Once you have that foothold, once you're talking to the customer about something specific, they tend to open up about the other problems that they're dealing with, right? These may be other projects on their plate or other expenses or costs they're trying to fight. Those become kind of obvious from the pattern perspective. The 80/20 test, those are almost always immediately adjacent to what you're working on because you're talking to the same user or same buyer. And if you kind of were to like write them out on a whiteboard, you could almost just draw circles around them. So we found that strategy works a lot better than let's build this product for this customer here and build this entirely different product for a different industry or different user. So the effect compounds if you have like a bunch of disparate products that are either completely unrelated or even sold to different customers. You're now acquiring entirely new customers each time, having entirely different conversations. And maybe you can kind of operate as a holding company, but it's not as efficient as if you're already there and having the conversation.
Welcome to the PE. I'm your host Jer Novak, founder of Banana Capital. Today's guest is Sanjit Biswas, co-founder and CEO of Samsara, the fleet management and safety platform. People that use our system are everyone from the frontline workers, so the drivers to the dispatchers, the safety managers, the mechanics. At the time of publication, Samsara is a public company worth over 26 billion. And we unpacked how exactly they went from zero to operating at over 1.6 billion in revenue in 10 years. I think the most important thing for us was focus. Staying focused on our customer, understanding what other problems we can solve for them. Our conversation explores how Samsara uses AI to impact the physical world, how they use AI internally, and how the products prevent over 200,000 deaths per year. So once you have AI coaching at the edge happening within a second or two, it really makes a very big difference. Sanjit has built two unicorns on hardware and we unpack how he did it including what most founders and investors get wrong about hardware.
Either folks underestimate the complexity of building hardware or they just say I'm going to avoid it altogether. Why he thinks customer first instead of product first and how to know when you have product market fit. We started getting asked a question like, "Hey, this beta product, is it a secret or can we tell other people about it?" How to get good at sales as a technical founder and how to spend more time with your customers. I remember vividly one of the customers had an interface and it was running on MapQuest. And it's like a kind of throwback to the '90s, but that was happening in 2015. We talked about getting his high school online in the '90s and the research project that turned into Sanjit's first company, Meraki, and its $1.2 billion sale to Cisco in 2012. I still remember the first time we got an order. It was like, "We'd love to cut you a PO." I was like, "That sounds great. What's a PO?" A quick thank you to Alana Goyle and Adi Verasami for helping brainstorm topics for Sanjit. A reminder, I publish two episodes of the PE every week exploring the world's greatest startup stories just like this one. Tune in next week for a conversation with Eric Vishner at Benchmark and check out last week's with Michael Kendan if you missed it.
Now, a quick message from Bolt before talking to Sanjit. This episode is brought to you by Bolt. That's Bolt.new, the easiest way to turn a text into a full stack web and mobile app. Bolt is one of the fastest growing software products ever, hitting 50 million in ARR 5 months after launching. And regulars of the show might remember when I had Eric Simons on where we went deep into their 7-year journey to overnight success, including almost shutting down a few weeks before launching Bolt. They're now hosting the world's largest hackathon with over a million dollars in prizes kicking off May 30th. Over 30,000 people have already signed up to build apps with AI. No coding required, 50 IRL meetups worldwide, and over a million dollars in prizes. Whether you're a founder, creator, student, or anyone sitting on an idea, this is the moment to try it. It's free to join and you can win real prizes like best oneshot prompt, top building public journey, and most likely to get funded. Sign up now at hackathon.dev. Tap the link in the description or head to hackathon.dev. Now, let's talk to Sanjit at Samsara. Sanjit, welcome to the show.
Yeah, thanks for coming on. Should be pretty fun. We were just at your kind of customer demo on AI and autonomy that I know we're going to talk about that a little bit. But I'm just curious real quick for people who aren't familiar. What is Samsara?
Yeah. So we help with the world of physical operations. So a lot of companies whether they're construction companies or utilities or logistics companies, they're looking to digitize. The way they do that is by putting sensors on their trucks and their other assets. They need software to help manage their operations. So we provide all of that and we also help them go through that digital transformation journey. So figure out how do you actually do it practically especially if you have tens of thousands of frontline workers.
So what does the kind of the products look like to do that?
So we have a few different products. Most of our customers adopt multiple of them together. We started out with GPS tracking. So understanding where your trucks and your people are in real time. We added safety cameras. So cameras that run AI models in them at the edge. And the idea there is when you're out on the road sometimes you develop a bad habit like you take a look at your mobile phone or you have tailgating habit or you speed. We can help break those habits by having the AI give you a quick warning or a nudge and it has amazing outcomes. So that has become now our number one product line. And then on top of that we also help track equipment, do workflows, kind of everything else that is related to operations.
And so who is a customer and a user for this kind of a product? Because it's maybe a little bit abstract for the average tech audience.
