Dan Gill，Carvana 首席产品官：公开市场上最疯狂的故事 | E1243

We IPOed at about $2 billion, peaked at about $60 billion, dropped back down to $400 million, and we're back to $40 billion. The fun thing about dropping by 99% is that the difference between a 98% drop and a 99% drop is another 50% drop.

Dan, I am so freaking excited for this, dude. I love the Carvana business model, so thank you so much for joining me today.

It’s really an honor. I’ve admired what you built, and you’ve got quite a record of impressive guests, so I’m honored to be here.

1. How Did Gymnastics Shape The Approach to Business?

Flattery will get you everywhere. I do agree it’s been a fun journey, though. I heard from a mutual friend that you were a US gymnast. That’s a rarity on the show, so can you start with how your time as an athlete impacted your approach to growing companies?

First, gymnastics influenced me in every way. I’m incredibly fortunate that I found it. I’m not a large human, and so I probably would have been pretty mediocre at almost any other sport, but I am a very intense person and a hardwired competitor. At least I had an outlet in gymnastics.

The shortest version is that I did it my whole life. I did it through college, I got to compete for the US, as you mentioned, and I took a shot at the Olympics in 2004. Leading up to the Olympics, I completely shredded both of my shoulders—ruined them. I decided to stop seeing doctors so that I could take my crack at the Olympic Games.

I went to try and make the Olympics in 2004. I was an underdog to make the team, but on day 1 I had the competition of my life. I did high school in Northern Virginia, and so the Washington Post said, “Local kid might make the Olympics,” and it was this big thing.

Then, on day 2, I fell on my best event, which you cannot do, and my Olympic dreams were over. I went back home, and the doctors said, “You can’t do gymnastics anymore. You’ve done irreparable damage, and you just would never be able to do it again.” It was just taken from me. I had planned to continue competing for a while, but it ended up being such a blessing in disguise.

I had surgery straight away and immediately started looking for jobs. The full circle of gymnastics is that it gave me a work ethic and perseverance. Those are the attributes that I seek out in others and try to surround myself with when we’re going to pursue enormous goals together.

Do you think young people today want to work hard enough?

No. I would say it’s not for everyone. I also think if you want to be exceptional, there is no other route. Exceptional outcomes require exceptional effort, period.

2. What’s the Ideal Background for a Product Role?

I try to instill that in my daughters, certainly, and look for it in terms of who we’re going to add to our teams. But I don’t think it comes for free or is the default.

I’m just throwing shit out there. It’s Friday; I want to have fun today.

You said something about adding people to teams. A lot of people go into product from consulting. How do you feel about the right background going into product? Do you like consulting? Do you not? What’s optimal?

We’ve had great success with a small number of consultants who have joined, but you’ve got to look for that very specific type. It is the person who comes from consulting and says, “I hate it. We do all the research, make these recommendations, and then hand it off to someone else to either do it or not do it, and I hate it.”

If you get the type of consultant who says, “I know everything, and now I should make all the decisions,” absolutely not. You need people who have been put through that meat grinder. Consulting can be very hard work and have very high expectations of output, but you need the people who hate it because they don’t get their hands on the finished product and don’t get to see it all the way through.

3. Hiring Process

We’ve had incredible success with that archetype, but not the type that thinks, “Now that I’ve seen the way a dozen companies operate, I’ll tell you how to operate yours,” because operating is hands-on business.

How important is it to have people in product who have delivered products to that customer segment before? Say you’re doing consumer social and they’ve done consumer social before, versus B2B. How important is that past domain expertise in product?

I would say it can be valuable, but it is not what we index on. When I’m hiring, I use 2 attributes, and I’ve tried to pass this through the organization. The 2 I look for are horsepower and give a shit. That’s what we look for in the hiring process.

The fun thing about Carvana is that there’s basically not a single person at Carvana who’s done their job before, certainly not at the scale that we do it. We’ve taught people about the automotive industry and put fresh eyes on a lot of assumptions about the automotive industry that are really long-held. Being able to question those assumptions is important.

How do you test horsepower in a hiring process?

4. Is Being a CPO Hard When the Product Isn’t Software?

I often ask what their favorite piece of technology is. If you’re going to say what your favorite app or favorite piece of hardware is, you better have gone into the advanced settings and know what’s buried under there. You should be able to compare and contrast it relative to competitors, explain why you use that product, say what you thought when you used the competitive product, describe the differences and why they made those decisions, and explain what you would do differently if the CEO handed you the keys to the product.

Those sorts of things can really assess horsepower. Give a shit, but my favorite question is to ask, “What is the hardest you’ve ever worked in your life?” I’m not necessarily asking about work or your job. It could be that in high school you had a sick parent, were trying to be on the track team, and were trying to get into school. It doesn’t matter. It’s about demonstrating that you’ve gone through something incredibly difficult and pushed through.

I want to hear that you have that intrinsic drive.

5. Was Founding Harder Than Being Carvana’s CPO?

Let’s fucking go. When did you work your hardest? Tell me—when did you have that moment where you thought, “That was hard”?

I founded and built a startup for 7 years. Was that harder work than being CPO of Carvana? It was quite different.

It is harder, I think. The CEO seat is the loneliest seat there is. There’s the adage that shit rolls downhill, and I think most CEOs know that the most toxic stuff rolls all the way to the top.

By definition, it means that all these incredibly capable people whom you trust to do so many things for your business—people who are really exceptional and whom you’ve chosen to work with—have decided they can’t make the decision and need you to do it. The CEO seat is the worst of the worst.

I didn’t have a single moment for 7 years when I didn’t feel the crushing pressure of making sure that all these people who had chosen to work at my startup were going to stay employed, have a great outcome, and that we were all going to win together. It’s a crushing experience.

I think we have a lot of former founders on the Carvana leadership team and throughout the business, so we bring a lot of that founder mindset and accountability. But it’s different.

Everyone says that being a PM is like being CEO of the product. Do you like that analogy?

