Carvana: Evolution of the Inspection & Reconditioning Process

Disclaimer: This interview is for informational purposes only and should not be relied upon as a basis for investment decisions. In Practise is an independent publisher and all opinions expressed by guests are solely their own opinions and do not reflect the opinion of In Practise.

To hear it from you directly, could you walk me through your background in the recon business and with Carvana? We can proceed from there.

Yes, absolutely. Before all that, I started with CarMax fairly early on. I was with CarMax for about 12 years, mainly in purchasing and reconditioning parts of the business. In between, I had my own dealership for approximately five years, then I moved to DriveTime and stayed there for over two years. One of the main goals at DriveTime was to implement a lean reconditioning process, which I helped to build out, establish, and implement. At that point, Carvana was still quite small and shared the same office as DriveTime at the headquarters. I had met a few of them and they asked me if I'd be willing to move over. Their model was very similar to the one I had at my own dealership, where we did predominantly internet sales. We would meet customers at the airport, deliver the cars to them, or they'd come and pick them up from us. So, transitioning over to Carvana was quite easy. I was very interested in the business model and focused on implementing and improving the reconditioning model.

That's fantastic. So, when were you at DriveTime and Carvana?

I was at DriveTime at the beginning of 2012 and then moved over to Carvana in October of 2014.

When did you leave Carvana? Why did you leave and what are you doing now?

I left Carvana this year due to a severance. We had downsized the region I was in significantly and there was no longer a need for me to run it. Currently, a business partner and I have been working on creating a used car franchise model that we plan to build out. However, we're still in the development and licensing phase at this point.

That's an interesting idea. It makes sense.

Yes, we hope so.

It's like a franchise. In fact, it should be better than a franchise because you could bring financing.

Exactly. The average used car dealer across the country sells 50 to 75 units per month.

I see.

The market is quite fragmented with numerous opportunities. Many businesses focus on specific niches, such as lifted trucks with large wheels and tires, or cars tuned for racetracks. It's possible to assist these businesses in areas they may not fully understand, such as purchasing at auctions and reconditioning.

Why do you think this hasn't happened so far?

I'm not entirely sure. However, I believe there's an opportunity gap. The industry is very independent. Even at new car dealerships, the used car manager often believes they're the best buyer in the country. This mindset of not needing support from others is prevalent, which could be where the opportunity lies.

Could you tell me more about your role at CarMax? You were there for 12 years. What were your responsibilities?

A significant part of my role was purchasing. I managed the purchasing department across multiple locations. CarMax operates on a hub and satellite model, with a larger store and smaller satellites throughout the same market. These satellites purchase, produce, and send out cars to the main store. They had more five to 10 acre facilities, as opposed to 30 to 50 acre facilities. I managed one of these areas and the purchasing team, which consisted of managers and people who trained others on how to purchase vehicles at auctions. A considerable part of the job was also buying cars directly from consumers.

So, you were in charge of purchasing for one of the hub stores for 12 years. That's impressive. Then you owned a dealership for two years?

Actually, I owned the dealership for about four to five years.

I see, four or five years. So you left CarMax around 2008, after joining in the late nineties?

Yes, I joined around 1996 and left in late 2007.

Understood. How did you transition from purchasing to reconditioning?

I was interested in a new challenge and wanted to expand my skills. At CarMax, purchasing and reconditioning were closely tied. We would buy the cars, and then review and approve the reconditioning estimates. We also underwent Toyota Way training, which gave me a base knowledge and understanding of reconditioning.

I'd like to understand Carvana's reconditioning operations better. Were you based in the Winder, Georgia facility? Is that where you worked?

The location is Blue Mound, Texas.

Blue Mound, understood.

When I first joined, Winder was the only facility that was operational. I was commuting between Winder and Blue Mound, essentially managing both. Once Blue Mound was sufficiently established, I focused solely on it.

And you were in charge of Blue Mound. What was your role?

Yes, I was the director of the facility. At its peak, we had nearly 1,000 associates working there.

To whom did you report while managing the Blue Mound facility?

I reported to the VP of ICs, Mike Rennie.

Understood. Is he still with the company?

Yes, he is still there.

Thank you. I'd like to understand Carvana's inspection and reconditioning process. I want to visualize it as it was around 2018 to 2019, when Blue Mound was being expanded. Then, I'd like to compare it to how it was just before the pandemic in 2019, how it was in 2021, and how it was in 2022 and now. I believe there have been significant changes over time. I'm particularly interested in understanding how the efficiency of the process has evolved. Could you guide me through this?

Absolutely. The core of the process has remained fairly consistent over time. At a high level, the reconditioning process involves six different stations where work is divided. Tasks are grouped together and broken down into roughly 15-minute intervals. This way, all stations operate at the same cycle time. After the inspection process and estimate creation, parts are ordered and the car moves to the cosmetic phase. This includes paint, PDR, glass repair, interior repair, etc. While parts are being ordered, work on the car continues, preventing stagnation. Ideally, by the time the cosmetic work is done, the parts have arrived and the car moves to the mechanical reconditioning phase. After that, it goes to detail, which is also divided into six stations based on cycle time. Finally, it goes into the photo process. One unique aspect of Carvana is that we don't have a brick and mortar.

I've had the opportunity to visit Winder and Blue Mound and tour them, so I have a clear picture in my head. I understand that the cars come in, go through a staging area, an inspection process, and then work is allocated for those cars. They are then routed across different stations for cleaning, mechanical work, paintless dent repair, and so on. After they are photographed, they are placed in a finished area. From the outside, this process may seem simple, but it appears to be quite complex and challenging to execute well, given the variations in the process.

Indeed.