Yeah. So, you know, maybe to help visualize it, I would think about a company like a DHL, right? Those yellow parcel delivery vans, there's thousands of them crisscrossing all over North America. We work with them to help improve their safety and their efficiency. So, the cameras help keep their drivers safe on the roads. They're driving thousands or even millions of miles at their scale. We help them understand, are they able to hit their route stops on time? Should they replan their route because the traffic or the weather's a little crazy. So, we do all of that. And then the people that use our system are everyone from the frontline workers, so the drivers to the dispatchers, the safety managers, the mechanics, kind of everyone involved in the frontline operation.
And so the way that they actually implement the products, how does that usually play out? Because when I think about a truck or a car like there's a touchscreen and like it's connected to the internet, you know, do you really like how do you why do you need all these different devices attached to the vehicles?
Well, a lot of it is around consistency of collection of data at scale. So, what we have is a little black box that plugs in. That's the GPS tracker that makes sure that you're getting high quality data across your entire fleet. Doesn't matter the make or the manufacturer like you know what year it was built or any of that kind of stuff. The cameras are all consistent as well, so they have consistent driver safety scoring and all that kind of stuff. So, the whole idea here is it's plug-and-play. They're up and running within minutes, and then they can start collecting data about their operations, but more importantly, they can manage it at scale. So, if you think about a company that has thousands or tens of thousands of trucks, they need something that just works out of the box. And a lot of the trucks, they're from, you know, like 20, 30 years old, right? Like they might not have some of these features already built in.
Yeah, that's right. I would say a 20, 30 year old truck is pretty unusual. Most of our customers have vehicles for 5 to 10 years because they're running them every single day. But again, the feature set varies among these different trucks and the consistency of the data really matters.
Interesting. Okay. And one thing I wanted to ask, so the name Samsara. What's the significance of that?
Yeah. So maybe a little backstory, my co-founder John and I, we've been working together for over 20 years now. So we met as grad students at MIT. We were both part of the same PhD program. We ended up in the same lab working on a project together. So we worked on that together. That was back in the early 2000s. It was related to Wi-Fi. That became a startup company called Meraki that we worked on for many years and that's now part of Cisco Systems. And then we started Samsara together. So Samsara is a Sanskrit word. It means the cycle of reincarnation and rebirth. So we thought it kind of fit and along the way we met a lot of great people and they now work with us at Samsara too. So it kind of has been a reflection of the team if you will.
Okay. I know I want to ask you more about Meraki but I know we want to talk about Samsara first. One really interesting stat that I saw was the Samsara product has prevented over 200,000 car crashes. I think like last year, what like that's an insane stat. How do you even track that and what are you kind of preventing?
Yeah, and it's one that we're super proud of. So, like I said, a lot of the times people develop bad habits that lead to risk out on the roads. You can kind of see a before and after effect. So, there's like a baselining that you're able to do. Imagine you have someone with a GPS tracker but no camera. You can get a sense for okay, per thousand or per million miles, how many of these incidents happen? And unfortunately, it's just statistical, right? There's a number of crashes that happen in the US every single year. So we're able to baseline versus that and then you can see what happens once you turn the coaching on. So once you have AI coaching at the edge happening within a second or two, it really makes a very big difference. It can be anywhere from 20 to 80% reduction in the number of accidents. And that's simply by changing behaviors. And a lot of the times the will and the intent is there. It's just bad habits form. And so if you can kind of get to the source of it, it changes the outcome.
So when you say AI coaching, I might hear that and think, you know, is it like a thing that lives? I don't know if you guys saw that recent demo of like the glasses with live on the in your view and the glasses giving you coaching and advice. Like is it the same kind of a concept with a driver in a vehicle or how does that play out?
Maybe less like literally in your face than that. It might be something as simple as a reminder to put your seatbelt on. A lot of folks don't know this, but in the United States, it's still the case about somewhere between 8 and 10% of folks don't wear seatbelts. It's one of the biggest things you can do to prevent accident or injury really. So that's an example of something that's like a slight nudge, right? And that's again like a form of AI coaching. There's also this idea of a safety score. So the reality is most drivers do a really good job on the road. And we see a lot of customers where they want to do recognition rewards programs for folks that are doing a great job. So the cameras are able to provide that composite score that could lead to like an employee of the month or you know you can do contests and things like that. So these are all different forms of kind of gamifying safety.
Interesting. Yeah. Because you think safety is pretty boring and why would you—pretty dry but at the other end like you do want to be safe because you know you're working in a hazardous environment with heavy equipment like you want to be safe. So then what are there other ways that you're using kind of AI to impact the customers and like the frontline users and workers in the product?
Yeah. So once we started using AI and we introduced AI coaching probably five six years ago, you started to see it was really useful for helping sift through tons and tons of data. So if you think about safety coaching, most of the time nothing's going on, right? And it wouldn't make sense for a person to be watching those videos. It's just like not practical. But the AI model is able to run continuously and find those patterns. We started applying those ideas to other sort of concepts within operations. So, could you replan your routes because you just keep missing the stops? Maybe it's not practical to run a certain route in a certain time frame. So, we can run AI to help identify things like that. Fuel coaching is another one. So, if you think about these fleets, they drive millions of miles. There's certain miles per gallon. You can affect that by how you drive. And so a lot of our customers will run little contests on like who's the most fuel efficient or you know managing vehicle idling, things like that. That cuts down both emissions and fuel consumption. So all of these are use cases for AI because of the amount of data.