I don’t. I think that being the PM means you need to be obsessed with every aspect of the experience, and you need to intimately understand the system.

What I mean by that is that any good experience has so many inputs and so many outputs. You should have a fingertip sense of the relationship between all of those inputs and outputs. If you’re going to make an improvement to one part of the funnel, you should have an intuitive understanding of what you think will happen as a result of that, and you should know how to prioritize as a result.

But the CEO saying, “I own a P&L,” is much less important than saying, “What is my total payroll? What are the payroll taxes associated with that? What are the total expenses?” There are a lot of things that you need to do as a CEO that I certainly don’t want product leaders to have to think too much about.

I think they need to be obsessed with the customer experience and with the metrics that they’re moving.

6. Top Lessons on Product Prioritization

Can I ask you about prioritization first? There are a couple of hotspot words that I heard there that I want to dive into. You mentioned prioritization. What are your biggest lessons on product prioritization that you’ve had over the last 10 years?

Carvana has been completely obsessed with unit economics from the beginning.

And I don’t mean that rudely, but there are some marketplaces and companies where you have to be elastic to unit-economic change. It’s okay to be negative for years if, at scale, it becomes really positive. Roku is a good example. A neobank is another.

I think you’re 100% correct, and that is our experience. But the Excel model that our CFO, Mark Jenkins, built 10 years ago has been quite prescient.

When we said this is where our long-term margins will come from, this is where the economies of scale will come from, and these are all the areas that we want to vertically integrate so that we have massive unit-economic advantages, that was laid out intentionally and in a very specific order. Some of that takes a long time to come to fruition, but it’s been built very intentionally.

7. How Carvana Achieves Margin Improvement

Can you help me understand how the margin improvement in Carvana happens? Is it that you get discounts on bulk buys of cars? Is it that you get better at transaction margin? Can you help me understand how that margin profile changes from really challenging to really good?

Absolutely. A lot of it is about vertical integration. When I explain Carvana to new employees or investors, I use 1 super-zoomed-out slide to start with and say, “Look, we sell a commodity product. If you can sell a commodity product and capture more profit than anyone else can—and I don’t mean raise prices; I mean capture more of the profit pools surrounding the transaction—then you have an advantage.”

In automotive, that means financing, insurance, accepting trade-ins, and being able to monetize those effectively. There are a lot of hands in the cookie jar of any given automotive transaction.

If you can grab more of the profit pools associated with that transaction and spend less on a variable-expense basis, you have a real advantage. At a dealership, that means the sales guy, the sales manager, the finance person, the finance manager, and all of those sorts of things. If you can drive those variable expenses down, that also means driving down the costs associated with third parties.

You’re inheriting their cost structures and profit margins anytime you use third parties. If you can do more of the work yourself as a first party and drive those expenses down, that transaction is worth more to you than it is to anyone else. You’re generating more of the profit pools, spending less, and having a greater profit per transaction that you can give back to customers in the form of lower prices, repeat-purchase incentives, and other ways that make you very, very difficult to compete against over time.

That was the vision: capture more of these profit pools through vertical integration, reduce the variable expenses aggressively, and now we have a massive unit-economic advantage.

Some of those expansionary products that you mentioned—from financing to insurance to trade-ins—will be more impactful than others. What was the most impactful from a transaction perspective?

Financing is a huge one. In automotive retail, about 90% of transactions are financed. Because the industry is so fragmented, financing is basically outsourced to lenders.

Dealerships generally get a lead-generation fee of about 1.5% of the amount financed, but the lender is making closer to 10% of the amount financed because these are multiyear loans and they’re collecting interest throughout, net of all their expenses. On a $25,000 car, that’s about $2,000 of profit per car in spread.

We built ourselves into a full-spectrum lender at Carvana. We do our own proprietary credit scoring, loan structuring, decisioning, and underwriting.

8. How Dan Makes Tough Product Decisions?

Talk to me about that product decision-making, because that is a very tough product decision to make. You said that all very quickly and simply, but it is not easy to do an underwriting engine. To build your own book, you’ve got to build a lending book behind you for that. It’s much easier to say, “We’re going to get origination fees and be a lead gen.”

It is, and you’re 100% right. But we saw that as $2,000 of incremental profit spread relative to anywhere else that someone can buy this commodity product. There was enough of a “there” there.

You’re 100% right that it involved originating loan pools and being paid zero for them for several years while you’re demonstrating to the market that you’re good at doing it. Ultimately, there was enough of a “there” there that, if we were rigorous enough in saying, “We not only have to get people to take a loan from us, but they have to be performant loans,” then we could make it work.

We were obsessed with that. One of the amazing things to watch was that we built that experience from day 0 knowing that we couldn’t monetize the loans, but we needed to start originating the pool.

Even from the beginning, we had 60% attach of Carvana’s financing. We did more than 10,000 combinations of down payment, monthly payment, APR, and loan term on a per-customer basis, precalculated everything, and made it transparent and intuitive. We had higher attach rates than we ever thought we would get from day 1, and then we just iterated on that.

Is simple always better in product? You said transparency there. I’m thinking simple UIs and incredibly minimalist design. Is simple in product always better in general?

Yes.

We’re going to edit out the “in general.” I love it. So many people do this. They’re like, “Well, it’s two-sided,” but, yes, simpler is better.

As product leaders, we spend a lot of time sanding the friction off experiences, and that’s a good intuition. I know that at so many points in Carvana’s life, we’ve had instances of trying to get a little too cute in terms of how much customization we offer to people versus just giving really thoughtful defaults.

On the other side, sometimes you pick your battles and need to take a really big swing. I think the best example for us was that, from day 1, we did 360-degree photography of every single vehicle on the site. You could spin the car around, fly in through the window, and do a tour of the interior.

It was a very complex and new experience for customers, but it was also fundamental in establishing Carvana as the future of car buying. We wanted it to feel different and engender this level of trust.

