Learnings From Starting & Scaling Convoy As A B2B Marketplace With Dan Lewis | EM Group Chat #194

Welcome back to Everything Marketplaces, where we talk with founders and leaders from some of today's top marketplaces. So this is episode 194, which is a really incredible group chat that I just had with Dan Lewis. Dan previously started Convoy, which has commonly been referred to as the Uber for trucking, and raised over $1 billion in capital. They were early pioneers in the freight space and ultimately faced challenges in the business that led to a technology acquisition by Flexport. Since then, Dan has joined Microsoft to lead their AI and Copilot for Business initiatives. So this is a great chat with Dan. We're going back to the founding story and first steps to starting Convoy. How they solved the cold start problem. We got to learn how Convoy evolved as it scaled. Dan shared some of the biggest challenges with it as a B2B, more operationally complex marketplace, some of his key learnings. We spent time discussing AI, and Dan shared examples of how transformative it is for marketplaces. We also had a great group Q&A. I really enjoyed this conversation. You're going to find it a great watch to the end. So Dan, as I mentioned to you right before we started recording, you've of course led the way for us all building B2B marketplaces. So I'm really excited to have you join us here today. I want to start off by saying huge thanks for taking the time to do so in advance. We have a lot that we're going to dive into with all of your incredible experience starting and scaling Convoy. But before we do, I thought it'd be great if we could start off by briefly sharing your background for those here that might not know you.

It's really nice to be here. First off, I love marketplaces, the whole notion science behind them. My background from Seattle originally, I worked in all sorts of things for a long time. I thought I was going to get into film actually in school and then ended up in management consulting. kind of the opposite, working in supply chain technology, worked in, you know, the US, Europe, Panama Canal for a while, kind of Mexico, and eventually ended up at Amazon and was really inspired by both the marketplace dynamics that existed at Amazon and also by the supply chain challenges that Amazon was facing and kind of the opportunity it was creating in the world to innovate with supply chain. I remember thinking that, you know, every company out there that's in physical distribution space is facing questions around are they going to be relevant as the supply chains change? And Amazon was leading the charge in that. So those kind of forces came together and inspired me to go start a company with Convoy.

That's an incredible background. Of course, a lot we're going to get into here, but I wanted to go back to the founding story, you know. So what was the founding story for Convoy and what were the first steps that you took to start at the time?

The founding story was, I really, I thought the business was really interesting of just, as I mentioned, transportation supply chain. And I felt like there was a lot of opportunity there because Amazon was changing the way people value delivery. So if you thought about before it, you know, the really fast delivery, the reason you would go buy something is because it was at a store near your house or because the price was right or because it was the right, you know, exact product you wanted. And then all of a sudden delivery time became a major factor in terms of whether someone bought something. If you could get it in 1 or 2 days, you'd want it. And I wrote down the supply chain is winning because I realized that if you're the fastest at delivering, if you can deliver faster than any other company, they'll procrastinate if they trust that you can get it there in a day or 2. So let's say Amazon says with Prime it's 2 days and you're a Prime customer, you're gonna say, well, I'll just wait till there's 2 days left and I can order because people naturally will kind of like fill up the space and wait. And once you're in that zone, you have no other option other than to buy on Amazon because it's the only one that can deliver that fast. So it became this local, probably too strong of a word, but almost this local monopoly for the individual person that had waited until there was 2 days left. It's like, oh, it's really powerful. People feel like they're in a position they have to buy on Amazon at this point. And so I realized that, you know, other companies would have to care a lot about fast delivery and that would become a differentiator for how retail worked. So that, that drove me to go out and look for opportunities in supply chain. And then I, we just went and talked to a lot of people, talked to tons of trucking companies, truck drivers, brokers, shippers. And after about 6 weeks of conversations, realized there was this extremely untapped opportunity around, you know, the hundreds of thousands of trucking companies in the country, the thousands of companies that ship freight. And matching those up in a more efficient way. And so that marketplace, you know, became the inspiration to come go and start Convoy.

And then as marketplaces, we of course, you know, face the chicken and the egg or cold start problem when starting out. So, you know, how did you solve for that one with Convoy?