Let's start with the 15-minute stations. Why is that the optimal approach? Why not route vehicles through the stations, knowing the average time it takes for a vehicle to be processed at each station, and then maintain the right amount of inventory in front of every station to maximize the system's throughput? Do you understand my point? Why aim to keep everyone operating on a 15-minute cycle? Why not say, for instance, this station takes an average of 25 minutes, so we want them to have one car in process and one car in inventory, and then route cars to manage flow through the organization?

You do have zones in front of each of these different stations that can fluctuate based on the variability of the cars. For example, a Smart car and a Suburban may come in one after the other. Even if they are both similarly maintained and clean, it physically takes much longer to process one than the other. Ordering parts for a car can also introduce a significant amount of variability. So, having some work done and ready in front of each of the next stations, as you suggested, does make sense. You do want to maintain a certain minimum and maximum number of cars in each zone. If you reach the maximum number, you would either slow down your inspection process or your detail process to maintain the right number in that zone in front of the next one.

I see. Let's approach it this way then. Can you provide some basic statistics? Roughly, how long would it take from the time Carvana purchases a car until it is ready to be sold to a customer, as of your departure in 2023?

From the point of purchase to the point of being ready for the website, it was probably closer to 18 to 20 days. As you probably know, with all the media attention it has received, you have to be very careful about when you have the title. Legally, from the point of purchase, they have 30 days to supply it. So, you could have the car completed well before that, but it still just sits until the title is there before you can put it on the website. There are a few different ways to measure cycle time. One is when Carvana is photo complete and ready for the website. From a reconditioning standpoint, you have no control over whether the title is there or not. So that's where Carvana's cycle time calculation ends.

How does that 18 to 20-day period compare to, say, 2019?

The time span was not as long as one might expect. It was probably around 16 to 17 days. Yes, the product had lower mileage, fewer parts, and required less reconditioning, which theoretically should have allowed us to get the car more quickly. However, I believe we just improved and streamlined the process over time, which helped us maintain a relatively close gap.

If I consider an average of 18 to 20 days, and think of that as a distribution, my understanding - though I'm not an expert on titles - is that the time it takes to get a title can vary significantly. This depends on the state, the financing partner, and whether the owner has any financing at all when they sell the car to you, among other factors. How should I interpret the distribution of these 18 to 20 days?

Are you asking what percentage of cars actually make it to the website at that point?

I was thinking more along the lines of fast cars versus slow cars. For instance, a car you pick up that has the title and requires light repairs, do you get it up on the site two days after you purchase it? And a car that, let's say, encounters numerous issues and is still sitting on the lot 100 days later. If I were to consider a standard deviation, a normal distribution around 18 to 20 days, what would that distribution look like?

As you mentioned, it is quite variable. Very few cars come with the title, so that's a relatively small number. On the other hand, not many cars will take up to 90 days. From an auction standpoint, you have the option to return the car if the title isn't supplied within 30 days. However, by that point, you've already invested in reconditioning and transportation among other things. These are costs you would have to absorb. So, you generally wait for the title to arrive. I believe the average car, with its title and reconditioning completed, is listed on the website within 25 days.

Is there a significant opportunity to reduce that 18 to 20-day period? For instance, I can imagine that cars with instant or immediate titles should be prioritized, because the critical path doesn't include obtaining the title. Is there an opportunity to significantly improve this process either by speeding up the title acquisition or by routing the cars that already have titles through the process faster?

Yes, I believe there is an opportunity to prioritize based on certain factors. However, this was never really implemented, especially after 2019. The prioritization was mainly focused on vehicles that had parts ready. This was because parts and supply chain issues became such a burden, making it difficult to get parts in a timely manner, if at all. Therefore, units were prioritized based on the number of parts needed or whether the parts were available. At one point, we had a model set up for prioritization based on the likelihood of sale for a particular make and model. However, if a car goes through the system in five days but then waits another 20 for the title, is that really helping us?

After 2019, did parts become a more significant delay than titles, hence becoming the critical path? Was the system managed based on parts availability? Is that why prioritization was based on parts availability and speed? Please help me understand how you source parts and how that creates delays. I imagine there's a wide variation. Some cars might only need oil and tires, while others might need a specific part that takes five weeks to arrive. Could you explain how you identify vehicles based on their parts needs and how you source those parts? Ben mentioned progress in the part sourcing process. I'm interested in how part sourcing in 2023 compares to 2021 and 2019.

Yes, availability was a significant factor. There was a high demand post-Covid, and like every other product, once the supply chain started to tighten up, we saw some progress in 2023 because parts became generally more available. Carvana transitioned all of its facilities.

Previously, a vendor managed the parts department. Now, we have an in-house parts manager and associates, people with a vested interest.

Was it O'Reilly?

It was Advance. The model shifted to Advance, which added distribution centers. They bought and stocked up distribution centers near the inspection centers to supply parts more quickly. However, by that time, we had already transitioned to in-house parts managers. They still supply a lot of parts. They just do it from a standpoint that they should have adopted in 2018 or 2019 to supply enough parts.

But practically, when a car comes in, it gets inspected, and a parts list for that vehicle is generated. How do you decide which parts to buy from Advance, which parts to buy from LKQ, and which parts to buy from somewhere else? How do you determine the right price for those parts? How do you source the parts?

There's a list of hundreds, if not thousands, of parts that Advance keeps on site. If any of those parts are needed, we order straight from them. If we need something outside of that, the parts associate employed by Carvana starts to look for the best price and availability. Sometimes, the best price might take two weeks to arrive. If we can get it for a slightly higher price, say $20 to $40 more, but it can be here today, then the cost of the car sitting there for two weeks makes up the difference. That associate needs to make the decision of where to purchase it from, whether it's LKQ, WORLDPAC, or a different distributor like O'Reilly. However, if Advance has it available, they're supposed to be able to get it to us within two hours, so we go with them first.

