you so much for joining us. Yes. And good luck with that and keep us posted on Oh, we will.

Jamye Carpenter:

On all the developments. He didn't tell you that he's really smart and he's really Dr. J. To me.

Angela Simoes: 

You're Dr. Coles?

Dr. John Coles:

Yep.

Angela Simoes:

Awesome. Well, thank you Dr. Coles for joining us. 

Dr. John Coles:

Thank you so much for having me. I appreciate it. It was a lot of fun.