So Convoy is a managed marketplace, first off. So we would take the spread between what we bought and sold trucking capacity for. So if a shipper or a beverage company comes along and asks us if we can provide trucking services, We would say yes, we would sell the service to them at a given price, and then we would go find a truck in the marketplace that could fulfill that job for less than we had sold it, and we made the spread. That's how the trucking industry works. So you take principal pricing risk on both sides of the trade. It's not like a GMV with a 15% take rate. It's an actual buy-sell spread. So that's why I call it kind of a fully managed marketplace where we take service delivery risk and also kind of pricing risk in, in making the marketplace happen. So we were thinking about how to get this started. We had to, you know, figure out early on, is it gonna be more supply-led or demand-led? Do we need to go get a bunch of supply and have that supply waiting for us, or do we need to go get a bunch of demand so that the supply will follow? At first, we actually thought it would be more supply-led initially, and we went out and talked to a bunch of truck drivers, and I think we probably talked to like 50 or 60 truck drivers. And none of them signed up. They were all like, nah, don't need it. Have a bunch of brokers I call all the time anyway. We're like, no, but there's this cool app and you can find all the jobs. It's super convenient, really easy. And they're like, I don't even know you. Why would I ever take work from you? You're this random broker I've never heard of before. And so we realized that we couldn't start on the supply side. And I was actually pretty worried at that point. I was like, okay, maybe nobody wants this. And so We went on the demand side and we started asking all these different shippers if they liked their experience working with trucking companies. And a lot of them didn't. They weren't happy with the transparency, the communication. They felt like they didn't know where their stuff was. They're always getting haggled over price. And so we said, okay, well, we can, here's this better opportunity. Would you work with us? We got a lot of the demand side to agree and say, yeah, we would actually want to do this. So what we did is we became really good, our secret sauce early on was being able to get a job in our system from the demand side that would pick up within 24 to 48 hours. And we would then go find a trucking company that was willing to do the job. So we used that as the, as the, every single job that came into our system was like the carrot that we would go offer to, we'd contact trucking companies, ask if they wanted the job. If they said yes, we'd say, great, it's yours. It's just waiting for you in our app. So just download the app and the job would be waiting for you. And then you can use the app to fulfill the job and do location and checking with your paperwork and everything. And so, you know, it was interesting. A lot of truck drivers would say yes, they're like, they wanted the jobs. They would actually download the app in order to get the job that was waiting for them. So the most important thing for us for scaling quickly was being able to very quickly take a unit of demand and then source supply, get the supply to fill out our paperwork, do a broker carrier agreement, give us their insurance documentation, their safety history, vet them. And so we had to build an extremely fast vetting and onboarding system for the supply side of our network. So taking in a demand, rapidly vet and onboard a new carrier to do that job, and then you match the two. So that was the cold start kind of like problem we worked on initially. We didn't, you know, have a lot of demand. So, and I remember looking at Uber and other examples back then and seeing that they were giving suppliers like $500 to drive for them. I'm like, that seems really expensive. And we tried that a couple times and drivers got kind of weirded out. They're like, why are you giving me money to, to, to take loads for you? I already do this for a living. So we stopped and we said, okay, well, the easiest way to bootstrap demand to get more supply in is we're gonna just make our own loads. So we rented warehouse space in 2 different warehouses in the Seattle area and ran like 4 pallets of, I don't know, sandbags or something like that back and forth between them. And each job would pay $150. But that allowed us every time we did one of those jobs, a new truck driver would sign up to work for us, download our app, learn how to use the app and view Convoy as a source of work. And it didn't take very long. We only did it a handful of times in Seattle and then we had 10, 15, 20 drivers on the platform. And as soon as we had that, you know, that active marketplace going in a local region, we were able to bring in some more work and go back and forth. But that's how we got it off the ground originally. Bunch of other challenges scaling that, but that was kind of the original bootstrap for the cold start problem.

Yeah, no, that's really cool to hear about, you know, solving the cold start problem. And it sounds like, you know, a very compelling way to kind of recruit some, some of the, you know, some of the drivers. So, and also too, I can imagine, did you, you know, constrain to like local markets to kind of build up initial liquidity or?

