Why Revenue Scale and Operating Leverage Are Different for Software vs. AI. (Tech Strategy – Podcast 69)


This week’s podcast is about revenue scale and operating leverage in traditional companies vs. software vs. AI. And these things can be really different. Especially software vs. AI.

You can listen to this podcast here or at iTunesGoogle Podcasts and Himalaya.

Here are the three factors I mentioned that interact for operating leverage:

  • Revenue
  • Operating profits
  • Economic value creation (i.e., vs capital)

Here are the books I mentioned:

Related podcasts and articles are:

  • N/A

From the Concept Library, concepts for this article are:

  • Valuation (Question 3): Operating Leverage
  • Valuation (Question 3): Revenue Scale and Growth
  • SMILE Marathon: AI/ ML

From the Company Library, companies for this article are:

  • N/A
This is part of Learning Goals: Level 7, with a focus on:
  • 34: 9 Questions


I write, speak and consult about how to win (and not lose) in digital strategy and transformation.

I am the founder of TechMoat Consulting, a boutique consulting firm that helps retailers, brands, and technology companies exploit digital change to grow faster, innovate better and build digital moats. Get in touch here.

My book series Moats and Marathons is one-of-a-kind framework for building and measuring competitive advantages in digital businesses.

Note: This content (articles, podcasts, website info) is not investment advice. The information and opinions from me and any guests may be incorrect. The numbers and information may be wrong. The views expressed may no longer be relevant or accurate. Investing is risky. Do your own research.