Could you explain how the parts associate uses the software and how this process has evolved over the past five years, comparing 2023 to 2019?

Yes, they do use a software, but some parts of the process are still manual. For instance, they might have to shop around on platforms like Ebay if they're having trouble finding a part. However, the majority of the process, let's say 80% to 90%, is automated. The software searches for the parts, compiles the results, and shows them where to find the best product at the best price and availability.

When the software is searching for the best price and availability, does it take into account factors like title availability? For example, if the title is 40 or 20 days out, are you willing to wait eight days?

Yes.

Is that integrated into the software?

No, it's not integrated with the title. The system is set up with the expectation that the title will be available by the time the car is going through the process. There's no cross-check for that.

It seems like there's a lot of potential to streamline the process by routing vehicles through the system based on title availability. There's also an opportunity to identify the critical path for each vehicle and prioritize them based on the amount of slack in their schedule. Additionally, the price versus cost of parts could be prioritized based on the slack from the title element of the critical path. Is there a significant project here that could save five to 10 days on vehicle speeds and reduce the cost of parts?

I agree with you. From a high-level perspective, there's a lot of potential. Although the reconditioning process has improved and has helped lower the cost of goods and increase the gross profit per unit, the main focus has been on growth at Carvana. The majority of the software, technology, and focus has been on the front-end user and building out a robust logistics network.

Despite being considered a technology company, Carvana hasn't invested much in improving the inspection and reconditioning process. This process is still relatively simple and outdated, and there's technology available that could be leveraged. There's a lot of room for them to refine this process and increase efficiency. However, their priority is still to improve the front-end user experience. I'm not saying this is necessarily wrong, but there's still a lot of potential to improve the reconditioning process. It's far from being as good as it can be.

Let's compare Carvana circa when you left, to Carvana in 2021. They claim to have made significant investments in productivity and have recently presented data showing a decrease in their inbound freight and reconditioning costs over time. First, do you believe this is true? Second, if we were to inspect the reconditioning process at Blue Mound at these two different times, what differences would we see?

The main issue, which is being addressed, is the aggressive and rapid growth. We were the first to market and had to get the volume out there. This led to a high dependency on vendors for paint work, PDR, interior, cosmetic work, and even some mechanical work in many cases. When things slowed down at the end of 2022 and the expectations for 2023 changed based on volume, the focus shifted to training our associates to handle the volume and transition all the vendors out. The goal was to do everything internally, which would be more cost-effective. This is where the change in price and reconditioning comes in. It's a big question right now, with the availability of used cars, increased mileage, and deteriorating condition. It costs more to recondition cars today than it did 10 months ago, as other companies like AutoNation and Sonic have reported. However, Carvana has managed to reduce reconditioning costs. It's not because they have different cars or only buy cars with lower mileage. Even to reach 300,000 units, they still have to buy cars that have higher mileage and are older. The way they reduced reconditioning cost was by phasing out and eliminating the vendors.

Let me summarize what I understood. They were growing quickly, which made it difficult to source and train more technical skills like paintless dent repair, painting, and some mechanical work. So they hired vendors, which came at a cost. The work might have even been done off-site, adding time and cost due to transportation. But with the slowdown, they've been able to eliminate the vendors and continue to further their staffing goals. Now they can insource paint, PDR, and mechanical work, reducing costs and improving efficiency.

That's exactly right.

Could you help me understand the cost to recondition a car? What constitutes that cost and how has it changed from 2021 to 2023? I imagine there was a KPI for recon costs per vehicle that you had to meet.

We can discuss some generalities. I don't know if they've ever publicly disclosed the numbers, but other groups like AutoNation and Sonic have reported (reconditioning) costs of $1,700 to $1,800 per unit to recondition a car. Carvana was never that high on average, but let's say it was $1,500 in 2021. If you've taken $200 off that per unit today, it's still higher than where you were in 2019.

I see.

The product is indeed quite different.

It's older and there has been four years of inflation.

That's correct

The $1,500 you mentioned, does that include inbound freight?

No, it only includes the labor and parts put into each unit.

So, it doesn't include overhead costs?

No, it only includes the labor for that specific unit. For instance, if four hours were spent on a car, then it's four hours of labor.

Where did you get the data point for Sonic and AutoNation? And what does it include?

The AutoNation data was a general figure. I believe AutoNation mentioned in their last earnings call that their costs had increased to around $1,800, and even up to $2,000 in some cases.

That's useful information.

Their reconditioning costs have definitely increased. To be honest, I've never delved into how they calculate their reconditioning costs. But I assume it's generally the same across the board. The main difference might be that some groups add a little more impact cost per car, such as a $400 fee to cover general administrative costs. Whether that's included or not, I'm not certain.

How should I break down that $1,500 or $1,200 between parts and labor?

There's a wide range there too. Especially with doing as much as possible in-house, you're likely to see more than half of that as labor. So, on an average unit, over 50% would be labor and less than half would be parts.

In 2021, would it be correct to assume that the time to recondition cars was much higher than the current 18 to 20 days? What was the duration?

It was easily in the low to mid 20s. There were so many cars coming in and most facilities had a space constraint. It was a significant challenge moving cars just to access others, which impacted the process. Furthermore, the staffing models were incorrect as they were based on cars that were three years younger with 20,000 to 40,000 fewer miles. So, when your average car added at least half an hour of labor, depending on the facility, and if you're processing 1,000 cars a week, which some facilities were, that's an extra 500 hours of labor not accounted for in the staffing model. If you were already struggling to maintain staff levels, it was extremely difficult to keep up. So, not only were there issues with the literal space constraint of having 8,000, 9,000 or 10,000 units in a facility built for 7,500, but the hours impact also changed significantly. Therefore, your average car was there a lot longer because it wasn't getting through the process as efficiently as it should have been.