Yeah, I was going to say it sounds like these are things where you probably could not just start a company initially and this is the product and how it works. You'd kind of need all this unstructured data that just kind of exists in the wild that you're now capturing, right, because of the product.
Yeah. And kind of going back to why do you need to install the hardware? Once you capture the data, you find the patterns in it. If you were just looking for one sort of attribute, you might miss all of this extra context. And then we can actually tie all the stuff together as well. So like what behaviors related to driving actually result in less risk, right? Things like that. You can figure out which vehicles do you need to maintain. So, we can see the fault codes on the engines. We can see how many miles are driven, how hard you're driving on those miles, that kind of thing. And help you come up with a better maintenance schedule.
And like is there any, I don't know, in your world like more physical transportation? I know if you're working in software or consumers, there's like tons of hype on AI right now. Is it the same in kind of the physical world like in transportation?
I wouldn't say it's the same kind of hype. There's a curiosity. I think everyone's interested in AI and people are using ChatGPT and wondering what else could you do with it. For our customers, they're really about just solving real world problems. So, you know, if you think about a construction company, for example, safety is a big deal for them, right? It's something where they don't want anyone getting hurt on the job in the first place. It also affects their revenue. So if you're known as a safe company, you tend to get more work and there are safety scores and things like that that get published. So they want to use AI to help improve safety for that reason. So it's a very kind of pragmatic viewpoint.
Yeah. Like it truly just like hits the bottom line immediately.
It does. And these are I would say asset heavy labor intensive industries. So we were talking about a food services distributor earlier. They do a lot of hard work, but most of their revenue ends up going right back into the operation, whether that's running the trucks or the warehouses or all that kind of stuff. So, for them, these little things actually add up to a lot.
Internally at Samsara, are you guys taking advantage like anything any cool new tools or way you've changed the team or operating due to some new stuff that you built or started using?
Yeah, absolutely. So we've been kind of early adopters of AI and large language models partly because we saw how useful it was in the product but also our background is you know we're tech folks so we're always kind of curious and trying new things out. A couple examples. One is, you know, for our sales team, we have something internally we call Samsara GPT and we were early adopters of OpenAI's enterprise licensing, right? So, we were basically using the enterprise version of ChatGPT. We tied that together with all of our case studies and we started basically crawling all of the internal data we have. And it's super useful if you're a salesperson and you need to you're talking to someone in a certain industry or a certain geography. We can now pull up a relevant reference for you really quickly, but you can do it while you're on the phone with them and it's just a little Slack bot that we built. So, we started with that and then that project has snowballed that's really heavily used internally and from there you started getting other ideas of well, can we put other information in here? Can we make it available to our customers? All that kind of stuff. So, that one was sort of like a sales assist or enabling tool to just give them more ammo to work with when they're on the phone.
With relevant information. Exactly. Because, you know, you have so many customers, it's hard to remember all the stories in your head. Or maybe you're new to the company and you're still learning.
It's really helpful to be able to just kind of ask a bot really quickly.
Yeah. Interesting. Okay. You've started two companies, two hardware companies. Well, there's a hardware component of it. Obviously Samsara is very much software focused. What do you think most people get wrong? Like I feel like as an early stage investor, the first thought with hardware typically is, "Oh, it's too hard. We don't like investing in those companies. They don't scale well." I feel like over the past couple years, we've kind of gotten past that and people realize that they can be really good businesses and sometimes better businesses. Just what have you learned or anything you think that most people get wrong when they kind of approach starting a hardware company?
Yeah. Well, I would say either folks underestimate the complexity of building hardware or they just say I'm going to avoid it altogether because it's hard. I think if you have a background in engineering, it's a little bit easier to kind of wrap your head around it. So for myself, I studied computer systems engineering as an undergrad. Our research project at MIT had a hardware component. So we were pretty familiar with it and we understood the sort of quirks and nuances of it. You also understand the costs of it when you work with it that closely. And we saw a pathway to actually building a product that would be a sustainable sort of product line to offer over the long term. So a lot of it is just knowing the details but it is extra complexity. I think the upside of it is it is a bit harder to compete with hardware businesses right? You have to compete against them in terms of the ecosystem and who your competitors are ultimately.
Okay. So, for example, our previous company, Meraki, was doing Wi-Fi access points when we started. We eventually did switches and routers and other kinds of things. There weren't a lot of companies doing that out in the market. And it's because you needed to figure out how to make the boards, write the firmware, ship stuff at scale versus like a pure software product. Someone can put it together really quickly in their dorm room and you might have a hundred competitors out there. So, very different market dynamic. Yeah. And specifically today, you've probably seen those like text to code it, right? Just like make me a whatever I need and it's just there. And that's a little harder to do with like an access point or GPS tracker or camera or something.