—–Transcription Below

Welcome, welcome everybody. My name is Jeff Towson and this is Tech Strategy. And the question topic for today, why revenue scale and operating leverage are different in AI versus software? Now that’s a bit of a dry topic. So why is this important? Because this is about valuation of digital companies versus traditional companies and valuations of. Companies that are more traditional, they’re starting to be more and more software infused, AI infused. And these metrics do tend to be different, depending if you’re talking about a more traditional business, physical products, goods and services, versus a software based business, versus an AI versus business. They are different in all three of those. So I’m gonna go into kind of, sort of how to think about those. Now in… tech news in this part of the world. Yet another week, yet another mega tech IPO out of Asia. Coupang, the Amazon of South Korea, so they say, has filed for IPO. The numbers are out. You can pull the numbers. They’re really interesting. Probably 50 plus billion dollar valuation. Who knows? I mean, these things take off. Last week it was Quaishou. We got more in the pipeline. It is just all the time. And what’s fascinating is there’s just a huge pipeline of these companies. The vast majority, I dare to say, that people out of the West don’t know. They do become familiar when they go public, but prior to that, generally not. And there’s a lot of them and they all came to, they all sort of ring the bell of, well, this is the biggest IPO of this year, right? It keeps happening. So I’ll be talking about coupons shortly. And. not in this podcast, but in the next couple of days. For those of you who are subscribers, I’m gonna send you out some data, sort of strategic frameworks around a couple of these companies starting tonight. One will be Zhong An, which is sort of an online insurance company out of China. It’s been public for a couple of years. Right after that, I’m gonna send you a Baicu. which is a housing online, online merge, offline housing, rental and sale platform out of China. And then Dada Nexus, which is a sort of on-demand delivery, almost infrastructure play. So I’m gonna send you those in the next couple of days. But yeah, there’s a lot of these companies coming. Honestly, I’m finding it hard to keep up. I’ve got a stack of S1s on my desk. They just keep rolling out. I’m working through them as fast as I can, but they keep coming. You know, so all of us who sort of have focused our career on Asia tech companies, South Korea, China, Southeast Asia, not usually Japan. You know, we’ve picked the right place, man. This is, it is where things are happening. And I can only imagine what it’s going to be like. Keep in mind 10 years ago that were virtually none of these. And now it’s just all the time and they’re bigger and bigger. They keep sort of hitting all the biggest numbers. Imagine what it’s going to be in five years from now. It’s kind of an awesome place to be. It’s a lot of fun. Okay, let me get into the content. But first, my standard disclaimer that nothing in this podcast or in my writing or on my website is investment advice. The information and opinions from me and any guests may be incorrect. The numbers and information presented may be wrong. The views expressed may be incorrect or may no longer be relevant or accurate. Overall, investing is risky. This is not investment advice. Do your own research. And for those of you who’d like to be subscribers, you can go over to jeffthousand.com, sign up there, there’s a free 30 day trial. See what you think. If you’re not a subscriber and you’ve been listening for a while and you wanna maybe give a little something back, feel free to go over to iTunes or wherever you listen to this podcast and give a review. Doesn’t cost anything, it’s free, couple clicks. It’s maybe a nice way of sort of giving back a little bit if you feel so inclined. Okay. Now standard company analysis, we’re building a spreadsheet, we’re trying to value a company, you start with the top line and lo and behold everyone starts arguing growth because growth rates especially if you’re doing sort of a future cash flow and discounting it back, it’s the major factor. You tweak the growth a little bit, no the numbers look great or the numbers look bad. It’s a big lever when you’re sort of projecting things out. People tend to be over optimistic, pretty common. I’ve been using the word revenue scale, which is, you know, let’s say we’re selling pens or Coca-Cola’s, okay, number of units times price, what’s the size of the revenue, how fast is it growing, what’s the overall scale this can be, is this going to be a local business, national business, international business? But you know, one of the things that happens when you start going from a business where your primary activity is the transaction. I’m selling cans of Coke. the activity, the demand is equivalent to sales and revenue. There’s no difference. That is the activity we’re in. But as we start moving into a more digital world, those things start to broaden out. And the example I’ve been giving quite a few weeks now is Nike China, which has great apps, Nike Running Club, Nike Training Club, or something like that, community activities. There’s a lot they’re doing in what we would call demand side activities that are much bigger or beyond a transaction. It’s like a transaction is a small monetization aspect of a larger type of demand activity. And it’s really easy to do this because the stuff they’re doing, the apps and the content and the community, you know, if that’s digital, it probably doesn’t have much of a cost structure. So it’s, you know, it’s this idea of like, if you talk to something like Alibaba, they will talk about… the economy they’re building, the platform they’re building, the ecosystem they’re building, which is a lot of people talking, chatting, doing all sorts of activities, and then