You've had experience with reconditioning at CarMax and Carvana. Why do you think Carvana could be, as you seem to suggest, up to $600 more efficient per unit than Sonic or AutoNation, or even CarMax? What factors might contribute to this?

Having intimate knowledge of the situation, I doubt they could ever surpass CarMax significantly. Honestly, I don't believe they're there now. They're probably pretty close, maybe slightly below where CarMax stands today.

Why would they be significantly better than Sonic or AutoNation if CarMax is already so efficient?

One major change is that every car purchased for Carvana, about a year ago, had to be reconditioned and put on the website, regardless of its condition, as long as it didn't have safety issues or frame damage. This was due to the need for inventory growth. However, Carvana has learned from this. Now, they assess during the inspection process if a car is worth the reconditioning cost. If it's not, they send it to the wholesale auction. This decision is made for a lot more vehicles than traditional dealer groups would. Although it seems like a small piece, when you add up a few of these a day or 20 to 30 cars a week, the averages start to add up.

Comparing the Carvana facility to the CarMax or AutoNation facility, it's much larger and has a higher throughput.

Absolutely, yes.

Is that fundamentally more efficient? Or is it just a different way to recondition cars? What's your opinion?

It's both. It's a different way to recondition, but theoretically, it could be more efficient. Compared to traditional used car dealer groups, it is more efficient today. However, they still have a long way to go to match CarMax's efficiency. They still have a lot of room for improvement. There's a general sense of encouragement that they've improved and reduced their prices, but there's still potential for further improvement. I believe this is a source of motivation for the Carvana team.

Where do you see the biggest opportunities for improvement? We've discussed routing vehicles to improve speed based on their critical paths. What other significant opportunities do you see to reduce the approximately $1,200 of reconditioning costs incurred per vehicle?

In my opinion, the inspection process as it stands is probably the area where the most significant improvements could be made. The process involves many people, which leaves a lot of room for errors, such as transposing numbers or missing something. These errors then have to be corrected, which requires the process to be repeated.

There is technology available today, some still in development and some already in existence, that can streamline the inspection process. For instance, using scans of the vehicles and automating measurements that are currently done by hand. This could not only save time but also increase consistency. It would also improve data capture, leading to more consistent readings from the inspection process.

The inspection process is the source of most rework, such as identifying missed or incorrectly identified issues. Therefore, investing in this technology could lead to significant improvements. However, I understand that it's a big investment, and that's probably why some companies, including Carvana, haven't adopted it yet. It would require a significant financial commitment and would change many processes. But I believe that this is where the most significant changes can be made in the future.

So, you're suggesting a more efficient front-end inspection process using technology to better identify necessary work and plan more efficiently?

Yes, but I don't believe it would be slower. In fact, I think it would be much faster.

I see. The overall process would be faster, but the inspection itself might take a bit longer. What would Ben Huston or Mike Rennie say if asked why they aren't implementing this?

At this point, I think it's a matter of cost. Implementing such a significant change and investing the necessary funds at this point is a big commitment. There's a strong focus on proving that Carvana can be profitable, or at least achieve a positive EBITDA. Undertaking a multimillion-dollar project is probably not something they're willing to do right now.

Why would it be so expensive? The inspection is not a significant part of the overall process, and the routing of cars would change as a result. You could test it on one line. Why would it be so costly?

I believe the initial investment in the technology and equipment would be quite expensive.

I see. But couldn't you test it in a low-tech way? You could have someone record everything in simple software, which could be developed in a few days. Why would it be expensive to test?

I don't necessarily think it would be expensive to test. However, if you're asking me what the pushback might be, I believe that would be it.

Let's shift our focus to the Blue Mound plant. How many production lines were you operating while you were there?

We had six production lines, three for each shift.

What was the capacity in terms of lines?

That was the maximum.

So it was operating at full capacity?

Yes, completely.

Interesting. So, they can't expand production at the Blue Mound plant. It's already running at its maximum.

Well, as of today, in 2023, I believe it's only running two lines.

I see.

Perhaps I misunderstood your question, but yes, at its peak, we were operating all six lines.

Understood. Now, let's assume the company starts growing again and they want to maintain and even improve their efficiency. They don't want to regress in terms of cost or speed. In fact, they want to maintain the capacity to continue implementing projects for continuous productivity improvements.

Correct.

How quickly do you think they could increase from, say, two to three to four lines at Blue Mound?

Realistically, it would take about 120 days to go from deciding to restart a line to full production. This includes the time needed to hire and train new staff. If you consider the need to train new leaders and all associates, and the fact that some of your existing staff may have been reassigned, I think 120 days is a reasonable and realistic target. However, I doubt they would set the target that far out. They would probably aim for 60 to 90 days.

That's quite remarkable. If you have two lines and you want to expand to six lines, based on what you've said, if you opened one line at a time and each took 120 days, that would total 480 days. That's not too long to triple production.

I agree. However, it also depends on whether we're adding lines one at a time. Typically, they would add two lines at once, which would make the 120-day timeline quite aggressive. But that's roughly how they've expanded in the past.

At the end of those 120 days, would that line be productive, or would it still have worse KPIs than a line that's been running for six months, a year, or two years?

If you're comparing line to line, yes, it would still not be as efficient as the lines that have been running for a longer period.

How long would it take for it to reach full productivity?

Certainly, within six months, which I understand is much longer than 120 days, you would expect a department to be running at the same Key Performance Indicators (KPIs). However, it's rare to have them broken down in such a way that one operates in isolation. For instance, if a car goes through the inspection process in line A, it doesn't necessarily stay in line A for cosmetic and detailing. Cars are pooled and cycled back through. So, the real question is whether the overall facility operates as it did when it only had two lines, not whether line A outproduces line B.