Do you think that we're going to get to a point where it's easier to make hardware? Like I feel like it has gotten easier over the years, but I think so. And I think Arduino and Raspberry Pi kind of became really great gateways and pretty good prototyping platforms. There's a lot you can do on a Raspberry Pi and kind of prove out an idea and then at that point you go figure out how do I make something custom.
So kind of talking about Samsara specifically. How did you first come up with the idea? How did it all come about?
Well, you know, there's the initial idea and then there's what the company ultimately focuses on product wise. We kind of started customer first with this company as opposed to product first. So maybe quick backstory with Meraki, a lot of what was going on was we had these ideas from our PhD research. We wanted to put them in a box, make it easy for anyone to build a big Wi-Fi network. So that was like product or technology first. Along the Meraki journey, we actually got to engage in a ton of different industries. So we got to see education. We also got to see retail. We got to see healthcare. And some of the most interesting customers we got to spend time with were industrials or manufacturing companies, people in operations because, you know, really you don't tend to get to see behind the curtain that often. How does a warehouse run or how does a manufacturing facility run? But they all needed Wi-Fi. So we got to see those up close. And it had always struck us that it was like stepping in a time machine. Like this is back in 2015ish. At that point smartphones had happened. The cloud wave had already happened, but you'd kind of walk through the door of a warehouse and you'd see, wow, there's like pen and paper clipboards everywhere. There are maps on the wall for people like planning their routes. Like it just didn't feel modern. And so that's where we said there's an opportunity here to have impact because that's what we wanted to kind of think about next was how can you be useful in the world.
So how did you figure out how to be useful?
Well, it was basically that okay, there's this underserved market. Let's go spend time with them. And our initial hypothesis was that, you know, maybe they need sensors, right? Like temperature sensors for their cold storage warehouses or for those sorts of facilities. So that's what we started with. We built that. It was the simplest kind of product we could get built. And the whole idea was to get close to the customer. We knew that it was probably not going to be the product, but it was essentially an excuse for us to talk to real people. It became very clear to us that their warehouses were not having refrigeration failures that often. It was 2015. That stuff had gotten really good. But the weakest link for them was actually understanding their operations. So they had no idea where their drivers were. They had no idea if stuff was kept cold, you know, that whole time, etc. And when they showed us the tools they were using, I remember vividly one of the customers had an interface and it was running on MapQuest. And it's like a kind of throwback to the '90s. But that was happening in 2015.
So how would they run on MapQuest? How would that even work?
So MapQuest had an API of some kind. And there were what they called breadcrumbs. So you get an update on the map every somewhere between 5 to 15 minutes. And just think about how far your DoorDash order moves in like 15 minutes, right? You can go from like waiting to pick up to like it's here. Exactly. And that was the customer experience. So when we saw that, we said, "Okay, what's the more modern way to do this?" And so we started with real-time GPS tracking. It kind of felt obvious to us like that would be something useful for the customer. But we knew that once we did that, we'd be able to kind of piggyback off that use case and find some other adjacencies.
So how did you know that in terms of there were other? You were just like, "Oh, I know there's more opportunities here." Like was there something you were picking up on that you just kind of knew like okay maybe this isn't the thing or maybe the second or third product but there's like a huge opportunity.
Well, so you know, it started with GPS tracking and that was just the dot on the map, can we show them in real time where the driver is? Then route stop arrivals were another one and no one had done the integration of like well you have a planned set of route stops and arrival times, if you upload those into the system, you mash them up with the GPS, you can kind of see did someone arrive on time. That was a very manual process. So that was the immediate adjacency. After that, we started learning a lot about the efficiency of the operation in general, things like engine idling. So people with delivery vans will often just leave them running because it keeps the air conditioning running. Turns out if you turn the engine off, that could save you 2 to 3%. We have customers that have $40 million fuel bills. So if you save them 2 or 3% that's like a million dollar savings. So all of a sudden the customers started coming to us with ideas saying is this something you could help with or we've been trying to figure this one out like is this something that technology can do?
Was there specifically in terms of like adding more products? Like how did you figure out what to do in what order? Or like was it a customer pull thing or were you just like trying and they took certain things and didn't take certain things? How'd you figure it out?
Well, for us because we were outsiders to the world of operations, we were always kind of we called it big eyes and big ears. Like we were always like looking to learn how does that operation run or what's on your mind or what's that binder for? You know, this thing on your wall, what are you tracking? And one of the areas that we started getting a lot of pull on was around safety and it wasn't even articulated that way. We actually started getting asked from our early customers, "We love your GPS tracking system, is there a dash camera that you recommend that integrates well with Samsara?" And what was going on was at the time people were buying consumer dash cams like the stuff you'd get at Best Buy or maybe you've seen in Uber, right? And it would record video to an SD card. So, it's basically a GoPro that you stick in the dash that would record the driver. Well, it was actually about recording what was going on outside. And what was going on really was in the US at least, if you're a large corporation and you're making a delivery and there's an accident, typically it gets blamed on the company, right? Like because deep pockets, insurance claims, all that kind of stuff. So, they were using it to exonerate themselves and say, "No, this is really what happened." That happens way more often than you might realize that companies are sort of wrongfully accused. So they would settle or there was no eyewitness or something like that. So they were buying dash cameras for that reason. They were kind of a bear to deal with because you'd have to pull the SD card. Maybe it wasn't plugged in. Maybe that operation's like 400 miles away and someone's got to FedEx you the card. It just wasn't really practical.