Yeah, initially we constrained to local markets and then we really were thoughtful about geography going forward. And that, that can be tricky because your customers are usually not just shipping in Seattle, for example, they wanna ship between Seattle and Portland or Seattle and Spokane or Seattle and Sacramento. And, you know, many of these companies are looking to ship all over the country. And so to kind of go to them and hold the line and say, well, we can give you this, this differentiated service with better visibility data, track it on an app, et cetera, but we're only gonna be able to work for you in a small portion of your business to start. And to hold that line and have like the resolve to say no when they're, when you're excited to grow your business, get more revenue and they're asking you to go to new parts of the country was difficult. But we actually held that line for a long time. And that's the, that was the second part of figuring out the puzzle. You kind of, what is that unit of marketplace? So I mentioned earlier, the first was like, is it gonna be supply-led or demand-led? And I think if you're facing a marketplace, if you're building a marketplace and that's your challenge, you need to look at the unique characteristics of it and kind of experiment. And so in our case, because we had generally at least 24 hours from when the demand side said, I want to use you to when we had to get a supplier to do the job because it was, you know, a job that picked up within 24, 48 hours, we had enough time to go get a carrier on board. So that model worked for us. Later when we thought about geography, We had to think about lanes. So lanes, the lane became the unit of marketplace for us. And that was Seattle local had a marketplace, but Seattle to Portland had its own marketplace. Seattle, San Francisco had its own marketplace, you know, San Francisco to Boise had its own marketplace. And so these lanes were interesting. You would kind of see a lane develop as a marketplace as we built supply and demand on that lane. And so we ended up building kind of a data system that would track the supply and demand balance on each lane and try to predict like when that was going to happen.

That's really interesting to learn about. And some great points you mentioned as far as for marketplaces and that we could probably have like an entire conversation, you know, on, you know, how Convoy evolved and kind of scaled. But could you maybe briefly walk us through, you know, how it evolved as you began to scale and then maybe what some of the most pivotal moments for growth were?

There's a lot of dimensions to it, but I'll focus on the marketplace aspect of it. But, you know, being a new company, there were a lot of different dimensions to scaling the business. On the marketplace side, we had to figure out kind of how to measure the success of these different lanes I was mentioning. And so, and how many of them there were and whether we were gonna care. And so at first we started thinking a lot about, you know, what are the overlapping lanes and what are the dimensions of our marketplace? For example, If you look at the supply side in, in the trucking industry and you break it down, there were like distance of the job was one really key criteria. There's a category of truck driver or local that wants to be home every night. There's another category that will do regional trucking that kind of plans to be home every week approximately. And then there were longer haul truck drivers that would have more sporadic schedules based on the demand and based on where they can go. And they would drive across the country and maybe be out for a few weeks. So if you imagine you look at your supply side and you're like, okay, well, one dimension of this, it's not just one supply of truck drivers, it's actually 3. There's local, regional, and long haul. And then you have to look at it and say, is it, is the job we're gonna run today a dry van, classic kind of cube behind the truck, or is it a flatbed job, or is it a refrigerated job or some sort of hazardous material? Does it need to go in and out of a port where it requires some special authorization? And so now you're breaking your marketplace down a little bit further by not only distance but type of truck. And maybe there's some other factors like some drivers don't want to take really heavy beverage loads and some drivers don't want to take loads of a certain type. You know, there's too much risk or they need a lot of insurance because it's really expensive. It's like, you know, computers, they don't want to take that. And so you end up breaking your marketplace down. So to scale our business, it was really recognizing what is the unit of marketplace And which of those units of marketplace are we gonna actually go after and scale? It wasn't just thinking, let's scale trucking. And so being really deliberate about defining these, categorizing them, understanding the market size of each one, figuring out which ones have overlap. It turned out that a lot of trucking companies that had a drive-in would also have a flatbed, for example, or some with a drive-in would have a refrigerated truck. And so we'd said, okay, well, looks like there's some pairing there, but those that have flatbeds don't really have refrigerated trucks. Typically you don't see that pairing as like a, the two. So, you know, figuring out, okay, here's all the breakdown supply. Now these ones really go together well. And then on the demand side, do our customers have short haul and regional and long haul? Do they have dry van loads and refrigerated loads or flatbed loads? And so you know, trying to match those up, figuring out for the supply side you're gonna build which customers actually have demand for that exact mix. And so, you know, for us, that was the key early on, was being deliberate about that. And the reason I mentioned before, it's a managed marketplace. So we did, we had pricing risk and delivery risk, which meant when our customers asked us to deliver a load for them, they were asking us to take responsibility for that being delivered, and we had to make sure it was there on time. And if we had to, if we took the job and it turns out the trucking market changed in the next 2 days or the next week and truck prices went up, we would have to pay more to source a truck. So we had principal pricing risk in the transaction. And so when you have that kind of risk, service risk and principal pricing risk, you have to be really thoughtful about the marketplace you're working on. You can't get expanded too fast. And so if you think kind of 2015, '16, '17, and even '18, up until like for those years, we really were thinking market segment by market segment for all the things I just described, matching supply and demand and trying to be really deliberate about measuring how much supply and demand do we need on one of these lanes or one of these markets to see like the cost structure get better, to make sure to see every driver start using the app and getting into that habit. To make sure we could predict how much, you know, how many trucks we had in that market so we didn't fail our customers, etc. So it's a, it's a bit of a deep answer, but marketplaces are complicated, and ours was a very, you know, complicated scenario. So we had to be really deliberate about it.