they’re only monetizing a small percentage of that activity. So their revenue line is a fraction of the overall demand side activity. So, you can view this as sales, what’s the sales of the company, what’s the growth rate. I don’t think that really helps very much, especially when you start moving into platforms, you wanna start talking about at least revenue, scale, and scale is a better word, I think. Or I tend to use demand side scale. And that’s what I’ve been putting in these PowerPoints with my little blue diamonds all the time. I’m always talking about demand side scale, users, activity, engagement, data. And then number four or five on that list is money. That would be the revenue and sales. So there’s this idea of when we talk about these models, how do you project that forward? As opposed to, hey, how many cans of Coke are we gonna sell next year? So revenue scale, demand side scale, how do we project that forward? And as you move, let’s say from a traditional company, we’re selling a product or we’re selling a service that is made of something physical. It’s made of molecules and atoms. I don’t know why I keep saying cans of Coke, Nike shoes, service agreement, consultants, things like that, that are physical products and services. You know, we can grow somewhat. We are often bounded by geography because it’s hard to move people and products across borders. Or maybe they’re just heavy. I mean, you can’t ship a can of Coke very far before the cost of the transport is greater than the price. So there are limitations on geography when we start doing physical things. Even bankers fly all over the region on a daily basis, but they don’t generally fly internationally every day. You have to start thinking about capital requirements. Am I gonna have to open 10 more Walmarts to sell more stuff? Well, that’s a lot of money. There’s a lot of capital in Walmarts. I have to increase the human capital, the number of bankers we have. What about route to market? Do I need to be in every store? You know, generally speaking, Physical products and services have an ability to grow that can have varying degrees of difficulty. And a lot of the most successful sort of traditional companies have sort of found a way to hack that. So McDonald’s got around all of that by doing franchises. Yes, we’re opening restaurants, but we’re franchising them out. So that makes the growth aspect a lot easier in terms of capital and other things. Movie theaters are like that. You know, it used to be if you wanted to see a play, well, every theater had a, you know, every town had a theater troupe. Every town had a theater and you would go down and that’s where the theater group would live and perform and so on. Or maybe you’d have a traveling circus that would come to town or a traveling theater group or something like that. Okay, movie theaters and movies on film sort of half hacked that situation where, okay, now the movie at least we can put on a truck. because it’s surreal and ship it around the world, but we still have to have theaters town by town. So we’re still mostly in the physical business, but we’ve hacked part of it. Not unlike how McDonald’s basically did the same thing. Okay, once we move to digital products and services, goods, then everything gets a lot better in terms of the growth side equation. You know, let’s say comparing, you know, traditional theaters to movie theaters to Netflix. Okay, Netflix is a digital good. It scales beautifully. Zero marginal production costs. It has more or less zero distribution costs, although they have to work out with the broadband carriers, some fees and stuff. And they can go globally like it’s nothing. So their growth ability as a purely digital company, their revenue projection is dramatically bigger. Right, not that everyone knows this. I’m not telling you anything you don’t already know. I’m sort of teeing you up to the third point, which is. That scenario gets even better when you switch from a digital service or product to a platform. Because once you move to a platform, you’re going from Netflix to YouTube, from Netflix to TikTok, where not only do you have the sort of zero marginal production costs, very low distribution costs and things like that, you’re not even creating the content because you’re using user-generated content, and platforms can scale faster than anything. They’re unbelievable at this. They can tick-tock and roll across the world in 18 months, 24 months, because of the production costs, the distribution costs, and because everyone in every country is creating the content for them. Their capital costs are almost negligible. So they’re sort of, when we start talking about, okay, revenue scale, demand side scale, you know, I put it in three buckets in my mind, traditional products and services, physical things. Okay, maybe four buckets. Second bucket would be physical products and service with some sort of hack, like a franchise or it’s part digital. Number three, digital goods and services, Netflix. And the number four, platforms, digital platforms to be specific. That’s kind of how I think about revenue side scale. And there was a great book about this that didn’t really get a lot of attention. It was a book by Sanjit Gopal. Chaudhary, God I’ve got to remember how to say his name correctly. I’m sorry Sanjit if you’re listening to this. I’ve talked about his book, Platform Revolution, which he wrote with a couple guys from MIT. Very successful, everybody knows this book. I’ll put the link below if you haven’t heard about it. You know, he’s one of, if not, you know, he’s in the short list of platform thinkers in the world because of this. But before he wrote that book, he wrote a book called Platform Scale, where when he started thinking, and this was just him, And from my take on this, this was when he was really digging into platforms for the first time, and he wasn’t thinking so much about the network effects, and he wasn’t