So what ended up becoming our second big product family was basically from customer asks of is there someone you recommend? And you hear that like three or four or five times and you say well if we built one would that be interesting to you and they were all willing to try it out. And so our very first dash camera was a prototype. It was basically a webcam that we bought off Amazon plugged into the USB port. I think my co-founder John hacked some software together over the weekend and we took it back to a customer the following Monday and he said, "This is amazing because now we don't need to pull the SD card, right? Like we can pull the footage remotely." So, it was an idea like that that has now become like an over half billion dollar product line.
Oh, wow. So, how was there a moment then when you knew that you kind of had product market fit and like you knew that this could be a company worth, you know, it was your second company. I'm assuming you could kind of you knew what you were doing, a little more experience. You could kind of be a little bit picky like you knew this was the big opportunity.
Well, for us, we didn't want to rush things because that sort of process of finding product market fit, it's really hard to do like in a kind of strategy session or whiteboard like you just need to make contact with the market. What was very clear to us and like I said the first product the beta product we built was the temperature sensor that was not taking off in the sense of like engagement on the product like people weren't using a lot of the graphs and things like that. What was taking off was the GPS tracking so people wanted to know where that truck was with the temperature sensor. And then we started getting asked a question like hey this beta product is it a secret or can we tell other people about it? We're like oh no tell other folks about it. What do you want to tell them about? It was the GPS tracking part of it. So when you hear that, that's like the first sign. Then you need to see like, okay, is this going to be a good business for us? Like, can we make the margins stable and really fund the entire company over the long term? And once we started doing the math on that, we said, "Yeah, this could definitely work."
How would you recommend that part of it, like the commercial, like is this a good commercial idea? Because I feel like there's a lot of times where you might build a cool product, but like when you really do the math, it's like it might be tough to make a business here.
Yeah. And I think this is where it being the second time through company made a big difference because we had an idea of what should a gross margin look like. And you know, I think at first, especially in our first company, we underpriced a product massively because we didn't understand how much it would cost to fund the rest of the business. It's like you know how much the hardware costs and you know how much maybe the software the cloud service costs to run but you don't think about sales commissions and support costs and all the other stuff that needs to go into making something sustainable and scalable over the long term. So if I were to come to you and tell you I have this idea and here's the idea, what in your mind maybe I think this probably is going to change over time but like if I'm first starting something what is probably a good gross margin to think about and then some of the other pieces like expenses that go into those unit initial unit economics like what would you kind of tell me is probably a good benchmark to clear like okay you can probably do something with this. Well, it kind of depends on whether the product needs to be sold or not. Like for us, we knew that in operations at least, people were going to want to have a conversation about it. They were not shopping for technology. They weren't going to download it. Like I feel like with a lot of open source software, you have people just doing product-led growth where they discover it. Your cost of sale might be very different. This was a little bit more of, hey, we're going to need to call people up, do things like free trials so they can check it out. And so if you had a business like that, I would say you need to have a gross margin that's north of 60%. Because your cost of sale might be 40% or something like that. And you can figure this out by looking at public companies that are in these industries, maybe not at all the same technology, but you have a rough idea of okay, what does it take to serve this kind of customer.
Yeah, that's something that's always fun for me is I just Google name of the company and investor relations. And public companies, they share a lot of information. You want to learn about them, just go to the page. They'll have nice presentations. They might have hundreds of pages, right? Each quarter, they put out earnings that you can go read. For a lot of people, like the first time you discover that, you're like, "Holy cow, it's like this company's sharing all their secrets."
Companies, it's all out there. Yeah, it's out there. And I think you can also learn a lot about an industry in terms of like, okay, why did they choose to focus on enterprise customers or why did they do SMB? So there's a lot of information like you said on those IR portals.
The investor days are often really good as well. And then for us the key was okay, is this an underserved market? Because you don't want to enter a market that's already got a lot of competition, customers are pretty happy with what they have. So I would say there's the gross margin side of things, which is could you make it sustainable? It's really hard to change the gross margin profile of a business massively over time. But then more importantly, can you be 10x better at something? And that's something that doesn't show up in the gross margin.
Yeah, that's fair. You might have like there's probably companies like Amazon doesn't have the highest gross margins on the Amazon e-commerce product, but there's extreme competitive advantages to just being like the lowest cost provider, right? No one can compete with you.
Yeah. Yeah. Exactly. And that's why it's like I think it differs industry to industry, product to product.
So, one question I had, I was just talking to someone who works with you right now. They told me to ask, what percentage of Samsara do you think was skill versus luck?