Yeah, no, as you mentioned, marketplaces are very, very complicated. So you got a lot of head nods at that, and it sounds like, you know, on, on the scale of kind of 1 to 10, Convoy is way more operationally complex compared to others. But I'm glad you actually mentioned the note as far as risk, and that's something that we're, you know, often talking about here as founders at the earlier stages, which is, you know, how much risk to kind of take on as a marketplace and when. So, you know, I guess like on that note, like how would you advise founders on how to think through this?

I wouldn't take on the level of market risk that we took on if I was doing it again. It was, it's funny when I, I'll tell you, I've never actually shared this. I don't think when I was first starting, I was looking at Seattle local trucking trying to understand how it worked. And there's a type of trucking which is kind of that straight truck, same-day delivery, urgent delivery. And I can't remember the exact term that we used at the time for it. There's several terms for it. And what's interesting, I went out and talked to a bunch of those truck drivers and they said, oh, the way it works in our business is we pay 45% or 40% of the revenue per job to our broker or our dispatcher that is finding us these local jobs. And I was like, oh, that's a really fat margin. That's 40%. And it's not like a spread, it's a take rate. So obviously if the job's, you know, $1,000, you're paying $400 to this dispatcher and it's a take rate. So I remember when I first went into this business thinking, okay, this is a take rate business with pretty good margins. 'Cause the first example I stumbled into and spent the first network of 10 truck drivers I talked to were all from that world. And I think if I would've maybe the first 10 conversations I had realized that it was actually a buy-sell spread high-risk, you know, pricey marketplace with like the delivery challenges, I would have been like, I don't know, maybe I should look for a different segment. Obviously, we could have changed it later, but we kind of stumbled into that. So I guess I would say when you're looking, try to minimize the number of dimensions of risk you have. I think it's different with AI and we can get into this a little bit. But if you're thinking about, are you taking on risk of service? Are you responsible for the quality of the goods or services sold on your marketplace? And delivered? Do you have pricing risk? You know, if you price wrong, do you potentially lose money or do you, are you on the hook? You know, is there a high cost of capital for servicing your business? Are you gonna like, when you're, are you getting paid faster than you're paying out or the opposite in your marketplace if you're managing the money in the transaction? Or is it like real time? You know, are there rapid changes to how your marketplace operates? throughout the year? Like, are there seasonal changes and regional changes, or does it all kind of operate similarly? And I think those are the kinds of questions that are really important to ask. And I think if you can find one that maybe has 1 or 2 that are kind of concerning, but not 5 to 6, that's probably a good place to start because these are all things you have to solve for and that obviously create a lot of risk for your business. So that's kind of how I think about it. Just identify the dimensions of risk in the space And pick one that has fewer if you have the option between a few good ideas.

And I'm glad you also mentioned that note as far as the AI. So we're going to get into that here in a little bit. But I did, you know, as far as Convoy though, you know, when it comes to like team building and hiring, I believe you scaled the team to what, over 1,000 employees? So, you know, do you have any tips as far as like when it comes to team building and hiring?