9. Is Different Always Better in Product Design?

To what extent do you think different is better in product? I always oscillate between “Customers like what they know, so just give them what they know,” versus the fact that sometimes a new product paradigm is amazing and differentiated and leads to word of mouth. How do you think about that?

Generally speaking, don’t make me think as a consumer. You need to feel comfortable, especially in something that’s not habit-forming. People learn to use Snapchat over time, and I found Snapchat utterly unintuitive at the beginning, but it was habit-forming and they did teach people a new paradigm that was ultimately different.

If I’m trying to find something on an e-commerce site, being able to search and filter, see results, and click through—you don’t need to fully reinvent that paradigm.

But, for example, on how we want to sell a car, you’re not going to get to test-drive the vehicle. You’re not at a lot, going and sitting in every single vehicle. We need to bring you as close as humanly possible to every single detail.

We want to show you if there are cosmetic things that we chose not to fix, and we’re going to take a high-resolution video and let you look at it so that we set really, really accurate expectations. That was completely different from anything that had ever been done in automotive retail because we had to solve a very specific consumer problem.

You mentioned the multiple different products there. I’m really interested to hear how you think about a review, so to speak, of what you’ve done. What did you do that you wish you hadn’t done, with the benefit of hindsight?

I would say my biggest regret is more of a team-structuring regret than a specific product decision.

For the first many years at Carvana, I said that my job was to build a machine that could parallel-process the needs of the business. There is a lot of truth in that. You need to build capabilities as primitives that can be snapped together, and for a vertically integrated business like ours, we do need to do many things at once.

But I took it too far. At one point, we had 90 very small teams working in parallel. That also means 90 different prioritization cues. Ultimately, there are not 90 number-one priorities. The fifth priority of a certain team might actually dominate the number-one priority of many other teams, but they’re not going to get to it because it’s fifth on their queue.

What we did in 2022 is go from 90 back to 8. We forced much more cross-functional prioritization, asked resources to be more flexible between teams, and asked people to jump into different projects and reprioritize more dynamically.

This ended up being huge in putting more effort and speed behind the most important priorities in the business.

If I’m an early-stage founder, what should I learn from that? Don’t allow for the silofication or fragmentation of teams? What’s the takeaway?

I think that’s exactly right. It is about being incredibly disciplined and rigorous about your prioritization. There’s such a temptation to want to do all of the things at once.

The product advice I give more than any other piece of advice is: if you can only change 1 thing, and you have to hold all other things exactly as they are right now, what is the 1 thing that you want to change?

You have to understand the complex system. We can all look at any product or piece of technology and think of so many things that could be better, but you have to say, “If I take this 1 thing that I want to change and make it perfect, what do I think is going to happen? How much lift am I going to get in conversion or, ultimately, in the unit economics of that business?”

You can’t change everything all at once. You’re too diffuse with your energy and effort. You have to force yourself to say, “If I make this thing that I want to change absolutely the best I can possibly imagine it being, what’s going to happen inside of this complex system?”

Then really focus your energy on the 1 thing that you have the most conviction will have the largest impact. You’ve got to serialize more, as opposed to trying to parallel-process.

10. How To Choose the Right North Star Metric?

I’m a startup founder, and you’re an angel in me. Thank you for your belief, by the way.

Yes, of course.

Your money’s just going up in flames if you invest in me as a startup founder. My question to you is: how do I know which 1 thing to choose? What is the advice on the right North Star metric and how to know you’ve got it right? Because you can walk down a path of many “1 things.”

Choosing the 1 thing doesn’t mean that there is only 1 thing. It’s about putting them in order.

What I like to ask founders when I’m working with them or investing in startups is: how does your moat get wider and deeper as you build? I want to hear how you’re sequencing those choices.

Once we’ve got the incremental gross profit of being able to originate these loans, and we’re making far more gross profit than anybody else, we’re going to plow that back into market expansion. As we do that, we’ll have a broader selection to choose from. That’s going to increase conversion of people who are coming to the site.

As we then increase conversion, we can hold more inventory. That’s going to get the inventory closer to the customer, which drives delivery times down, and that’s going to increase conversion even further. I can add more inventory, and these things build on one another.

It is a flywheel—thank you, Jeff Bezos. I want to hear how you’re going to sequence many “1 things.” I just don’t want to hear that you’re doing them all at once.

How do you think about staging product expansion? That’s the other hard part. I totally get you not doing it all at once, but you have the opportunity to create a product suite, as you’ve done brilliantly with Carvana. When is the right time to add another product?

Working in enterprise software early in my career, I was a Geoffrey Moore acolyte. You’ve got Crossing the Chasm and other books to rely on.

That was one of my favorites. I read it when I was 14 or 15 at boarding school. I was single for all of my teenage years, so, shocking loser.

I think that what we like to hear is: how does that particular effort accelerate the spin of the flywheel?

Let’s use Carvana as an example. We could license so many pieces of software that we’ve built. There are a lot of companies in our industry that would love to use software that we’ve built. The idea of us generating $10 million of incremental SaaS revenue or $100 million of incremental SaaS revenue doesn’t actually make the core business better.

That is contributing to the core flywheel, as opposed to building an ancillary product. We like to see the ways in which these product offerings accelerate one another and make the core offering better and better and better.

11. Carvana’s Biggest Product Decision Mistake

It has to go back to the core offering and add value there. What product decision did you make that was the biggest mistake in Carvana’s history?

It’s a fun one and somewhat counterintuitive. One thing that we did early on is that we wanted to offer free shipping for everyone.

We’re an e-commerce business, and we want to give you a broad selection. Our cars are in these 100-acre facilities around the United States, with 5,000 or 6,000 cars on the ground. We have a nationwide logistics network that we own and operate, and we wanted every car in the network to be free for every customer in the network because the truism of e-commerce is that you must have free shipping.

Consumers love free shipping, so you think, “That seems reasonable.” But in our case, the costs are very high. We can’t rely on FedEx, UPS, or USPS to deliver a car to you.