I think it's hard to put a percentage on a sort of thought experiment like that. I would say it was really both. Like we had to be lucky enough to kind of come across an industry that was underserved that is this important, right? This large and massive. But then to have the skill to execute on the hardware, the software, the sales, the support, like there's a lot that goes into serving these customers.
So maybe 100%, 100%. Like you kind of need both.
Yeah. Yeah. I've heard people say like a good way to define or maximize the opportunity for luck is like surface area, shots on goal, volume. Because I mean startups and any company is just like a series of experiments. You know, you're just so if you're constantly trying and doing something like if you take one shot and you miss, right, you're out of the game, but if you have 100 shots, thousand shots, eventually the 999th one it might finally hit and you got lucky but you were disciplined enough or skilled enough to take enough shots to get lucky.
Yeah. And that's also why it's hard to say we're just going to start scaling the company. You have to go find that product market fit. Usually that's a feedback loop process unless you've been in that industry or you're like a technical expert and just kind of know. But what we found again with Samsara, we had not spent a lot of time in physical operations. We were like admirers of it and kind of curious nerds. Like we were reading all these books and stuff, but you really had to just go get involved. And once you do that, that's when you start taking those shots. And it's just a bunch of conversations or at least that's what it was for us.
And then what was I know kind of the early days you had initially there was kind of like a new regulation that came out and like gave you guys a really big boost in like kind of the first year, the first year of like selling a product. Kind of what was that and how did that play out?
Yeah. So maybe for a little context, there was something called the ELD mandate. And that basically meant for folks driving long haul or heavy trucks that there were a certain number of hours they were able to drive per day before they had to kind of clock out or stop. And that was basically to help manage fatigue on the roads. In the US that had been done by pen and paper for decades and ELD was electronic logging device. So it was basically let's digitize that because it was easy for people to lose their logs. You know the paper would get lost, all that kind of stuff.
You probably fudge the numbers a little.
You could probably fudge the numbers. And you know, I would say the majority of the industry was compliant, but it was a real pain to deal with all that paper especially if you've got thousands of drivers and having to hang on to like piece of paper.
So that was a mandate that had kind of come up a few times and then not gotten passed. It ultimately did get passed in 2017. So it's technically the second year of the company that we were selling product. We actually didn't know about it when we started. We didn't know about it for the first year and then customers were asking like hey are you guys going to build an ELD. So this kind of is a repeating story, right? But for us, you know, just being customer led in that sense we said well help us understand like how does this affect you, etc. And they were saying, well we need to comply with it, but it also means we're going to need to put devices in the cab, which meant there would need to be an app. And that felt like something that we could do a really good job with, which is building modern software that would be part of that frontline driver experience.
So, you have this concept of customer feedback loops. How exactly is that? Is that an example of it or how does that usually play out?
Yeah, very much. So because this worked for us so well in the early days of the company, we made it kind of fundamental to how we operate. We run feedback loops in sort of formal and informal ways. So the formal way would be we have a big customer conference every year. It's called Samsara Beyond. So we'll have a lot of customers come in person. We'll take a whole day where we're like sitting around in groups like workshopping ideas, showing them prototypes, things like that. That's something we're able to do now at our scale, but in the early days we were actually just bringing those prototypes with us to a customer meeting or demoing it on the screen. So we still try to do that throughout the year. And what's great is as your customer base grows, the amount of feedback you can get grows too. So if we want to learn about the construction industry or the waste management industry or something a little more esoteric, it's really easy for us to get feedback in those areas now because we have tens of thousands of customers.
Is there a good way to sense out the quality of the feedback or what's worth listening to versus not?
Well, you kind of want to look for patterns. So, you know, it's kind of like an 80/20. Is this going to be relevant to the majority of our customers at least in this market segment? And if it is, then it's kind of worth investing in. It's worth building. If it's in the 20, it's not that it's a bad idea. It might be more of we should integrate with somebody or, you know, we should help you write software against the open API to do that.
Another thing that came up when I was talking with some people at Samsara, they said that out of any CEO they've ever worked with, you spend the most time talking to customers. Which it's kind of funny that maybe some CEOs are known for not talking to customers. How do you think you're able to do that so well? Or like if I'm somebody who's like, I want to spend more time with customers, but then also not lose focus the rest of the process of running a company, like how do you do it?
Yeah. I think you have to make time to spend time with the customer. For me, it's I'm always curious, right? And one of the most fun parts of my job is getting to go on site with customers because they'll give you a tour. They'll show you the operation.
And you can see everything from, you know, we were talking about autonomy and robots, like you see warehouse robots moving around to you see how a big waste management facility works like serving an entire city. There's really fascinating stuff.
What's that one? That actually sounds crazy. What is that?