Would this involve relying on third-party vendors, or would it be about maintaining the current staffing model to sustain this pace of expansion?

I believe it would involve maintaining our in-house and internal associates to do all that work within 120 days. You could probably turn it faster if you relied on vendors.

Would the Inspection and Reconditioning Center (IRC) management team be primarily focusing on expanding the new line, or do you think there would still be bandwidth to continue working on projects to enhance overall productivity in other parts of the facility?

The focus would certainly be on improving other parts of the facility. The question of whether there's enough bandwidth is a reasonable one, but that would certainly be the goal and expectation.

Can you open one line and then immediately start opening the next, or do you need some period in between to stabilize operations, set a new baseline, and ensure all your productivity stats are where they need to be before you can open the next line?

You can definitely open one line right after the other. Most of the experience at the facilities today involves opening lines simultaneously. The biggest challenge to that is hiring the right amount and the right kind of people. This is usually what slows down the process considerably. Especially in markets where you've built greenfields in Bessemer, Alabama, and West Memphis, where there's a smaller pool to hire from. This becomes a challenge when you're trying to open more than one line at a time. I believe that Carvana's future expansion will be slower to maintain metrics, particularly reconditioning costs. In my opinion, they won't expand anywhere near the rate they did before. Much of the previous expansion was proof of concept, and now it's about proof of metrics. I think the expansion would be much slower today if they started to expand again than it ever was in the past.

Taking a step back, how fast do you think Carvana can expand annually in percentage terms before they start to encounter operational challenges that begin to impact their profitability metrics?

The company was established and prepared with enough employees and infrastructure to handle over 500,000 units sold. The question is when they will start aiming for that target. Currently, they're around a 300 target, but if they increase to 350, 400, or more, it depends on where they set their sights. From an infrastructure perspective, the capacity for 500,000 units is already in place. Furthermore, the potential of the ADESA facilities has not been fully utilized. Therefore, it's reasonable to suggest that with the current infrastructure, they could potentially handle close to a million units.

That refers to the physical infrastructure, correct?

Yes, that's correct. Please continue.

How quickly can they increase their workforce without negatively impacting performance, while maintaining or even improving it?

For each line and facility, a 120-day timeframe is probably the most aggressive realistic estimate for bringing in the right amount of people in the right amount of time.

If I'm expanding a facility and I have six lines available, which line is the most challenging to ramp up? The first, second, third, fourth, fifth, or sixth? Is there a difference?

It's more about shifts than lines. Each line operates in two shifts.

So there are twelve shifts in total?

In the Blue Mound scenario, we had three production lines running on the first shift and the same three lines running on the second shift.

Okay, so there are six shifts across three lines. Understood.

The second shift is more challenging, regardless of whether it's line one, two, or three, because it's at night. Currently, Carvana has essentially eliminated the second shift and everything is running on the first shift. As they begin to add more shifts, which will predominantly be second shifts, it will become more difficult. Now that I think about it, the 120-day timeframe might be more challenging than we initially thought.

I understand. However, I was considering the possibility that if a plant is operating with two shifts, hiring for a third shift might be easier because some employees could be promoted. If the plant has been running on two shifts for six months or a year, there might be people ready for promotion and skills built up within the organization. Conversely, hiring the first person or trying to rapidly increase the workforce could lead to inexperience. Is there something like that which makes it easier or harder to expand as the plant matures?

You're correct. The first line is always the most challenging to hire because it involves recruiting team leads, reconditioning managers, and other roles. Ideally, you'd have around 110% or 120% of your line hired and ready to go, as this is necessary from a leadership standpoint. If you're able to hire from within and add your line leads, then line two would be easier to fill, and line three even easier. However, at some point, you'll face diminishing returns as you may not have enough talent and you may spread your existing talent too thin, putting someone in a position they're not quite ready for yet.

You mentioned it takes 120 days to hire two shifts. Is that for one line or one shift?

I would say one line.

So that's two shifts?

No, I apologize for the confusion. One line refers to one shift running one line of production.

I see. So, if we take your 120 days as a standard, in 720 days a facility could be fully ramped up, assuming you follow this pattern. However, it seems that might be a bit aggressive. If we added a 30-day buffer for each ramp, then we're looking at roughly three years to go from zero to full production.

Yes, that seems reasonable. It's still quite aggressive, but not as aggressive as our growth rate a few years ago.

Indeed, it's not even close. Currently, they have many facilities that could add another line. So, in theory, they could ramp up quite quickly. However, as the growth becomes larger relative to the number of facilities, it becomes more challenging. Now, how did Blue Mound compare to other facilities in terms of the key performance indicators (KPIs) that mattered? And what were those KPIs?

I'm interested in Blue Mound's performance compared to other Inspection and Reconditioning Centers (IRCs). How did you rank and what were the key KPIs you were held accountable for?

The three main KPIs were cost, speed, and quality. We measured how much it cost per unit to process cars, how quickly we did it, and the quality of the work. We used a quality control (QC) process at the end, or customer feedback after sale. The ranking was very fluid. You could be a hero one week, and the next week you could be at the bottom. When I left, we had 17 facilities, and Blue Mound usually ranked around seventh or eighth. However, we had plenty of weeks where we were in the top two or three.

We've discussed cost and speed, but not quality. How did you track quality? What was the defect rate? And how did you determine whose fault it was if there was a defect, whether it was the IRC's fault, the trucker who damaged the car, or perhaps a misunderstanding on the customer's part about the value proposition? How did you figure out what was your fault and what wasn't?

From an after-sale perspective, as you've identified, determining whether an issue was pre-existing or occurred after the sale is not straightforward. This issue didn't rank high on our priority list. Our main focus was to fix the problem, do what we could to keep the customer satisfied.