How much does it cost to ship a car?

It depends, but we were shipping cars upward of 1,500 miles.

Right.

If you’re going to move a car from Philadelphia to Fort Worth, Texas, and while it’s en route the customer says, “Sorry, I bought another one. I used the free option you gave me for the one you were sending, but I bought a different car,” that’s very expensive for us.

Now we’ve got to restock that car and find another customer. Meanwhile, these are depreciating assets. They’re losing value with every passing day.

What we came to over time is that we can’t allow for that, and we need to charge for shipping when you’re moving a car from a long distance. We want to offer free shipping if we have a car that’s close to you. It’s still a wonderful value proposition to say, “This is free delivery to your door,” and we want to offer that.

But if we’re going to move a car a long way, we added a nonrefundable shipping fee to create a little bit of friction at that moment. We’re going to start moving this thing, but it’s going to cost you money, and it’s nonrefundable.

What ended up happening is that we unclogged our network from wasted excess transport. That brought delivery times down for everyone, gave us fewer late deliveries from this sort of excess clogging of the network, and we sold more cars.

By taking away free shipping, we actually sold more inventory.

What’s your primary customer-acquisition channel today, Dan?

Like many companies, we’ve come around to the fact that brand advertising is incredibly powerful. We want Carvana to mean something as a brand to our customers, and so we spend a lot of time storytelling.

12. Top Lessons in Product Storytelling & Marketing

What are your biggest lessons or pieces of advice when it comes to effective product storytelling today and product marketing for founders listening?

I think that people like to conflate storytelling with sales, and they like to shit on sales. They say, “I don’t want to be in sales.” The truth of the matter is that you’re in sales, and you better be in sales.

You better be in sales to attract incredible talent, but you should also be storytelling because it’s motivating and inspiring for your teams to know what they’re chasing after.

Let’s contrast that with not having a more fulsome story. If you look at a particular product and say, “We’ve got to make this product massively better,” an engineering mindset may say, “I see the thing that’s broken, and I’m going to fix the thing that’s broken because nothing else matters until we get better at this immediate next step.”

But you’re actually trying to get many people to collaborate and work on this thing. Telling them, “No, we’re trying to get way over there, and when we get all the way there, this is how much better the unit economics are going to be. We can give a better customer experience and a totally differentiated offering, which is going to help us carry more share of market,” makes people more motivated and excited to go chase after that.

You need to be able to paint a high-resolution picture of what that future looks like, so that when people get completed with what you’ve asked them to do, they understand the why and can keep going autonomously and independently.

Being able to paint pictures of the future is absolutely requisite, and it’s something that you should work really hard on.

What’s a good picture of the future, and what’s a bad picture of the future? I hate “Zoom: Make everyone happy.” Who’s going to say, “No, I disagree. I want sadness instilled upon everyone”?

TwentyVC is very explicit: we are not for beginners. If you don’t get what we talk about, no worries—this is not the show for you. We are for or against a brand, and that is great: Nike and Adidas, McDonald’s and Burger King. What is good? What is not good?

Exactly as you said, be intellectually honest and be transparent about what you are and what you’re not.

For Carvana, we have always said that we intend to build both the largest and the most profitable automotive retailer ever built. We’re unapologetic about the fact that we are building a highly profitable business, and we’re going to do that by offering the best experience, the broadest selection, and the best value for consumers in a way that allows us to capture market share that no one’s ever captured before.

13. The Best Experience Hurt Margins the Most

Where does the best experience hurt margin most?

It certainly does at subscale. Trying to move cars across the country involves so much underutilization of assets. These are big 9-car haulers going down the freeway, and if they’ve got 3 cars on them, it’s not a very good use of expenses.

But you have to be willing to move those trucks and get cars to customers. If you say, “I’m going to wait until the 9-car hauler is full,” then I can only tell you that I’ll get you your car somewhere between 3 and 11 days from now. That’s not a good customer offering, it’s going to take a long time, and it’s low-converting.

You have to be willing to absorb some of that underutilization for a period of time. But once you get to the point where you’re fulfillment-constrained instead of demand-constrained, there’s so much incredible optimization work to be done.

We’ve done a lot of software building around intelligent labor utilization, logistics-network utilization, and fulfillment utilization to make the unit economics really sing. That takes time.

Do you know the lesson I have from interviewing 30 or 40 of the best CPOs in the world?

Lay it down.

I’ll tell you the lesson: the best are not product people, weirdly, or they don’t sound like product people. They sound like intense business leaders who understand unit economics and product better than anyone else. The good just get product, and the bad don’t get either.

The combination of getting business, unit economics, and product is what makes the best. Do you agree, or is that a stilted view of what a great CPO is?

I completely agree. You have to be unapologetic about the fact that you’re building a business.

We say we do not win by shipping features; we win by moving metrics. Ultimately, while there are so many input metrics and you should have a clearly defined, unit-economics-driven hypothesis for why we’re taking on any given initiative, and you should have leading indicators that you’re measuring from minute 0 of launching a new initiative and looking for evidence that your hypothesis is valid or invalid, if it doesn’t drop to the bottom line, you have not contributed anything.

It has to ultimately show up inside of the business. We are unapologetic about that. It is everyone’s responsibility to care about it: product, engineering, analytics, design, and operations.

It is everyone’s responsibility to care about the initiatives that we’ve taken on and what we’re prioritizing, and to see that all the way through to economic impact for the business. So, yes, I think it’s absolutely the case that everyone should be rigorous there.

I’m going to ask a rude question.

Do it.

Is it difficult being a CPO in a company where the product is almost not the software product? What I mean by that is that delivery time is really important, letting me know where it is and when it’s coming, and then ultimately it’s on my drive and it’s as I thought it would be.

That is really the product. The pixels that I discover it with are important, but nowhere near as important as the other 2. Is that difficult being CPO in a world like that?

Not at all, because the technology is what unlocks all of those capabilities.