Well, this was just a couple weeks ago. I was out in the UK and we got to visit, you know, we're in London and it's obviously a very very big city. We got to visit one of the regional waste processing facilities and you see the conveyor belt, you see the garbage trucks coming in, conveyor belts, all the sorting happening, all the recycling, both a mix of manual and automatic. A lot more automation than you might realize. Because they're using things like computer vision to go figure out, okay, this is recyclable or this is not and a little arm pops out and sucks the little can over. So things like that you don't get to see. And then at the end of the process, a lot of the stuff gets compacted and it's like Wall-E, if you remember that movie from Pixar. And just like huge stacks all the way to the ceiling. You wouldn't get to see that or you know you could maybe find it on YouTube if you really dug for it but that's the kind of stuff that I love about the customer visit and in that process you get to learn a lot about how often do garbage trucks break, right? Turns out pretty often because they're lifting tons and tons of material hundreds of times a day, right? So that's the kind of stuff where it's real world problems they just become really apparent once you set foot on site.
Yeah, like I'll have my garbage, I'll be pushing out to the curb and I'm like this thing is 100 lbs and the arm has to be designed to at that weird angle, you know, like pick it up and tilt it into and then just do it over and over and over again.
Yeah. Exactly. So there's all of that and then I learned about a lot of other interesting challenges like every once in a while someone will say you skipped my stop, right? And so that'll be a pretty common customer complaint but they're having to make thousands of stops on that kind of daily service route so statistically that means like 10 people will say you need to come back out and pick up my bins. So things like that make a big difference just to hear the stories.
Is there if I'm someone who wants to talk to customers more and I think like I might not have the time or, you know, how would you recommend getting myself in front of them more often if I'm like a product leader or founder?
Well, first just remember you can just do things, right? Like, you know, pick up the phone or jump on that Zoom call or even better go on site if you can. I think it's a really good use of time and it's worth kind of prioritizing. So, one of the things I've realized is in the same way that people budget money, you need to budget time, especially as a founder or a CEO, and be deliberate about where is this time investment going. And I think understanding your customers, there's kind of nothing like it, right? Like you have to go spend the time. You can't do it by proxy.
Whereas like you could probably always spend more time in some of these other functions. So you have to just be deliberate about the allocation.
Yeah. Any other interesting customer visits that you've done recently?
What's been maybe one of the most interesting or you know you went somewhere and like the thing just blew your mind or you learned like a new fact or statistic where it just changed the way you thought about something.
Well last year we have some customers in the airline industry. So they take you on the kind of behind the scenes tour of how these airlines operate.
And so we got to go, we were out at Dallas Fort Worth airport and they took us out onto the ramp or the tarmac and you get down there and you realize whoa this place is huge. And so we asked like how big is this? And they said well DFW airport is larger than the island of Manhattan. Like so things are big in Texas. This is really really big. And it was also a reminder of the conditions. Like it's really hot out there and there's the Texas heat but there's no air conditioning. It's like all reflecting off the ground and these folks are working 8, 10, 12 hour days, right? So just getting to see that up close.
There's no way they let you just like run around, you know, the airfield.
Yeah. But then we got to see the kinds of equipment they're operating and we learned all about how the carts that carry the baggage, they actually end up migrating to one side of the airport over the course of the day. Flights come in and it's kind of like shopping carts or something. They just move around and you can't actually service the plane unless you got the baggage carts ready to go. They would spend hours searching or hunting them down, making phone calls, trying to find them.
So stuff like that you don't find out unless you go on site.
Interesting. Yeah. I think the craziest like airport statistic was I believe it's the Denver airport. If you just kind of overlay it on a map over top of San Francisco, it's like bigger than like greater San Francisco. Airport's bigger than the cities.
Yeah. So but I guess like when you really think about airports, they're like and flying just like the aviation industry. It's like a modern marvel of the world is like we've just got hundreds or thousands of planes just like every minute or two flying in and taking off at places. It's pretty cool.
And it's just this huge complex operation.
So this is kind of going back to if you're curious, you can just spend all day like asking questions and be like, how much does that tire cost? It's like whoa $5,000 that's expensive, right? Like so those are the kinds of things that I think bring us a lot of energy which is just like how do we help solve more real world problems.
And you have this framework that's called the 70/20/10 framework for allocating capital.
70/20/10 hopefully not 7/3.
Yeah, you need that extra 63.
So what is the 70/20/10 framework?
That's related to kind of time horizon. So if you think about, one of the jobs of founder or CEO is capital allocation. Like what are we investing in and what are we investing for? On one hand, you could invest entirely for the long term and place all of your bets on things that are speculative and way out there, but then you're probably not fixing the bugs in your products for your customers or making enhancements and things like that. So, one of the things we noticed as we got larger was we had to come up with a split, right? Because we wanted to work on interesting new long-term ideas and now it's experimental ideas related to AI and things that are moving really quickly, but at the same time, you probably have a pretty big customer base that you want to make enhancements for and kind of bug fixes and that sort of thing. So what we kind of found is 70/20/10 fits. If you say like what's happening on the road map for the coming year or coming two years like near-term road map, then what do we want to allocate to for the medium-term road map and that might be new geographies or new market segments or strata. And then what are some ideas that are high risk but also high reward like if this cracks open this could be a really big product line and that would be in the 10 kind of category.