thinking about a lot of the aspects talked about in platform revolution. The number one question I believe he was focused on is their ability to scale, that platforms can scale faster than any other type of business model, hence the title platform scale. I’ll put the link in the show notes for both of those. And I think That’s a question we don’t talk that much about. That’s pretty important. So, you know, the point of this podcast is, how do you think about revenue scale, demand side scale? Why is it different for traditional versus software versus platforms versus AI? And I’ve sort of given you the first three. It’s the fourth one that I’m struggling with right now. This is why I’m thinking about this, which is when people talk about digital, oh digital scales. They’re usually talking about software. They’re talking about things that are done by software engineers where you kind of design the machine like you’re making a robot or a car engine. It’s a human activity where we’re designing something that will run. You know, people call software invisible engines. When you move to AI, depending what type of AI you’re doing, you may well be doing a similar thing like you have a world you’ve created, like the Netflix content library, and you’re using AI to try and predict what most people will wanna see next. Okay, so you’re using prediction, AI is cheap prediction, but you’re doing it on a system that has been sort of human-made, and therefore is very orderly, has rules, it’s very defined. You could say the same thing about using AI within, say, Amazon logistics hubs, where you’ve got a… warehouses that you have designed. You’ve put the sort of lines on the ground, the robots follow the lines, you’ve put the cameras, you control the environment, so the AI is making predictions against a limited number of controlled variables that you have created this game. That’s one type of AI. The other type, which is what most people are talking about most of the time, is this idea where you’re using AI, cheap prediction. to respond to the real world that we don’t control, where the car is going down the road and a rabbit runs across the street, what does the AI do then? And we don’t control that world. And when you start talking about that world, you’re no longer talking about engineering, where you’ve built something that you control. That’s a lot more like doing science research, where we’re studying some aspect of nature, or the human body, or biology, or chemistry, that… the external reality has created and evolved and we’re just trying to understand it and it’s far more complicated than anything we create. This is why if you’re an engineer you can get a lot done in the life. You know you can design rockets and stuff. But if you try and create a new drug that’s going to do something in the human body, I mean it’s good luck with that. They almost all fail. I mean it’s really really difficult to do science on the physical world. or human biology or something like that. It’s far easier to write software or to do underwriting stuff. Or these are all human manmade games as opposed to a very complicated, messy reality. Okay, when we start applying AI to a messy reality, what’s gonna happen with the weather? How is this car gonna go down the road? What happens if a little kid runs out in the road? Something like that. It’s very, very difficult and it’s not clear. that it scales at all in the way that we think software scales. And people tend to equate these. They say, oh, it’s all digital. It’s software. It’s AI. No, no, no, no, no. Software and AR are very different things, and they have different economics. And that’s becoming more and more clear by people who have launched these companies. There’s Andreessen Horowitz, the big venture capital firm in Silicon Valley. They’ve been writing some good stuff about this where they’ve invested in a lot of the early AI driven companies and they’re like, look, the economics of these things are different. They are just different. Their argument is it looks like the economics are a combination of services and digital, not just software. Let me rephrase that. The economics are a combination of software and services, not just software. because a lot of people like to start software companies because they’re on, they grow so quick and they’re great, it’s Microsoft operating system. Now this is a lot messier. And it’s not clear that they scale. It’s not clear you can take Netflix from California and then show it to people in Florida and it’s gonna work fine. If you have trained your AI to drive around the streets of San Francisco, it’s not clear that that AI is gonna do anything nearly as well in Italy or Japan. or Thailand, you know, AI may end up being very local for a lot of this stuff, which means you’re gonna have to spend a lot of time training data. You have to get the training data, you have to sort of train the algorithms. They may be very specific for each region, for each activity, for everything. And they also may decay in their usefulness very quickly because the world outside, yeah, it was good for predicting crop patterns in northern Thailand for six months. But then the weather changed and nature changed and now it’s giving us the wrong answer. So we have to get a whole lot more training data. We have to train the algorithms again. Maybe that their effectiveness decays fairly quickly and whenever you sort of do that training exercise, it usually takes a lot of people. You’ve got to grab all the data, you’ve got to clean the data, you’ve got to label it, you’ve got to retrain the AI. That’s a lot of human activity by very expensive people. Then the algorithms work pretty good for a while, but maybe they decay. and then a year later they don’t work very well. So the whole scaling question appears to be very different depending what you’re doing. That’s sort of point number one for today. Revenue scale, demand side scale, you wanna think