I mean, one thing I did was I spent a lot of time really actively recruiting on LinkedIn. Like that was not, it's not necessarily a rocket science observation, but I had an approach to reach out to people directly with a very, with a subject that made it obvious I was not a professional recruiter. That was important. And a pretty like lightweight message that was just talking about how exciting this opportunity was and if they wanted to connect. I found that that led to, you know, half of the great hires that Convoy made were kind of just direct in the first 3, 5 years were just direct outreach and not hiring recruiters for a long time. We didn't hire, we didn't really hire like a recruiting agency or anything like that for a very long time. We did a lot of it really proactively because you're trying to figure out, you know, you're trying to really attract people that are looking for that kind of an opportunity. So if you reach out the right way and you position it the right way and you talk about the challenge the right way, you're going to attract people that are looking for, to be part of a story, to be part of something challenging, to be part of something kind of inspirational and exciting. And we were really trying to filter for that because we needed that early on at the company, given how many unknowns we faced and how challenging this was. One thing we also did, which I found to be useful early on, this is really tactical, but we would try to interview really in a really challenging way. Like we would make the interviews hard specifically because we wanted people who are really strong to be attracted to that. We found that the harder the interview, kind of the higher quality candidates would be more attracted to that. They would be like, feel challenged and they like that versus maybe if they went to an interview that was easy. So we actually tried to do that a lot. I'd say in those early years, again, we're forming the culture of the company. I remember one of my engineers like, this person's amazing. They're so good. Like, I don't think we need to interview them anymore. And I was like, go give them the hardest interview they've ever had. Like, make it feel like they have, they're really earning it, you know, to be part of this. So I think when you have a startup or like an early stage company, you have that opportunity. I wrote down like all the problems we were gonna have to solve. I really like to start from that orientation. Like, what are the problems our business is gonna have to solve? We're gonna have to, there's a bunch of things that'll be challenges for us. And so who's gonna be really good at solving those? Like, what are the people in, you know, that have the, have had to face similar problems before? And I thought about it that way because I didn't like the idea of like job title matching. I was like, I need a, VP of ops, but what does that really mean? So I was like, okay, well, the challenges we're gonna have to solve over the next year are these 3 things. Who else is out there solving those problems right now, independent of title? Like, what are the— what's the problem similarity? So that was one thing we thought about pretty early on. We wrote a lot of that down. And I'd say, like, from a financial perspective, you know, we did raise a lot of money over time. We raised almost $1 billion of venture or like private equity money. which is pretty crazy given the, but again, the scope of the managed marketplace and all these things we can get into, like why it was challenging to scale that nationally. But we were trying to be very frugal in terms of like salary and what we could afford, especially in the early days. So we would always make 3 offers. I won't say always because it was only probably the first 3 or 4 years, but where we would give a low, medium, and high salary offer coupled with a high, medium, low equity offer. But the reason we'd give the 3 offers is because we didn't want to get to the point where we had a great candidate get to the end of the process with us and then like come back and ask for twice as much as we were offering or come back with some like really big request that we couldn't afford. So by giving multiple offers, we created the range of negotiation. Like we kind of, by giving these multiple layers, you kind of, the person would get 3 different offers, see the high-low cash, the high-low equity, and know there's like increments of sort of negotiation that we're comfortable with because we're showing them what the increments are based on what that looks like. Does that make sense? So little things like that did help a lot with being able to land good people. But the actual attracting people was all about just friends of friends, networks, and lots of LinkedIn outreach.

And you did mention as far as fundraising and raising, I believe, over $1.1 billion in capital. So quite a bit. But if we kind of go back to the earlier stages, what were some of your, I would say, Fundraising learnings from like the earlier stages?

I mean, it was venture capital for a while, then private equity, and then there was some debt involved as well. I think the early, I mean, it's storytelling. Well, there's, I guess at the very earliest stage, there kind of were 3 things. One, do they think that the idea is, can they understand the idea? Like, do they understand kind of what you want to do and can they succinctly explain it to somebody else? That's like an important factor. If it works, is it going to be a big outcome? Like, does it have the potential to be a significant outcome? And do they believe that the team early on is able to do it? Like, do they have confidence that you can actually pull this off because of whatever qualities? Those 3 things were the most important things. And, you know, and that all comes like delivering that comes down to storytelling, but also their confidence in you can come down to storytelling a lot because An early stage company is telling stories and it's telling stories to customers and to partners and to vendors and to investors and to employees around what the potential is of what we're building because it hasn't happened yet. That's the whole point of raising venture capital is you're gonna take strategic risk. If you're just taking execution risk, which means you're using a known playbook, you know what you're gonna do and you're gonna go execute it better than other people. That's a different kind of risk. You probably don't need venture capital for that. But if you're taking strategic risk, which means I have a business model or an approach that no one's really— it's different and we don't know if it's going to work. Then you have to be really good at kind of like knowing how to tell your story. And so I'd say those 3 things were what mattered the most to angel investors and early investors. And then what brought it home was storytelling.

It's going to be really helpful. And I did want to transition to talking about AI because I know you briefly mentioned it and that's a topic that we're always discussing here in the community now. So, you know, how are you seeing AI impact marketplaces? And, you know, specifically maybe when it comes to like the future of B2B marketplaces, how do you think it'll impact it?