And this is 10% of capital, time? Like what do those numbers mean exactly?
Yeah, it would be more related to operating budgets.
That would be specifically related to R&D spend. So, if you just look at the R&D budget and where does it go? That's kind of how we think about it. If you look at the company, there are other teams. There's sales and support and finance and all that kind of stuff. So, those are less time horizon based. It's really R&D that we're talking about.
Okay. Well, and talking about R&D, so you have this theory of like innovation in concentric circles. How does that relate to all this?
Yeah. So, we were talking about customer feedback loops earlier. What we kind of discovered over the years was once you have that foothold, once you're talking to the customer about something specific, they tend to open up about the other problems that they're dealing with, right? These may be other projects on their plate or other expenses or costs they're trying to fight. And those become kind of obvious from the pattern perspective, the 80/20 test, those are almost always immediately adjacent to what you're working on because you're talking to the same user or same buyer. And if you kind of were to write them out on a whiteboard, you could almost just draw circles around them. You're like, okay, this is an idea that's really close to what we're doing today. This is the next layer of ideas, and these are immediately adjacent. And then you can kind of draw them further and further out. So we found that strategy works a lot better than let's build this product for this customer here and build this entirely different product for a different industry or different user. You don't get the same leverage of the relationships and your sales team and all that kind of stuff.
And maybe this is a rhetorical question but getting you to keep talking like why is that important that they're close?
Well, it really is about getting leverage, right? So most of our customers, for example, they don't just buy one product. They actually license many different products on the platform dependent on their operation. So if you're an airline customer, you might have a lot of equipment that you're trying to track alongside your trucks and things like that. And it's the same sale, right? So when you're having the conversation, you're like, this is what we would recommend to solve all of these problems. So the effect compounds. If you have like a bunch of disparate products that are either completely unrelated or even sold to different customers, you're now acquiring entirely new customers each time having entirely different conversations and maybe you can kind of operate as a holding company but it's not as efficient as if you're already there and having the conversation.
Yeah. Talking about margins and being profitable earlier, like that's ultimately like I think what makes companies sustainable. By the way, is, you know, you got to think through the margins, but you have to think about the operating costs and then can this thing run on its own? Like it sounds like common sense for a business but with venture I think there is often a tendency to just focus on the ideas and we'll figure it out later. You have to have some way to figure it out because later eventually happens if you're lucky. Do you, is, when you say like later, like is there a cutoff where you feel like too far in the future and like, and maybe what you do too with Samsara, like how far ahead do you plan? Like is 10 years in the future just we can't allocate to that yet? Like is there when you're thinking about that time horizon where there's like an era of importance and when you can or can't do something? Like how do you usually think about that?
Well, I think on the financial and margin side we have to see a pathway where it makes sense on a whiteboard or a piece of paper of like okay this kind of pencils out, like there's a way here. You may not be there in day one or year one and that might be you're going to get some benefits to scale or when it comes to AI for example we know that model costs are getting more affordable, like cost per token's coming down so you can say okay there's a trend line here and so we're going to intersect with that. That's more kind of through a financial lens. On the technology side of things, it's pretty hard to predict 10 years out, right? Like I don't know if you saw ChatGPT coming 10 years ago but that was really hard to predict.
I didn't see it coming 3 years ago.
Exactly. And so those are the kinds of things where you want to be dynamic. So we'll look for long-term technology trends. That kind of goes back to like a Moore's Law kind of cost curve or the cost of bandwidth is coming down. Those things you can kind of bet on and see around the corner. But every once in a while you get like a big tectonic shift and that was like mobile phones back in the 2000s or LLMs. Now, that's where you just have to hit refresh and start hitting refresh pretty often.
How do you know like maybe with, maybe that's obvious like with ChatGPT like everyone describes it as the ChatGPT moment. But how do you know if there's a moment that's worth refreshing on or kind of like taking a step back and like maybe we should change something or...
Well, I think this is more art than science, but ChatGPT came out in like November of '23. It was like Thanksgiving if I remember correctly.
And at first it seemed like a really cool novelty. But it's when you know something's going to be useful, right? And if you remember those early models, they hallucinated a lot. They weren't super up to date and things like that.
Yeah, there was a cutoff, right? There was a knowledge cutoff.
And it's not actually clear that those models were able to reason either the way that the newer ones do, but you could kind of see the potential. And so this is where actually being a tinkerer helps, right? Like you kind of use it and you see okay this is how it could be really useful. And I wouldn't say that we upended our road map at that point but it was like okay we need to start playing with this and that's like I said earlier we were a very early enterprise adopter of OpenAI. Like we started working with them, it was because of that intuition that we developed from being kind of in the details.
Yeah. So a really interesting thing I wanted to I guess get your thoughts on or opinion on and I read that Samsara is the that I know of the only company that's doing over a billion in ARR that's growing over 30% a year that also has positive free cash flow. I mean that's pretty impressive like you think of the universe of companies that exist that could do that. Like how did you get so good at just focusing on growing quickly, doing sales, being profitable? Like any specific lessons or takeaways?