If you really want to get a car 1,000 miles away as fast as possible and profitably, there is so much technology to be built to do that. You have to have all of these vertically integrated systems. You cannot have humans in the loop for every single decision.

Our articulation of that, when I would hire engineers early on, was that there was all this thing about certain companies in Silicon Valley being engineering-led. I would say, “We are not engineering-led. We are customer-led.”

You need to be completely on board with the fact that we’re trying to unlock more and more value for customers, and by doing so we’re going to win a very big game by a very large margin.

Is everyone not customer-led? Ultimately, I certainly hope so. But this notion of being engineering-led can sometimes be trying to romanticize and give too much control—or a perception of control—to the idea that it’s really all about the technology, that if you build it, they will come.

I don’t think that’s singularly true. You need to have a go-to-market muscle, tight feedback loops, analytical rigor and obsession. There are emotional aspects to using any product that really matter, along with psychographics and all of these other things.

You do have to actually be customer-obsessed, and I don’t think that every organization evangelizes it in the same way.

You mentioned tight feedback loops. Every product team has product reviews. Bluntly, some are done brilliantly and some are done terribly. What have been your biggest lessons in how to do effective product reviews?

You can’t hide from the metrics. Several years ago, we took on an operating cadence where there is at least 1 executive-team member, sometimes more than 1, who meets with every single team in the company every single week.

We review what you said you were going to do last week, what actually got done, what you’re going to do this week, and how we’re seeing that in the results.

As mentioned, you have to have a hypothesis for why we’re taking anything on, and those hypotheses have leading indicators of whether they’re true or not.

What do you do when there’s a team that consistently has a chasm between what they said they were going to do and what they actually do?

Reconfigure the team.

Generally speaking, we like to have single-threaded leaders who are highly accountable for those initiatives, the prioritization, and the execution of those initiatives. We look to that leader and really try to understand their diagnosis for why it’s working or why it would not be working.

It is that leader’s responsibility to ensure that it is working or to change course.

To what extent do you listen to team members on product reviews? I think, “We should do this,” and you say, “That’s great, Dan. Back in your box.”

We love the ideation coming from everywhere. We have a number of product managers in our organization who were customer advocates. They were on the phones listening to customers and dealing with customers all day, every day.

We have ones who were loan underwriters, and others who did home delivery of vehicles. They know how the experience actually works, where the tooling has been falling down, where the frictions are, and how to make those gains.

We love that coming up from the bottom, but we have a very high bar for rigor. You better have a hypothesis for why this is going to matter.

A good framework for that is the magnitude of impact times the frequency of occurrence. You can say that something is a horrendous experience, but if it happens 1 in every 50,000 deliveries, that’s not a very good thing to bring to the table.

Even something with medium pain that happens 1 in every 50 deliveries is much more important for us to pay attention to. You can distill that into a unit-economic impact. You can say, “What is this doing to our cost per unit?” or “What is this doing to our profit per unit?”

We ask everyone to collapse their ideas into that unit-economic framework so that they can be compared apples to apples.

Gustav from Spotify, who I think is one of the best CPOs in the world, always says, “Talk is cheap, so we should do more of it.” I fundamentally disagree with this. I think speed of execution is everything, and I actually think dictatorial product leaders generally work best.

You’ve got to appease people and make them feel heard, but ultimately we’re going to go with what we think is right. To what extent is he right versus me?

Ultimately, clarity of thought and clearly articulated strategy are very important.

I really agree with you on the importance of writing. Do you do an Amazon-style 7-page document? Do you do slide decks?

We do written communication. Written communication is hard to hide. You have to distill your perspective into clearly articulated prose, and you can’t hide and say, “That’s not what I meant.” You wrote it down, as opposed to if you just presented it in slides.

We really like written communication for PRDs and for proposals of major initiatives in the business. We’re also okay with rather visually heavy things.

What do you think are the most non-obvious other skills that product people, or the best product people, have? You mentioned writing, and we mentioned being business-minded. Are there any others where you look at your team and think, “The best have this less-obvious skill”?

I think high accountability is the number-one thing you have to look for, where people take it personally when things aren’t succeeding.

It’s that willingness to go the extra mile: to watch the customer-session replays and take notes, to do product teardowns of competitive products and compare and contrast, to sit and read every word of copy on the page out loud, and to feel embarrassed when it’s not intuitive.

14. How Should Founders Use Take-Home Assignments?

That willingness to take it personally and want to get it right is the number-one thing that we look for. It’s the willingness to go the extra mile and sweat the details.

A lot of founders worry about take-home assignments when they have someone amazing—an incredible talent coming from a top company—and they think, “I don’t want to give Dan a take-home assignment. He’s so important.” How do you advise founders on take-home assignments? How do you do them right in product, and what should you not do?

Let’s go the other way first and say that the single best way to hire someone into an organization is for them to reach out to you with a vision for how your organization gets better. They’re so passionate about it, and they already have ideas.

Whenever anyone asks me, “Could you connect me into any companies?”

When does that happen, Dan? I would love that. That’d be great. Thanks.

It is rare, but it happens. There are those people who, at least when they say, “Can you connect me into company X?” I say, “Why? Why do you like that company?”

It can’t be, “I heard they’re doing well,” or, “I see their stock.” It needs to be, “I love the approach they’re taking, and I want to be a part of doing this.”

Even if they don’t have that, I’ll encourage them: you need to go and create that for yourself. You need to write down why you believe in this business.

There is nothing more magical than being a founder and having someone tell you, “I love what you’re doing, and I want to be a part of it.” It’s the most intoxicating feeling because you’re so self-conscious and have so much pressure on yourself. To have someone who sees it, gets it, and wants to be a part of it is an intoxicating feeling for a founder.

Let’s start with that: when people are applying, they want to be very specific about why they’re applying. They’re not just lobbing in a résumé and hoping that you say yes.

Second, when it comes to whether you can have them do take-home assignments, absolutely. You need to.