So I'll give an example with how Convoy was working and how I think I would do it differently if I was building it today, which, which answers the question. We had to get all of the participants onto a platform so they could transact. So if you look kind of like pre-Convoy, the way the trucking industry worked is you have a truck driver, you have a broker, you have a shipper, and they're all calling each other and texting each other and sometimes emailing each other. To make a deal. And so there's a lot of negotiation, data sharing that's happening over the phone, over text, and over email in natural language. We wanted to scale this and make it more automated so you could be a carrier, you could go on and see all the jobs in one place. They would be ranked and, you know, the relevancy would be high for you and you could negotiate in this app with the shipper or to kind of like with against the price. So like we created an app for drivers to go see all the structured data of every job, bid on those jobs, share their location, share the paperwork, share the information via an app that allowed, it's like a common platform that they could learn to let them interact with Convoy. And then the shippers would have their own kind of app experience or that they could use to then go interact with this. So we had to structure the experience on both sides. for it to work. The only alternative to that was to call them and talk to them in an unstructured way using language. But we didn't have, you know, GenAI at the time, so we had to structure it so you could automate the process. If you fast forward to today, many of those things could be done using natural language with AI. So again, you could out— you go, it's sort of like before Convoy built this platform, it was all phone calls and text messages. Well, that's all now possible again with generative AI because it can use natural language and it can talk to you like another person. So instead of building a separate platform for the marketplace where people go meet to conduct the transaction through automated outreach over text, phone, etc., using AI, and they don't have to go learn how to use your platform and discover it, so you can just meet them where they are.

I'm glad you mentioned the point as far as, you know, reaching people where they are, right? We have a lot of marketplaces that are leveraging as far as like for direct kind of sales outreach as well. So kind of a related question that I actually had on that that I was like, be curious to learn more about is like, How would you maybe suggest, you know, founders think about like positioning, you know, like selling kind of like AI or kind of agents, you know, when they're selling to like, I would say, you know, like SMBs or more kind of legacy offline businesses?

I mean, I think what I would do is I would look at what are all of the things that a marketplace is, a multi-sided space where people meet to conduct business, sort of find each other and get something done, right? And so outside of your direct marketplace, What are all the activities that each participant is doing to like get their content or service or product ready to be shared on the marketplace or to make a decision about, you know, what from the marketplace do they want to buy or transact? Like really draw out the activity map that they have to deal with and then think about what agents can replace those steps for them so they can transact and, and kind of like prepare their stuff faster. Or if it's a decision that can give them higher confidence in the decision they're making. I think those are general, but like those are 2 areas where I would, I would start if I were breaking it down.

Yeah, no, it's definitely gonna be helpful. It's just something where we're at that's kind of top of mind right now in the community. Someone wanted to kind of ask about that and expand. Well, cool. So we're gonna make sure we can save some time for the group Q&A. So we're gonna jump into that here in a little bit. But you know, right before we do, did you have a few tips maybe that you'd like to share with the early stage marketplace founders?

It takes persistence like marketplaces. I think that story I told where I went and talked to like 60 carriers and none of them were interested. That was a hard moment. And I remember I'd hired somebody around that time who was like interviewing people, telling them that nobody was signing up. Like, we gotta like not say that. We gotta have confidence. We're gonna figure this out. So I'd just say like persistence is just so important in a marketplace because it is hard to find that right thing, but that's, it's sort of persistence. And then like I said, break it down, like really understand the unit of marketplace you're talking about. Like where does it happen? What are on the supply and demand side? What are the, you know, how do you simplify it? You want some good advice someone gave me actually on this was in your perfect world, every unit of demand that you have can be serviced by every supplier in your platform. Like that's a perfectly liquid marketplace. where like everything, there's a perfect match between suppliers and demand, supply and demand. So the closer you can get to that without making your, you don't wanna make your business infinitesimally small, but the closer you can get to that by limiting like the dimensions of your marketplace, the easier it is to get going.

Helpful tips. So cool. So we're gonna kick off the group Q&A. Hey Molly, do you wanna come on?

I'd love to. I was really curious what your process of rapid vetting looked like. For context, I now run a marketplace for kids' after-school activity camps. So there are some parallels about really being cautious around trust and safety. And one of the things I found at Bellhop was initial vetting was a challenge, but continuous compliance monitoring and management was even tougher. So I'm just really curious how you managed that.

I mean, we vetted our demand, but our demand was mostly medium and large enterprises. And our biggest customers were like Starbucks, Costco, Walmart, etc., like these kind of pretty large enterprises. We wanted to make sure they were gonna be good partners, but we weren't like vetting their creditworthiness or like, you know, safety operational stuff as much. Most of the vetting was on the small carrier side and trucker side. And a couple things. One, we took it really seriously very early because we were taking principal risk and we were on the hook for the pricing, the delivery service, et cetera. One of the first hires we made was a quality and compliance person who came in and they were, you know, doing that job for 6, 7 years. And we started with as much over-the-counter data as we could get. So we looked at what are all of the sources of information out there that already exist that let us kind of vet these carriers. There were government resources that would track and monitor the carriers. There were private companies that would, you know, have the insurance data and information. And a very unique thing we did was Because most brokers were calling and kind of manually doing this, they wouldn't be like checking the status of the safety, the safety status, the insurance status, the compliance status of all of their carrier before they called them and brokered the deal. They would maybe check later, like right before they, or they would check like once a month and maybe they'd miss it sometimes. We automated that. So we very quickly found all the people that we thought would have the data we needed. to vet a carrier, built direct connections into them. So every time we had a carrier that we were about to get a job to, we would first check to make sure that they were compliant, or we would check like before the job started. And therefore we didn't have a— we went from like 5% of jobs. I remember early on we had to pull back 5% or so because the carriers actually didn't meet the qualifications when they actually started. But a few months in, we were able to automate it and that pretty much went to zero after that. So that was really critical, that like automated vetting versus having to remember to do it or having a period of vetting.