about traditional products and services, software and AI as different when you’re starting to map these things out. The other idea for today is operating leverage, which. I think it’s super important. It’s on my very short list of numbers I’m always trying to figure out for companies. Doesn’t get talked about nearly as much. Everyone talks about growth rates of the top line, and everyone talks about operating margin, but somehow on that list, like operating leverage gets left off, and I think it’s arguably one of the most important numbers. So I wanna talk about that a little bit, and then how it’s different on software versus AI versus traditional. which is another sort of interesting complexity. Okay, for those of you who aren’t familiar, operating leverage, I mean, it’s just like financial leverage. You know, you can buy your house, I don’t know, $100,000, and you can pay all in cash, so it’s 100% equity, no financial leverage, or you can buy it with 50% debt, so 50% cash, 50% loan. and you’re levered up 50% and then what happens? Well, if the house goes up, let’s say 10 or 20% from 100,000 to $120,000, if you weren’t levered, you know, you made $20,000, so 20% return. And if you had taken 50% financial leverage, obviously the returns are much greater. Don’t know what they are off the top of my head, but you can do the math there. It’s basically the same idea. It’s as a business grows, which is revenue, not sales, but revenue. As the revenue grows, how does the operating margin increase with increased revenue? So it’s the rate of change of one versus another. If you have a company and it’s making $100,000 and let’s say your operating margin is $20,000, when you go up to a $120,000 revenue or sales, at that number, what is your operating margin? Is it still 20% or did it go up to 25%? Did it go down a little bit? How does the operating margin as a percent change as the revenue goes up and down? And people talk about this for software digital because obviously everyone always wants the scenario where you have a fixed cost, the company starts growing and suddenly the profits start increasing dramatically because it’s got a lot of operating leverage because most of the costs are fixed. Now in the example I gave, if your business went from $100,000 to $120,000 in sales and your operating profit went from $20,000 to $24,000, you’re basically still at 20%, which tells you, okay, this company is almost entirely variable costs. Because the operating margin went right up, you know, same percentage as the sales grew, as opposed to getting bigger or getting smaller. Operating leverage of one. So here’s a… People say this is a measure of how sensitive profit is to change in revenue. And I gave you the one about, you know, it’s a simple business. You went from a hundred to 120, but there’s other ways to think about it. Okay. What if your business goes from one city to another city? Yes, the sales went up, but it’s also kind of a different scenario. You could have a lower or higher margin operating business in Chiang Mai versus Bangkok so as your revenue grew you could see very different operating margins. So it can change with geography It can change with various different customer tranches It contains change with customer types It can change, you know If you get a greater density of customers within a certain geography like a city then you can see a change So it can change in a lot of ways beyond just the fact that look we have one core product and now we’re a little bigger and our leverage increase because we have so many fixed costs versus variable costs. Now there’s a lot more variables that can be under that beyond the just basic fixed cost versus variable cost scenario and that’s really where people tend to get tripped up. As I say, oh we have a great business we’re expanding from one city to another city and the operating margin changes. You could say, oh we have one product and now we’re tying a second product to it and the margin and maybe they’re tied together, things like that. So there can be a lot hidden within those changing numbers. But generally speaking, if the operating profit goes up faster than your revenue, then you have high operating leverage. If the operating profit is more or less immune to changes in your revenue, then you probably have low operating leverage. And the equation, I’ll put it in the show notes. Operating leverage equals percent change in operating profits divided by percent change in sales. Now the common sort of approach to this, is the whole, okay, let’s just look at the fixed cost versus variable cost. There’s your operating leverage. And that’s true. It’s just the most simple case and I don’t think it’s the particularly interesting one. It’s pretty much analogous to the example I just gave of buying a home with debt and leverage. I mean, it’s kind of like that. So you look at a business like, I don’t know, like a theme park, a hotel, luxury stores, things like that. And it’s like, okay, you’ve got a fairly fixed cost structure. When I used to run a hospital, I hated running the hospital because all virtually almost not all the costs, let’s say 70% of the costs were fixed. You couldn’t change anything. I mean, it’s like you couldn’t say, oh, we have low volume today. Let’s send the cardiologist home. Can’t do that. And say, oh, let’s not have the radiology techs here. Well, can’t do that. You basically have to keep the whole place ready for whatever walks in the door. And then your variable costs are more along the lines of consumables and drugs you actually use. But, you know, if your hospital is empty, you start losing money pretty quick. And hotels are like that. Theme parks are like that. I mean, Disneyland and stuff like that, like Disneyland Shanghai, Disneyland Hong Kong, when those things lose volume, they start to go into the red almost immediately. You can’t do much about them. I mean, it’s not like you can close off half the theme park. and say all those rides aren’t available, please pay us for a ticket. No, they want the whole park, generally speaking. So that’s something pretty good. You can think, okay, this company’s what percent fixed costs? 