This was something I learned very early in my career when we were selling enterprise software. I had a sales leader say, “How can you hire a rep without making them dance?” That really stuck with me. How can you possibly hire someone whom you’re going to trust to speak to your customers without hearing them pitch a product?

You have to hear that. In product, how can I trust someone’s judgment without asking them first to put their thoughts in writing, break down a problem, and show how they would approach it?

15. Should New Hires Work on Existing or Neutral Products?

How do you think about the challenge of whether to get them to work on Carvana, where you have a huge amount of asymmetric information and you’ll probably think that everything’s not great because it’s not at the level that you are after having spent years there, versus a neutral product where you start on a level playing field?

It is an asymmetric advantage, but we always have them do Carvana problems. We’ll try to make it something that is relevant to us.

We’ll stand up a dummy SQL database and say, “Query this and tell us what you would do strategically to increase attach of warranty products on a sale, and what’s that going to do to the unit economics of the transaction?”

We’ll give them a bunch of historical data, and it’ll be nonintuitive so that you have to ferret out those nonintuitive realities buried inside the data. Then you come to a conclusion on what you would change in the interface to make it better.

You’re right that it’s an asymmetric advantage that we understand it quite well, but you get to see the level of effort, critical thought, and versatility there. We aren’t hesitant.

We also don’t hire super-senior people. We like to grow our talent from within, so I wouldn’t have the issue of, “This person’s such a big deal that I can’t possibly ask them to do homework.”

Do you agree with this idea of a 10x engineer? Hire the 10x engineer and overpay them?

I think there’s no question that there are individuals in almost any functional organization who are manyfold more effective than others. I think that often has more to do with work ethic, accountability, pride, and those sorts of attributes than just raw intellect.

16. Does Every Public Company Need an AI Story?

I should be honest: we are not building AGI here. You can have singular insights that unlock something, but I do believe there are people who are better at turning big problems into small problems and executing with urgency and accountability. That can really move mountains.

You’re not building AGI; you’re delivering cars. But that is very important. You have to, I’m sure, come up with ways to tell Ernie about how you’re leveraging AI to make your margins and product better, because you’re a public company and every public company needs an AI story.

Do you agree that every public company needs an AI story? As a CPO today, how are you thinking, “How can I leverage AI to get as much juice out of this as possible?”

100%. We’ve been all in on AI for years.

Zoom out for a second and say that a computer can only give an answer to a customer or take an action for a customer if the computer has access to that information.

In automotive retail, there are so many human decisions. You’re negotiating the sticker price of the car, the finance rate, the trade-in, and all of these individual things. In our world, all of these systems are deterministic and algorithmic and have been built that way from day 0.

We can have an LLM tell you, “Harry, given your credit score and the VIN of the vehicle that you entered, you have $3,000 of positive equity on that car applied to your financing. That unlocks a 7.4% rate, and this Tesla Model Y will now be $612 a month with $1,000 down.”

That information is knowable in our world, and it is not knowable in the physical dealership world in the vast majority of cases. We lean very hard into AI because we think we’re structurally advantaged to leverage it and offer increasingly frictionless experiences to our customers.

BYD is crushing it in London. They’re everywhere now. I’m a believer in Adam Smith’s invisible hand and minimal market intervention, and then I saw the other day how the unit economics of these Chinese cars are being subsidized so heavily.

It is not a fair game when the Chinese government pays for 30% or 40% of the cost. To what extent are Chinese cars a net negative for your business, and do you worry?

For any new brand that wants to enter the United States, think about what it takes to get cars in the hands of customers: how much capital and how much time to get land, get zoned, get staffed, and start to fulfill, and then get those cars in the hands of consumers.

Today, Carvana’s first-party logistics touches 90% of US driveways. If you can get any car to either coast of the United States, we will get it to 90% of Americans within a week.

I actually think our infrastructure is unbelievably well set up to facilitate OEMs and brands getting cars into the United States and distributing them.

Would you sell Chinese cars?

I think that’s a complex geopolitical question: what are the right moves for the government to make in terms of allowing US automotive manufacturers to thrive and grow, and protecting them from this very subsidized unit-economic model that could come in?

I’m betting on Dan. Do you think Carvana should sell Chinese cars?

I think we should be selling other people’s cars, and we want to be Amazon for cars. We want to be able to distribute the cars that consumers want.

Wonderfully ambiguous. That was beautifully done, by the way. Career in politics?

I don’t think Americans want Chinese cars. You guys are very different. The sentiment against China from you is very different from the UK. It is surprising to me that the UK doesn’t care, but anyway.

I think Americans are largely that way, but I don’t think the American government is that way.

If you go to Middle America and say, “Hey, this is from China,” they’ll say, “Fuck that.”

Americans do love cheap, though. We have plenty of Chinese goods, that’s for sure.

That’s true.

What do we not see about Carvana that we should see? What does the world not understand about this business that makes you think, “That’s the really special thing”?

I think the amount of vertical integration that we’ve taken on is pretty staggering, and it’s hard to see from the outside.

17. Are Chinese Car Subsidies a Threat to Carvana’s Business?

Early on, there was a misconception of, “I can buy a car online. I go to Google, search, and see cars.” But that’s a listing site. You’re going to go to a dealership and transact that car. It’s not an e-commerce model for automotive.

When you look at the number of companies that we’ve had to build, we are a full fintech, we’re a lender, and we have one of the largest automotive reconditioning companies.

What do you think you vertically integrated that you shouldn’t have done? I still look at Shopify and Stripe. Stripe’s partnership is incredible, but why Shopify is never a vertically integrated Stripe, I have no idea.

We pulled back on a couple of bets. Even bets that I mentioned today—we have examples where we said, “This is such a logical adjacency; let’s take it on,” and then we pulled it back.

Why did you pull it back, out of interest?

That was during the 99% drawdown of the stock, when we needed to say, “It’s time to focus on the core, make sure that the unit-economic model is crystal clear to the world, and get on better footing.”