That's a great question. Hey Dan, I actually wanted to ask a question that a founder sent me in advance and I added to the doc, which is around, you know, the market expansion, a new market launch and expansion playbook. So, you know, what did that end up being? And did you, and they were asking specifically, like, did you use like a GM model in markets like Uber, for instance?

We didn't have like local GM teams because, well, I mean, we had a GM team in Seattle. We were just starting and we hired about 5 people from the Seattle local office from Uber who were kind of doing the local Uber office. So we had, we initially thought that would probably be the right model. Uber had to kind of meet the driver, make sure the car was used, was like good enough to drive, make sure that all their stuff, because Uber was creating a new workforce, like A lot of these people weren't doing this job before and their car was not being used for this purpose before. So they had to have a very hands-on system and it was entirely local. Like everything they did was local. So having people in like a city for that local experience and making sure the drivers and the cars were vetted was really important for quality. In the trucking industry, we quickly realized that, you know, everyone we're working with is already a truck driver generally, or they're going to truck driving school, having to go through like a formal qualification process and find a truck and get authorized with, you you know, with the government to drive. So there were other people doing some of that like work and there was an existing set of truck drivers we could tap into. So we didn't have to go verify that like they had a truck that worked. Later on we did more verification stuff, but like initially we're like, look, let's trust you can do, you have a truck. And even though some of it was local, a lot of our lanes and the unit of marketplace we talked about became between cities. It was like Seattle to Portland, Seattle, Sacramento. And so the number of locations expanded really quickly. Even if you only had 10 cities, the combination of those cities is like pretty significant. And you're thinking not only about like the experience in either city, but maybe the experience of that driver between the cities with like truck stops and, you know, stations where they might have to pull over. So we started thinking about it a little more holistically. So we assigned people to broader regions. So we did have people kind of assigned to regions and then we had people assigned to this, the different parts of the marketplace. So flatbed trucking, dry van trucking, you know, on the shipper side, refrigerate, beverage companies, CPG companies, retail. So we really kind of actually assigned ownership more at the players in the marketplace and then regionally based on the marketplace geography.

It's gonna be super helpful, really interesting to learn more about too. So hey, Danny, did you wanna come on? Saw you raise your hand.

I found it really inspiring the way you talked about how you hire people in the early days, especially kind of reaching out directly on LinkedIn and, you know, going through the interview process and being quite rigorous. I guess if you were thinking to start something today, you know, how do you think your approach to hiring would change? Would the org structures be the same? Would the type of people you'd be hiring Would they be the same number of people? Like, how do you think your approach would change?

I mean, I think definitely we over time hire fewer people. I think there's more we could do using AI today than we could do back then to give people like more leverage. I think early on I'd probably hire very similar types of people because relationships still matter a lot in this industry, for example, and it doesn't matter for every industry, but in our industry it mattered a lot. And so having somebody, you know, I remember when I was first starting, I kind of wrote down 4 different people I thought I would need. Like, I need somebody who understands how trucking works, like trucking and operations work. I need someone who understands how marketplaces work. I need somebody who understands sales and the demand side of a marketplace, and I need somebody who understands operations and the supply side of the marketplace. for our business. And I actually had a physical 8 by 11 piece of paper with a 2 by 2 drawn like this and the dimensions at each of the ends of the points. And I wrote every name I met into the boxes based on where I thought they fit on that grid based on their skills. And I'm like, I'm gonna work my ass off till I hire one person in each of these areas or somebody who can do both of them or whatever. I think I ended up hiring 3 people that covered 4 of those. But I would do that again and that was deliberate and it was because you do need that like understanding of those spaces and it's important to put a person against the major problem areas of your business. Those are the major problem areas of my business, you know, and that took, obviously that was not like overnight. That was the first 6 months kind of hiring people. And then on the engineering side, I think I would probably hire more generalists. builders who can, who have a broader set of skills to get started than I did at the time when I hired several different specialists in different areas of engineering because, you know, we didn't have the same tools we have today.