We had monthly discussions where we reviewed all the cars from your facility that had post-sale issues. We would examine the photos taken in the photo booth and try to determine if the issue was visible in the picture, whether it was present before or occurred after the sale. However, establishing precise KPIs to track this is still likely a work in progress.

Could you tell me about the focus on process and rigor in your work? I'm interested in understanding what a typical meeting with you and Mike Rennie would have been like in 2019, 2021, and what it might be like in 2023. What were the main topics of discussion? How frequently did these meetings occur? What were the expected outcomes? I'm trying to understand the focus and rhythm of the organization. What would have been the top three items on the agenda for these meetings? How would you have approached these items?

The difference between 2019 and 2021 was significant. In 2019, our primary focus was hitting the production target. We were concerned with volume and believed that we could tighten up other costs later. In retrospect, 2021 was even more focused on production. We were trying to catch up with demand and needed to increase our unit production.

The most frequent discussions I had with Rennie, Ben, or others were about production. We questioned if we could hire enough people and if we could meet our production targets. These were weekly conversations with Rennie, and it wasn't uncommon to have additional phone calls in between.

With Huston, our meetings were more on a quarterly basis and focused on broader topics. We didn't discuss specific cars at Blue Mound, for example. However, things changed in late 2022 and 2023. We started to focus more on actual costs, such as parts and reconditioning costs. We aimed to reduce these costs and align them with our targets. This became more important than the production number itself, especially when we were reducing the overall number of cars on our website.

The second most common topic of conversation was cycle time, or how quickly we were getting things done. Even in 2019, we had many discussions about how we were calculating cycle time. Should it start from the moment inspection begins, when the car arrives at our facility, or when we've paid for the unit? In some cases, it took seven to 10 days for a car to reach our facility. Should this be included in the cycle time? From Ben's perspective, if we've already paid for it, the clock has started.

We were still establishing many of these scenarios in 2019. However, the main topics of conversation in 2019 and 2021 were always production numbers, cycle time, and staffing. Could we get enough people to keep things running? This focus shifted towards cost efficiency later on.

When did that shift occur?

It happened in late 2022.

So, mid to late 2022? I would have thought it might have been earlier, perhaps in March, April, or May of 2022.

It could have been late spring or early summer of 2022 when we really started to focus on costs. The quality metric was also included in the overall production metric conversation. If cars had to be repainted or if a part was missed and had to be reordered, it significantly impacted our overall production. So, in essence, we were discussing how to minimize these errors and reworks to help reach our production number.

Could you take me through the process, down to the brass tacks? I understand you were in a meeting with Ben and/or Mike Rennie last week and there were some issues. Could you provide a specific example of what could go wrong, something that could increase delay, cost, or quality issues? How would you perform a root cause analysis and implement changes to your processes? Furthermore, how would these changes be integrated into the system to prevent the same problem from recurring? Please walk me through the full cycle of root cause analysis, diagnosis, and process change, using a specific case if possible. For instance, something like changing a battery.

I would say that the biggest bottleneck for most facilities is probably the paint. There are many variables to consider, especially as Carvana was ramping up and building pre-2019. The initial focus was on highly skilled painters. We needed to train them for this volume. There are plenty of painters we could hire, but no one has seen this kind of volume from a single facility.

The real skill, however, lies in the bodywork. A great painter can't improve poor bodywork, but excellent bodywork can make an average painter look good. We overlooked this aspect and it took us a while to realize that we needed to shift our focus and train people to do the bodywork correctly.

The bottleneck was at the paint stage, which is the slowest and hardest to build up at any new facility and even maintain. This is due to the variability with weather changes, humidity, and how you mix the paint. It has always been the bottleneck.

The solution became using vendors. We started sending our cars to vendors to get them done. This process not only took longer, but we also had to pay for transport and inflated vendor costs. However, we were able to meet our production target and get the cars through.

Reconditioning vehicles may seem simple, but it's extremely complex. We were essentially running seven to eight different verticals of a business. We would often discuss why we struggled to get these cars through paint and then had to wait for parts. This led us to the conclusion that we needed to take on parts in our own department and supply our own parts in a reasonable and consistent amount of time. We discussed what changes we needed to make and where in the process we needed to order parts.

We considered whether we needed a larger buffer for painted cars so we could get more through and have everything flow more directly. We questioned whether our min-maxes were wrong and if we needed to change them in each zone to ensure we had enough cars in process at one time. This would change the formula for how many cars we're keeping in WIP and how many cars we have to keep on a lot at the same time. We had to determine where the change needed to be made if we weren't able to fix and solve the issue. That was a simplified version of that conversation.

So, how did this feedback into the process? Were each of the IRCs allowed to operate more or less according to how the director of the IRC thought it should run best? Or did you find yourself increasingly following a shared playbook, where different lessons from one IRC would affect your operations at another? For instance, we've learned that the right buffer should be three cars at paint, and that's what we're doing now. That's what we're doing everywhere. Or was it more like, in Blue Mound, they've made it work with five. How did you actually think about this?

I think you've described it accurately. There was probably a target of, let's say, 65% to 75% of everyone running the same way in 2018 and 2019, across every facility. At that point, nothing was a greenfield. Nothing was built from scratch. At Blue Mound, it's eight different buildings there. It's not under one big roof. It's not run like a Bessemer or one of the ones that were built from the ground up. So, you've just got to make the best out of what you have.

This then evolved into us developing a playbook with best practices at each facility. We wanted everyone to be running more like 85% to 90%, if not more similar than the way it was before. So, yes, it kind of became a case of you having the autonomy to run your facility as you see fit, just get to the metrics. And as we learned from that and took best practices, that playbook was developed.

Do you think if and when this company goes back to growing, that they can, by adhering to these playbooks, return to growth without sacrificing much of the progress they've made?