That was part of the refocusing.

Dan, the market cap today is $50 billion.

The market cap was $5 billion. We IPOed at about $2 billion, peaked at about $60 billion, dropped back down to $500 million, and we’re back to $50 billion for the second time now.

You dropped down to $500 million?

Yes.

Oh my God, dude. As a leader, what do you say to your team when it’s at $500 million?

With the benefit of hindsight, it was so healthy for us to go through that together.

18. What’s the Most Misunderstood Thing About Carvana?

What you say to your team is, “Here’s what the rest of the world misunderstands about the truth. What you’re reading in the Wall Street Journal or any of these publications is one version of the truth. Let’s be really clear about what that version of the truth is, and then let’s articulate to you our version of the truth. We’re going to check back in in 30 days and tell you which one is more true.”

In 30 days, did we do exactly what we said we were going to do? Is everything that we told you checked out? Yes, it has. All right, let’s roll that forward another 30 days.

We were doing an enormous amount of communication through that: here’s exactly where we’re going, here’s how to express that in unit-economic terms, always our obsession. This is what we’re going to do in terms of cost per unit. This is what we’re going to do in terms of gross profit per unit, and this is how we’re going to see through this storm.

We shared those results again and again and again. There were the people who opted out immediately at the beginning of that drop and said, “This isn’t for me. This thing’s never going to make it,” and, “I always knew it was never going to make it.” There’s that crew.

Good riddance. Please move along. Some people just couldn’t handle the swings, and that’s totally understandable.

But for those who chose to stay, it has been so galvanizing as an organization to go through that process together, rebuild that trust, and say, “Wow, you were right. We were misunderstood from the outside, and we did do exactly what we said we were going to do.”

Now we can ask, “What are we going to do next?” Everybody’s charged up and ready to run at that hill.

With the benefit of hindsight and having survived, I think we’re in better shape than we’ve ever been as an organization and really poised for an exciting future.

19. Keeping Morale High After a 99% Value Loss?

$500 million. That’s quite a drop. I remember someone—we mentioned our friend earlier—showing it to me at around $3 billion or $4 billion, I think it was, and I thought, “Seems rich. It could go low.” Then I thought, “Oh, that’s like a nice 100-bagger.” Jesus.

What’s amazing is that it’s a 100-bagger in public markets.

Yes. It’s been an incredible journey, with just incredible, wonderful people who really care.

We’re an organization based in Phoenix, Arizona, and there’s a little chip on our shoulder that we’re not Silicon Valley or whatever. But the truth of the matter is that I would pit our team against any team in the world.

Why has being in Arizona helped you over being in Silicon Valley?

You have to opt into the mission. We kept a very high standard around things like grit, accountability, problem-solving, and bias to action.

Once we found the right people, they were fully bought in. If you have those characteristics, you can be employed elsewhere. You could choose to go many different places, but you’re in Phoenix and you’ve chosen to be with Carvana.

We’ve really built a long-tenured group. Ninety percent of our senior leadership has been with Carvana since before 2019, well before the drop. Our executive team has been together for 10 years, and we’ve built real cohesion.

I’m still getting over this 100-bagger. That’s going to keep me up at night. You’ve ruined my day.

Listen, I want to do a quick-fire round. I’ll say a short statement, and you give me your immediate thoughts. Does that sound okay?

I’m into it.

What’s the most common reason early companies do not get product-market fit?

I think founders are often in such a hurry to found something that they aren’t rational enough about what they pick to go and found.

There are a lot of companies out there that are features and not products. You better have a very clear vision for how that feature is a wedge to something much, much bigger or something indispensable for an organization—not, “We’re just going to build it and wait and see.”

I’m obsessive about this notion of how your moat gets deeper and wider as you execute. I think a lot of founders lack that vision.

What do you know now that you wish you’d known when you got into product?

That it’s at least 50%, if not more, about people and not just making product decisions.

Fortunately, I love people. I’m a prosocial guy, and I love my teams. But the real leverage is in getting lots of people to move with urgency, care, understand the mission and vision, and attack it.

That’s a lot of what it’s about. It’s not just making product decisions and saying, “Let’s go do this.” You’ve got to get a lot of people to run hard along with you.

What’s the most recent wow moment you’ve had with a consumer product, and why that one?

I have to give you 2. One isn’t crazy recent, but it’s fun. It was the first time that my wife tried Google Maps inside virtual reality. She was tromping around our living room like Godzilla, looking in our apartment window, and experiencing this immersive experience.

I’m still a big believer in what’s going on in the metaverse, and I’m very excited by where that all can go.

The second one was literally 2 days ago. I got a Rivian, and I’m obsessed with it. There’s just so much craft in every detail of that car.

The headrests have a plaid inlay in the middle of them. When you’re setting the flow direction of the air vents, there’s this beautiful animation, and you can save it as your default. When I get in the car, the air vents point in a specific direction, and when my wife gets in the car, they point in a different direction.

There’s just a lot of craft in that car, and I love it. It’s beautiful.

God, we really do live in hard times, don’t we? When I get in the car, the air is in this direction. Goddamn, life is tough.

I wouldn’t want to live in any other period of history. I’m happy to be here. Do you know what I find funny, and it’s going to ruin the whole interview for you? I don’t even have a license. I’ve never driven a car. I don’t get it.

20. Quick-Fire Round

You don’t have a license? That’s wild. I hope my daughters never have to drive. I think they’ll be self-driving.

I live in London, literally 100 yards away from my office, and I take Uber. That sounds awful, but I’m a total PR-man diva, so everyone comes to meet me anyway.

Jaguar’s rebrand: terrible decision that will hurt the company, or weirdly genius move to get attention?

I’m not a fan of the rebrand, if I’m honest. I think Jaguar already means something.

When you’ve built brand equity for so long, and the brand has stood for quality, innovation, luxury, and all these really great things, I think you could have reinvented the product and really let the product speak. You could tell the story of why the product is different and better, as opposed to trying to reinvent the brand, which already had plenty of value.