25% fixed cost versus 75% variable or the inverse. Fine. I don’t think that’s that much, I don’t think it’s the most interesting case. I think it’s more interesting when you start thinking about how the costs are gonna evolve with the company over time. The picture I gave you was kind of a static view of the company. Here’s what we sell, we’re a luxury hotel. If our revenue goes up, therefore our cost structure changes because this is our fixed versus variable. You know, most of the companies we’re looking at, they’re evolving very quickly along a lot of dimensions. And that’s more interesting. How is this gonna change over time? A lot of tech, you know, overall, all the costs decrease. The fixed costs and the variable costs tend to decrease over time because software keeps making things cheaper and cheaper. So if you’re doing a free, you know, a cashflow analysis, and you’re looking at a software company and you’re looking three to five years out, you have to assume, yes, there’s certain fixed and variable, but you also have to assume, look, they’re probably all going down. Maybe the people costs are gonna go up or your real estate in Palo Alto is gonna go up, but generally speaking, tech costs just keep dropping. Your server costs are going down. Your software licenses are going down. I mean, so it’s not fixed like a hotel. I mean, these lines tend to trend down. There’s also interesting scenarios when you look at these in good times versus bad times. If there’s a recession everybody tends to get lean and mean. You know this is what’s happening in Thailand right now with the tourism industry pretty much shut down. A lot of retailers, a lot of hotels are getting lean and mean. They’re cutting every single cost they can. They’re replacing people, they’re putting machines in, they’re doing a lot of stuff because they’re trying to survive for the most part. Generally, I think you get a more compelling picture of the true long-term cost structure if there’s some sort of short-term recession, because then you see what these companies look like when they’re lean and mean. And that’s probably what the future is gonna look like. If you’re in boom times and everyone’s growing fast, hey, we’re all making money, those cost structure numbers are probably not gonna be accurate long-term. Truth is the numbers are probably gonna look better than that long-term. probably lower costs, probably maybe not as aggressive marketing spend, things like that. So you wanna think about sort of good scenario, bad scenario, hyper competitive, not competitive, the evolution of the business model, the evolution of the product service mix, all of that stuff is gonna impact operating leverage. And I think that’s a lot more interesting than oh, your fixed costs are A and your variable costs are B, therefore you lever it up this much. I think you got the point on that one. Now that can be a little hard to get at, especially in a business that’s changing pretty quick. I tend to keep an eye on fixed assets as well because fixed costs and fixed assets generally go together. If you have 10 buildings, you’re gonna have a certain fixed cost associated with maintaining and running those buildings versus five buildings. That can give you kind of a good floor on what maybe your fixed cost structure is gonna be. And tangible fixed assets are easier, obviously. but you can also kind of get at the intangibles sometimes. So these things all sort of go together. Keep in mind, we’re really after three different numbers here at the end of the day. We’re after revenue, which is the growth question. I mean, we’re after operating profits, and then obviously operating leverage is a big part of that. And then ultimately we want to know economic value creation. We want to know does this company generate, preserve, or you know, wealth or not. And so you have to factor in the capital deployed and the cost of capital. So those fixed assets are important and ultimately that’s going to tell you the economic value and the share price. So it’s all three of those. And they all kind of relate to each other in some ways. I mean, operating profits obviously follows from revenue, but it also tends to follow a lot on, what fixed assets did you use here? And what are the fixed costs associated with them? So they all kind of tie together in various ways. Which brings us back to this whole software versus AI question, how is operating leverage different in this scenario? People love software, one because it scales, which I mentioned, but they love it because of the operating leverage. That’s really what people like. This idea of do we do things with people or do we do things with software? And it’s pretty much what it’s, you know, the term I like is digital agent versus a human agent. If you’re running a bank, you have a bunch of bank tellers. Those are your human agents. A company like Ant Financial comes along and says, we’re going to do the same business, but we’re going to use no human agents. We’re going to use quote unquote digital agents who can make decisions, who can make predictions and things like that. And you know, the economics look amazing. And that’s where this operational marathon, the smile marathon as relates to machine learning is so important. No, I put that as, I called it machine learning. I also called it an AI factory and I called it zero human operations. That once you remove the last person and it’s software that runs your core operation. you get an unbelievable amount of operating leverage. I mean, it’s pretty amazing what they can do. And that’s kind of what Ant Financial is doing. And this company I’m gonna send out tonight to the subscribers, Jongan, they’re trying to do the exact same