I believe so. I think they've got a much better, clearer direction and ability to either start a new facility, or in this case, probably expand your leadership team at the existing facilities. And you've got, like I said, more of a playbook that someone can transition to rather than just needing you to figure it out quickly.

On a practical level, how much did they manage to save in terms of parts, both in terms of dollar amount and also, more importantly, probably in terms of speed? How much improvement were they able to generate there from insourcing more? I believe they also did more aftermarket parts, stuff like that.

Right. And I think a significant part there, too, is once you expanded your talent level or your ability for the associates internally, there are things that you can just repair rather than order a part. For instance, some of your interior trim pieces that were scratched or cracked in half, like being able to plastic weld something and get it back and make it look as if you can't really notice that it was ever that way before. So that had an impact on that as well. But overall you're probably talking, let's call it $200 a unit. I think that's maybe not exact, but pretty close.

And that's from when to when?

I would say that's probably over a full 12-month period.

Sorry, they saved $200 a unit from what year to what year?

In terms of parts, I would say 2022 to 2023.

That's a significant savings. And that's mostly, I guess, best sourcing. That's the idea of best sourcing and.

Sourcing and time. If it's costing you, say, $10, $15, or $20 a day, that's a factor to consider.

That includes speed as well.

Yes.

Have you seen progress in titling speed? Could you tell me about how the titling speed has evolved over time? It seems it improved, worsened, and then improved again. Could you walk me through that story?

I believe most of the improvements in titling have occurred more recently, after I left, to be honest. Everything slowed down after Covid. Titling, which might have been somewhat of an issue pre-Covid in 2019 and before, wasn't as glaring of an issue. It really became apparent when everything slowed down with title exchanges post-Covid. There are stories out there, and Carvana was certainly one that received criticism, but from a percentage standpoint, they weren't anywhere near some of the biggest offenders. I'm sure you've heard all that. I don't want to get into that detail, but they struggled with it for sure during the 2020 to 2022 period. They took some media hits, and eventually reached the point where they decided to let professional companies manage the whole title process. In some cases, it's even faster to go to the DMV and get a replicated title once you show proof of ownership. Once they transitioned to that, I think you've seen them improve significantly.

I thought the plan to fix titling was to move away from third parties and fully insource titling.

It's been internal, but every market had its own third party to use. I think now there's one group helping them do it all.

But more specifically, in the IRCs, how much did the delay in titling specifically affect the time it took for cars to be listed or saleable? How much did that evolve over time? For instance, cars that were available in seven days might now be available in 14. I'd love to understand how that changed.

From an IRC perspective, it had virtually no impact.

Right, because you weren't routing cars based on title availability.

Correct. The main issue was that cars would go out to the part of the lot where, for all intents and purposes, the website existed. In these huge zones with thousands of cars, a crew would go through and check on cars that hadn't moved within 15 to 30 days. That was the only time you'd really know if a car wasn't even on the website yet.

So, tell me about the trends you observed. It seems like you didn't see it. It would have mattered, but in a strange way, it might not have because you would have reconditioned a car that you didn't have a title for. And then it might have gone on the site or not, but if it did, it would only have been available after some distant period of time when you would have the title. Is that the idea?

Yes.

So, theoretically, a better approach would be to recondition the cars that can be listed for quick delivery, while allowing the cars without title to wait. Have you noticed the title times compressing again after they extended in 2021 and into 2022? Or have they remained extended?

I believe they are compressing now. They probably started compressing over the last six months, but it was a challenge throughout the entire period from 2021 to 2022.

Could you tell me about your experience with cancellations and car movements related to orders that were not ultimately fulfilled? Was that a significant issue for you while managing the IRCs?

It didn't really impact the IRC or the reconditioning process directly. Once a photo was completed and the car was listed on the website, it was no longer the responsibility of the IRC, from the Blue Mound facility. Although I didn't directly oversee logistics, there were some indirect connections, and the reporting structure was somewhat unclear.

However, they would have been loading and unloading cars on your end. So all of that activity was happening, but it wasn't your issue, correct? Did you ever have to re-detail or recondition cars that were moved for orders that were later canceled? Did that affect the workflow?

We had a post-production team that handled some details on cars. Mainly, if the logistics team needed a specific car to be staged and ready for sale, they would retrieve it from the lot. If they noticed that the car was still dirty or had an issue, the post-production team would handle it at that point. That was somewhat under my purview, but it wasn't a direct IRC task.

I've always thought that Carvana should sell tires and rims, and offer customers the option to fix any annotated scratches for an additional fee. As the person who managed the IRC, how operationally challenging does that sound to you? Does it seem like a total nightmare or not a big deal?

To be honest, this was a significant topic of discussion around 2017 and 2018. We considered adding ancillaries, like not including a second key in the reconditioning process and offering it as an add-on during the sale. Tires, wheels, window tinting, and other ancillaries were part of that conversation. However, it would have required expanding parts of our facility or building another one to accommodate these processes.

But don't you already change tires? Couldn't you just run cars through the tire changing station more often? You would be changing more tires, but it would be the same process, right?

If you weren't at max capacity. In the Blue Mound scenario today, they could absolutely do it. But with six lines, there wasn't room for it.

So, the lines weren't designed for it?

Yes, the lines weren't designed for it. You couldn't even plan or staff for it because you never knew when something would be available.

Cutting keys seems incredibly straightforward. Is it operationally challenging to produce more keys? If you offer people an extra key, some percentage might take it and you could make $100, correct?

Yes, it does slow down the reconditioning process significantly. Especially with high-end German cars, you can't cut the keys yourself. You have to get them from the dealer, and they're not always quick to assist. For instance, to cut a Mercedes or a Tesla key, you need to have the title. This leads us into another issue, as we're not even supposed to have the title yet.

I see, that makes sense.