I don’t disagree with you. I understand that they want to appeal to a new demographic, specifically in Asia, but the heritage element is what Asia loves.

What 1 piece of advice would you give to a product leader starting a new role today?

Ultimately, you do need to understand the whole system that you’re stepping into. Product is really about having a systems mindset.

You have to understand all of the inputs and outputs and where the leverage is. I’ve mentioned this a little bit, but my number-one piece of product advice is always: if you can only change 1 thing and everything else has to stay constant, what is the 1 thing you’re going to change?

Imagine that future where that 1 change is as perfect as you can make it. How does that impact the business KPIs that you’re responsible for? That’s what I always tell product leaders.

What other function has the most tension with product?

We’ve fought this hard. Carvana is such an operationally intensive business, but we have to be hand in glove between technology and operations.

I think sometimes operations has a reputation for wanting a Band-Aid for a pain that exists today, and sometimes product thinks they’re smart and knows more. “Don’t worry; we’ll handle it.”

But you really need product to live in the weeds with operations and intimately understand the challenges, experiences, frictions, and deficiencies. You want those folks to be very close together and embedded.

Product and operations can sometimes be in tension, but we work really hard to slam them together and make them best friends.

What about the way we build product today will be completely different in 10 years?

The siloization of functional organizations is going away. AI is reimagining what it means to write code, do analytics, and even do design.

The most versatile problem-solvers and thinkers are just going to be able to cover more ground than they’re able to today.

Will Carvana have more or fewer people in 10 years?

It’s an interesting question because there are aspects of our business that are very people-centric, and I think we will want them to be people-centric forever.

The person who shows up on your driveway and delivers your car—even if that car were a self-driving car—I think there’s still that customer service and the face of our brand. That’s going to be really important, and it scales slightly sublinearly in terms of labor optimization but is going to need to scale with unit sales.

On the other hand, we’re very transparent that I do not think that to sell twice as many cars, we need twice as many product managers or twice as many software developers. We should be expecting incredible leverage over our fixed-expense base in those areas and expecting those people to cover more ground over time.

How will AI impact your customer service, specifically support teams?

Ultimately, we believe that people don’t want loans, trade-in values, or loan underwriting. They want their car. They came to get a car.

We need to remove as much friction as humanly possible while also giving the customer confidence along the way that they’re making a good decision. We want to automate everything and make everything self-service and intuitive.

If we can hit an API to get data on you to help our underwriting decision instead of making you provide a document to us, we’re going to do that. But if at any moment you’re not feeling confident or you’re nervous, we want you to be able to pick up the phone and talk to someone who’s knowledgeable, confident, and competent.

We want to give you that option.

When I say CPO, in support today, are we hiring more support people or replacing them with AI?

It’s both.

We want to have the best people, retain them for longer than you would see from a typical customer-service agent, because they’re exceptional. When you have a business as complex as ours, you need to specialize.

You have people who specialize in the early part of the transaction, people who do the loan underwriting, people who do the registration work, people who do the title work, and people who talk about the delivery because it’s so complicated.

As we make those systems more integrated and intuitive, and AI can handle all the low-hanging fruit, I want the person who answers the phone to be able to traverse any of those systems and say, “Here’s exactly what’s going on with your car. I’ve got you covered. I’m going to follow up with you in 2 hours when I have more information, and we’re going to be good to go.”

We want the caliber of those people to be just off the charts, and we want AI to handle as much of the low-hanging fruit as possible.

Final one. As a CPO analyzing other companies around you, which company’s product strategy have you been most impressed by?

This is a popular one, I suppose, but I love the go-to-market of SpaceX.

First, if you’re going to drop the cost of getting to space by an order of magnitude, I have nothing but respect for the operational discipline required to do that. But simultaneously, you’re going to use that advantage to build out Starlink, which is likely to be an insane cash-cow business that will allow you to continue to invest.

It’s brilliant and really to be admired because it takes such incredible vision, discipline, and execution to do it. That’s what I love the most: disciplined execution and operators.

I literally tweeted today that I’ve been in private markets for 10 years and I’ve never seen institutional demand so concentrated toward 1 company’s stock as it is now. Before, it was always, “I want Figma,” or, “I want Wiz,” or, “I want Notion.” Now everyone just wants SpaceX.

My question to you is: if you enter SpaceX at a $360 billion valuation, there are multitrillion-dollar companies, but you also have to grapple with the fact that SpaceX’s mission is to make us multiplanetary. The payoff timeline of terraforming Mars feels very long.

You have to buy into this notion that it’s about becoming a multiplanetary species, as opposed to just building out Starlink. Can Starlink alone be a $2 trillion business or whatever?

I don’t think about it as much from the perspective of its current valuation. I think about it as, “I love the way this business is constructed and how clever it is.”

It’s able to take on such an enormous swing while also building something so profitable and useful.

Last one, I promise. OpenAI at $160 billion, Anthropic at $40 billion, or xAI at $50 billion—which one do you buy?

I wouldn’t invest in any of them.

I think the foundational models are trending toward commoditization, and it’s going to be difficult to capture all of the gains relative to what Google, Microsoft, and others can capture.

Do we not agree, though, that cloud is commoditized and, despite that, AWS is still very valuable, Google Cloud is still very valuable, and Azure is very valuable? Is there a difference between those comparisons?

No. I think those things are commoditized, and they add value-added features on top of them.

I think that what you’ve quoted are companies that are already valued at $50 billion to $160 billion apiece. There’s an enormously long way to go to accrue the kinds of revenues and margins that will justify significant upside on top of that.

You also have to believe that they’re not going to be accrued by the cloud providers bringing these foundational models into existing distribution channels.

I wish we had a dinner party. It would be much more fun.

Listen, dude, thank you so much for joining me. You’ve got such a fun energy, and you’re great at what you do. It’s been really fun.