thing in insurance, but not as successfully. So pure software companies tend to have pretty amazing operating leverage. A lot of the ones we see in China that we’re talking about now are what I’ve called digital physical hybrids, where yes, it’s a software company, it’s probably a platform, but it has a significant physical component. So Alibaba and Lazada, they have all these warehouses and all these people loading boxes and all. It’s a very physical, operationally somewhat intensive business underneath a digital software business. And sometimes that can… Actually, one, when you add a physical component, it really does help your defense ability. It can be probably your strongest moat long-term. That’s certainly the case for JD and Alibaba in China. I mean, try to replicate all these warehouses they’ve been building for the past 10 years. It’s almost impossible. And that tends to defend you against a lot of digital attackers. companies that are pure digital like TikTok or let’s say YouTube, their existence is a lot more volatile. One because you’re sort of selling to what people want to see right now as opposed to just buying staples. But two, they don’t have a physical component to protect them. So they’re very dependent usually on network effects to defend themselves. Okay, but we look at something like Alibaba and hey, I kind of like that you have all these physical assets and I like that not only do they defend you. They also kind of give you more operating leverage because a logistics network has significant leverage as well. It’s like you got digital leverage and you got physical operating leverage. I like both of those. As opposed to say, Maytwan, Grab, where you’ve got a ton of people riding around on scooters. Okay, we’ve gone from digital to physical, so you’ve got a much greater defendability, defensibility, because it’s hard to replicate all those riders. But those writers are mostly variable costs. So you’re not getting a lot of operating leverage on your physical component, even though they’re still getting it on their digital component. And that’s kind of Maytwan Gojek. Great operating leverage on the digital side, the physical side tends to be mostly variable costs, not awesome in that regard. But increases your defensibility, you know, everything in life is a trade off. Which brings us to AI. What is the operating leverage of AI versus a traditional software company? And as I kind of mentioned earlier, it looks like it’s a combination of software and services. So that’s not great. I mean, it’s closer to Meituan than it is to YouTube or it is to Alibaba. Yes, we’ve got this great car program that can see the road and tell what it’s doing and that’s amazing. But we have to keep retraining the data all the time, or retraining the algorithm and getting all the new data because the road conditions keep changing. And there’s so many long tail cases all the time that you’re continually having to get more data, you’re having to scrub the data, label the data, train the AI, just to keep your effectiveness at the same, whatever, 95% level. You’re always losing ground. So you’re always having a significant human services component as opposed to just pure software. So it looks like the operating leverage is not awesome. And it looks like it might even be bad. Like at a certain point when you grow larger and larger and you start taking on more and more use cases and more and more long tail situations, it may be that. not only do your costs not decrease, your costs may increase as a percentage of your revenue. That’s kind of, yeah, there may be diseconomies of scale in that regard. So it’s not clear at all to me at all what operating leverage looks like for AI focused companies that are dealing with the external world. Now, if they’re dealing with a manmade human made world, like we’ve built these warehouses and our AR moved the… the robots along the warehouse we’ve created, then it looks a lot more like software. But once you start going from a purely engineering game to a, you know, let’s go out and do science in the real world game, the operating leverage looks like it takes a major hit. That’s just my working conclusion, I don’t know. Most of these AI companies haven’t gone public yet, so it’s hard to get a look at their numbers. But that seems, so I’ve been listening to venture capitalists. What are they saying? Cause they see the numbers and that seems to be what I’m hearing, but we’ll see. So anyways, that’s kind of one of my favorite, if not my favorite numbers operating leverage. And it looks like when you go from software to AI, it’s not nearly as attractive if it’s in the external world, as opposed to in a controlled scenario, like I’m gonna show you which videos you most wanna see next on Netflix, or I’m gonna move the robot in the warehouse. or I’m gonna move the delivery cart within the business park that we’ve all designed it. Anyways, that’s kind of where I’m in my thinking right now. Now, I guess last point, my recommendations for those of you who are investors looking at these companies and trying to make a decision, my general thing is focus on operating leverage a lot on digital companies, because it tends to change significantly. in ways far beyond, well, your fixed costs are A, your variable costs are B. No, no, it can really change with the products you’re doing. These are generally companies that are evolving quite quickly. So there’s a lot there that you can tease out, and I think it’s a great place to find stuff that other people aren’t seeing. Because let’s say you’ve got a good sense for the operating leverage of a company, and everyone else is just looking at the operating margin. Oh, this company has this 40% operating margin. we’ll just project forward, but you know, not only is it 40% operating margin today, you know it’s got high operating leverage. So you are fairly confident because of that, that the operating margin in two years is