We're gonna try to get now one last question if we can. I know we're almost outta time here. Hey, Peter, do you wanna come on? I saw you raise your hand.

I was rewatching your FreightWave chat from 2024 and you talked about the risk founders face, like execution risks and strategic or even existential risk. when trying to innovate. With AI changing how people solve problems, do you have any updated advice for founders navigating these risks today? If you're using AI to solve the problem, you are almost automatically in the strategic risk category because you're doing something that hasn't been proven yet. The point I was trying to make on that one was I learned a lot about like that, those kinds of risks in terms of a business you build and I think, you know, I can't— we were in a very traditional industry where people were like, why are you raising all this money and hiring all these people and hiring all these engineers in trucking? It makes no sense. Like, why would you do that? Like, it's like, it's so easy to be a broker. Like, why are you— why is this hard for you? And I think what I realized, a lot of people who have really only taken on execution risk, which is like, hey, I apprenticed, I've— or I know somebody ran this business, right? There's this business model. I'm gonna do it better. I'm I'm gonna out hustle them. I'm gonna get the deals. I'm gonna like run my shop better. I'm gonna run my team better and I'm gonna do a great job at this business model. That's awesome. And I have a lot of friends that do that and that's like a very worthy pursuit that is very hard as well. You have to execute really well, but you know that it's gonna work if you can execute really well generally because it's a proven model. The strategic risk, and I'm just giving you guys this context of what I, what he was referring to, is when you're setting off in a direction that no one has really proven yet and it's not sure, you're not sure it's gonna work. That is very different. That is what venture capital is about. And that's why it's often hard for people that aren't doing that to understand why people are investing money in these things. Like, I don't understand why would you invest money in this thing? It's, you know, it's because it's a totally different type of risk and it's like, if you haven't done that before, it's really hard to understand what that even means. So that's what we're doing when it comes to AI. There aren't a lot of examples yet. I mean, there's some, but like there aren't a lot of examples where AI has like replaced the human model entirely and is like solving the problem. So saying I'm gonna use AI to like, to offer my customers a new type of experience that they're gonna be comfortable with, that's gonna work with the quality that it's gonna need to work over time and it's gonna scale, et cetera, is something that hasn't really been proven. So you are taking strategic risk and I think that's why there's a lot of venture investment going into it because Categorically, it is new and it's unproven. I guess, you know, probably if you were to do a 2 by 2, it's like, is the business model proven in general or not? And has AI been shown to be able to do it or not? And so, you know, if you're taking on a business that inherently has strategic risks, it's a new business model in a new area and you're trying to use AI for it. That's, you You know, that's going to be really a challenging thing to go do. I think you're seeing a lot of people take more traditional businesses that have a working business model and try to improve the performance of that business and lower the cost using AI. That's like a pretty common model we're seeing as well right now. Both are really interesting, but that's kind of how I'd break down the risk between them.

That was a great last question and kind of way to wrap things up here. So we're almost out of time and I want to be conscious of that. But, you know, Dan, huge thanks for taking the time to join us here today for the group chat and also being transparent and sharing some of your, all of your experience and kind of learnings with us all here. So this is definitely a real treat to have you join us as a special guest. And I actually had one last question for you, and that's my usual closing question, which is, you know, if you could go right back before you started Convoy, you know, what would you tell yourself about marketplaces specifically?

I would tell myself, one, I would do it again. It was awesome. I love, I actually have, I was deeply fascinated with marketplaces before Convoy. So I'll just, I'll share that too. I spent a lot of time thinking about them, researching them, meeting marketplace companies. I just thought it was fascinating all the different ways you could have a marketplace. I think that I would, again, I'd go back to the, that like complexity thing. I would say, hey, maybe I should limit the complexity of what I'm doing a little more in some ways. I think that's probably what I would do. Try to have fewer dimensions of complexity so that you can execute better on the ones that you can control.

That's gonna be great parting advice for us all here. And then last but not least, time for a quick plug. Where can we keep up with you at?

I've been less active online recently, but I'm on LinkedIn. Definitely connect with me on LinkedIn if you, if you'd like to follow up from this. Daniel Lewis at Twitter as well.

Awesome.

I'll include all links in the recording and you'll have lots of marketplace founders coming your way. So we once again really appreciate taking the time to join us for the group chat. And thanks everyone for the great questions today.