Even something as simple as ordering and fitting floor mats into a car can slow down the process. This is another task that could be done as an ancillary service to cut down on process time.

I understand. But it seems like all these tasks were put aside to focus on excelling at the core reconditioning process.

Yes, that's correct. We were not willing to slow down the reconditioning process or divert production to add ancillary services.

That makes sense. To clarify, when you mention recon times of 18 to 20 days, is that from the time of vehicle purchase or from the time it arrives at the facility?

That would be from the time it arrives at the facility.

What do you think the ideal recon time should be?

I believe the target should be 14 days from the time the car lands on your property to the time it's ready for website photos. The way we calculate this is by starting the clock when the car is brought in for inspection. That's what you can control. Because one week you might have 30 cars arrive, and the next week 700 cars could arrive.

I see. So even if the cars land at your site, you only start the clock when you scan them in because the inventory at the front of your site isn't something you can control. There's variation in that flow.

Carvana should control that, but the Inspection Center (IC) itself doesn't necessarily control that. This changed when Carvana shifted from relying heavily on auctions for inventory. Previously, third parties would bring in cars in large quantities, often when the IC wasn't operating.

Now, with first-party consumer sourcing, the process is much smoother and more predictable. This also frees up lot space. Are cars sourced from consumers generally easier or harder to recondition, or does it just depend on the year and mileage?

The factors are really the year and mileage of the car. For instance, CarMax had a lot of data suggesting that if a car is purchased in DFW, it's more likely to be resold in the same area. This is due to market preferences and styles. Although this was considered anecdotal data because the exact reason couldn't be pinpointed, it was consistently observed. For example, if DFW was buying 150% of what they needed, and we decided to send cars to Houston, those cars always sold slower in Houston.

That makes sense. We've covered a lot of ground. Is there any other aspect of the historic or potential evolution of the Carvana IRC model that you think is interesting or important for the next five to 10 years that we haven't discussed yet?

From a five to 10-year perspective, I believe we touched on it, but it's worth reiterating the importance of updating the reconditioning model to be more technology-driven. There's a lot that Carvana could do to improve the inspection process by leveraging technology, which is a cornerstone of the company. I believe there's significant room for improvement in that area.

Did you notice any increased attention from Dan Gill and his team starting in 2022 when they shifted their focus to productivity? Did you find that the product team was focusing more on what you needed to be productive, or was that not the case?

That wasn't the case. I've had a lot of interaction with Dan Gill, who is one of my favorite people at Carvana. He's always positive and forward-thinking. He's done a lot for Carvana, but there wasn't much focus on how we could advance the technology in the ICs. Most of the attention was on vetting third-party software for tracking reconditioning. In my opinion, there's not much out there that's effective. The best one is CarMax's, but it's not for sale. Carvana did invest in a team to build their own software program, the Carly system. That had the most impact on the inspection centers and was a significant step forward. However, as the programmers often say, if you don't succeed at first, call it version 1.0. The Carly system is probably still in that stage and has a long way to go.

Given the lack of technology, how did you measure the productivity of your stations in terms of speed and quality? How did you determine the order of tasks for a car that needed multiple things done? How did each station decide which unit to pull from its inventory next? How were these routing decisions made without a comprehensive technological system to optimally guide cars through the system and measure the performance of all the nodes?

The process is still very manual. If you've toured Blue Mound, which I may have shown you, you would have seen the writing on the windows. After inspection, the cars are routed. A porter reads the instructions on the window, gets into the car, and physically moves it based on the information from the inspection. However, there's no system in place that optimizes the process by instructing which group of cars should go where. The cars are simply stacked.

How do you measure the performance of the nodes? For instance, how do you assess the work of the paint centers or the dentless repair technicians? How do you differentiate between good and bad performance?

We mainly look at how many tasks they complete. Each group has a production board, and we also track defects.

So, you track how many tasks are completed and how many defects occur. How do you decide how many cars to start working on to maintain your tempo? You need to identify your pace to avoid inventory buildup. How do you set your tempo at the start?

Generally, we plan two weeks ahead. For example, if we want to produce 500 cars in two weeks, we aim to inspect 500 cars this week.

The challenge with that approach is that if the cars take longer to refurbish, you'll have a work-in-progress (WIP) issue in about five or six days.

That's correct. We usually have a meeting at the start and end of each shift to manage such variability. For instance, we might need to inspect 25 more cars today to build up the next zone.

So, you need to build up a little WIP if you're falling behind. It's not necessarily a one-in, one-out system, although that could be an approach.

Yes, we need to build up the WIP zones throughout the whole process. Sometimes, we might need to shut down one inspection line because we have enough cars in the overall WIP. We just need them distributed from the cosmetic stage to the next. In such cases, we can use some of the labor to move cars around and support the process in other ways. This is an ongoing, daily manual task.

Absenteeism during Covid must have been a nightmare.

Indeed, it was a nightmare.

Right.

Generally, our model was built for an average amount of people taking paid time off (PTO), but we saw absenteeism rise from 7% to between 15% and 20%. It was probably the biggest challenge at that time.

But we're still not at a point where we can accurately predict how long it will take to refurbish a car before we even buy it. This would allow us to price the cars that we can refurbish quickly differently from those that take longer, ensuring a steady flow through the IRC, for example.

I believe they're significantly closer to having enough data points to start. For instance, you could say that the average 2017 Honda Civic needs these parts and takes this long to repair, so let's have those parts ready.

Understood.

While that sounds simple, it's quite challenging. Even starting with that data in the first place, such as when the car comes in for inspection, and saying, "Make sure you look for this because it's a common problem for this year, make, and model."

Is there anything else I should be asking about?

As you mentioned, we've covered a lot. There might be something else, but nothing comes to mind at the moment.

I appreciate your time. This conversation has been helpful.