gonna be closer to 48 or 49%. Now, if you see that, one, you’re gonna recognize there’s a lot more value coming. And two, once those numbers start to move, everyone else is going to see it too, so they will immediately revise their estimate and the price will respond. It’s something that other people don’t see and when it starts to happen the price will reflect it very quickly because they’ll see the operating margin expand like oh we were underestimating its leverage right. It’s a great sort of natural way to capture something. Yeah look at sales, look at revenue, look at the operating leverage. If you can, break it down into the key factors and the drivers. Resist the urge to just project the current operating margin forward. Well, it was 40% this year, well, it’ll be 40% next year. I’m feeling optimistic, I’ll call it 42. People are just making this stuff up. Try to work out the operating leverage and its drivers. And you might be surprised where that number actually is. Most, not most people, a lot of people are just guessing. 41, let’s call it 43. You don’t know where it’s coming from. But you’re looking for a margin expansion, a margin contraction, that sort of thing. Other points, that’s pretty much all I wanted to talk about. Operating leverage, very different industry by industry, fixed cost versus variable, very different at various stages of the product lifecycle. Very different at various stages of a platform. When you start going from one platform to multiple platforms. Very interesting. You know, there’s a lot you can really take apart in this. I think, I mean, for me it’s on my short list of three to five numbers I always want to know. Okay. I think I’ve sort of talked that particular topic to death. The key concepts for today, for those of you who are subscribers, obviously operating leverage. I’m putting all of these under valuation, which in my nine questions is question number three. And other ideas, Smile Marathon, which is basically machine learning, AI, AI factories and zero human operations. I’m putting it under those concepts. It all goes under learning goal 34, which is my nine standard investment questions. If you’re not familiar with my nine questions, I’ll put them below. I’ve started to take this apart. I still owe you parts two and three on this. These are my nine standard investment questions. I always sort of try and force myself to answer as I go through a company. And this all goes under question number three, which is evaluation. Okay, that is it for content for today. As for me, I’m just puttering along here. It’s, I’m past the point of being stir crazy. I don’t know whatever comes after that, but I’m more than stir crazy. I’m itching to go. Started teaching a little bit here in Bangkok last week at another. group at Sasson ramp up. So that was kind of fun. But apart from that, yeah, it’s too many hours in the apartment. Now on the good side, I suppose, I’ve been doing sort of indoor bike training. Ruvy is the program I use, R-O-U-V-Y.com, which is you pedal on a bicycle in the room and then… you watch the screen and you can choose routes around the world and they’re real routes. I mean someone filmed them with a camera on top of a car or something. So I’m currently riding a bicycle in Switzerland. I did Japan a couple weeks ago and then I did some Italy and now I’m in some mountain range in Switzerland. So that’s kind of interesting in a stir crazy trapped in the apartment sort of way. Which I guess is good because of the kitchen shape. I’m hoping things are about done. I’m starting to plan trips out of the country and I’m sucking up the idea that I’m going to probably have to go through quarantine. Get on the plane, go do what you need to do and just accept that I’m going to spend two weeks sitting in a hotel every time I come and go. Anyways, we’ll see. But yeah, that’s about it. Not a whole lot going on here. getting awesome at Ghost of Tsushima. I’m like, I’m really good as a samurai it turns out. Because you know, that’s a skill I really needed at this point in life. So I feel pretty good about that. Anyways, that’s it for me. Doing well. I hope everyone is doing well. And for those who are subscribers, I will send out the notes on Jong An tonight. So that’ll be kind of a bit of a long day. I know I’ve sent you a lot of content in the last. Week or two. I don’t really, I mean, I’m sort of giving you, I think more than would be normal to read through. I mean, it’s a lot of thinking and it’s a lot of strategy and it’s, you know, they’re long. My recommendation on that is go through them multiple times. That’s how I do it. I don’t, when it’s a lot of thinking, I try and do three or four passes and then, you know, I maybe I’ll go through it once and I’ll make some notes. Then I’ll go through again and I’ll read my notes like a week later. I’ll just put them in a stack on the corner of my desk. and then at some point I’ll go back and I’ll type them into my checklist. But I generally give myself three to four passes for complicated, important stuff. And what I’ve been sending you is pretty, not heavy duty, but I mean this is not light reading. And so if you’re feeling a little overwhelmed, like it’s oh my god, here’s another 2,000 words on, what did I send you the other day? Monster energy drink. Which I think is a pretty neat company. You know. Don’t just let it sit there and overwhelm you. And print it out. I actually kind of recommend this is print it out. That’s how I tend to read things is I print them out. I get a pen. I sit, I flip through the pages, I mark it up. And then I sort of put the stack somewhere and I come back to it later as opposed to trying to read it on a screen. I just find I learn better that way. But anyways, I’m well aware that I’m sending you a lot of content in the last week or two. So anyways, that’s it. I hope you’re doing well. Have a great week and I will talk to you next week. Bye bye.


Leave a Reply