If you ever need to explain something to somebody else, you really have to understand it. And I think writing in public is maybe the greatest forcing function for that.
So the most effective board members I've seen are the ones who ask simplifying questions. Running a business is incredibly complex, right? You think about all the different dimensions that there are: the competitive dynamics, the product, the engineering, the financials, the people management component of it. And there's, you know, there probably one to two priorities that really matter for a business. And there's probably one or two mental models that are useful for some short period of time in managing those priorities. And exceptional board members, I think, are the ones that are able to just focus on those things.
Gartner ran a survey, I think two or three years ago, where they looked at cost of customer acquisition for software companies, increased 60% monotonically every year. Cost of customer acquisition will increase for the industry. There's an inflation, a CAC inflation that is inexorable. And if that's the case, well, a company that has a five or a 10% cost of customer acquisition advantage compounded over seven years will need to raise like half the capital to achieve the same ARR.
So one of the lessons I learned is if a customer chooses a vendor, they're probably not switching for three years typically, right? And so, well, if that's the case and you're early in a category, then you want to acquire as many customers as possible because if they're not switching for three years, even if you have a product that needs a lot of work, the faster that you can grow, the more opportunity you have to influence that user or the buyer, get as much feedback about the product, and shape the ecosystem in the way that's advantageous to you.
And so I think there's a very real argument that startups should try to almost grow uneconomically, right? If you look at these coding agents, they're growing uneconomically because just like a social network, they believe that the distribution ultimately will be able to be commercialized. If you're starting a SaaS company today, you have a market that's five to 10 times the size that it was three years ago.
Yeah. Right. Awesome. You have incumbents that are, some of them are probably at their weakest point in a very long time. Also awesome. And then the buyers themselves are asking to be educated because they don't know. None of us knows how to use AI. But if you can find a subject matter expert who's figured out a particular new workflow that is really effective, well, you can grow really fast as demonstrated by many of these companies that are now on trajectories that were unimaginable five or 10 years ago.
Welcome. This is Second Act, a podcast where we document how SaaS leaders chart the exacting foundational shifts of scaling up. And I'm your host, Chris Subramanian, co-founder and CEO at Chargebee. In this season, we are diving into AI's all-out enterprise embrace and how that's accelerating second acts everywhere, reshaping software categories, business models, and teams to meet never-before-seen stakes.
Now on to the episode. With dozens of thoughtfully written S-1 analyses, rigorous breakdowns of major data infrastructure and now AI developments, all published at an insane near-daily frequency since 2010, this week's guest, Tomasz Tunguz, has chronicled SaaS as a devoted historian. As a VC at Redpoint Ventures where he spent almost 15 years, he backed nine unicorns including breakout companies such as Expensify, Lucid, and Customer.io. In 2023, he founded his own firm, Theory Ventures, with a $238 million first fund followed by a second fund at $450 million announced late last year.
In this wide-ranging conversation, we spoke about Tomasz's early fascination with venture and its three infinite learning curves, the lessons that have made him such a prolific writer, the habits of exceptional board members, why AI is forcing today's startups to take on complex second acts way sooner, his long-developing thesis on the importance of go-to-market leverage, why he thinks monetization is a perennial, monumentally difficult challenge, his favorite recent reads including Alberto Echo's Chronicles of a Liquid Society, and much more. Let's dive in. Tomasz, so excited to have you on Second Act. Welcome.
Oh, thrilled to be here. Thanks for having me on, Chris.
Thank you. So let me start with something that you had written about. You had recounted recently how walking along the Embarcadero in San Francisco once, you had heard an angel investor talking to a founder and had wished you could do that someday. Can you tell us about that early fascination and how that led you to Theory Ventures eventually, and when you walk the same path today, what are the conversations that you find fascinating today?
Yeah, great question. I was walking along the Embarcadero with my wife. We used to live in SoMa, which is a neighborhood in San Francisco, and we would walk to the farmers market every weekend. And I remember there was a restaurant there, Palomino, and there was a seed investor or an angel investor talking to a founder, and it blew my mind. I mean, you know, you kind of understand, well, this, I was really new to Silicon Valley, and it kind of blew my mind that it's just these conversations happen everywhere. You go into a restaurant or a coffee shop and these kinds of conversations are happening all the time.
And it also, I hadn't been exposed to venture capital before I joined the industry, and so it kind of blew my mind that this was an industry, this was a job. And so that, you know, I think created a desire to be in the industry. And here we are like 17, 18 years later with Theory. And I think it's an incredibly wonderful ecosystem that we have.
Where the other part that is different than I think from a lot of places is people pay it forward, or people just do nice things for each other there. And I remember getting started with Theory, a lot of people who I knew tangentially or just a little bit went out of their way to help me, and I think that's just a very small slice of all the generosity that exists within startup land.
Wonderful. I share this sentiment as well. For us, the inspiration was Joel Spolsky's Joel on Software blog, and then...
Oh, it was so good. I remember reading that. It was phenomenal.
The best, right? From early on in the career, it shaped how we thought about the kind of company we would want to build someday, and we were able to keep the dream alive for like 10 years. But like we used to look forward to reading that every week, and sometimes if he doesn't publish often enough, we would be waiting for the next one to come through.
Yeah, he was one of the first. I mean, he wrote a lot about different kinds of business strategy and... yeah, wonderful, wonderful content.
Right, yeah. So you also mentioned that you saw two infinite learning curves, right? One was about figuring out how an industry evolves. The other was more an unobvious one, was about helping founders achieve their potential. How do you approach those today, having done this for like now 17, 18 years?
Yeah. So the initial appeal for venture capital was just, I mean, I like you, love technology and it moves really fast. So to be in an industry where in a sense you're paid to stay on top of it or as close to the bleeding edge as you possibly can, I thought was one of the infinite learning curves.
I think the second learning curve is as companies go through hypergrowth, the business and the people within those businesses change, and to be an effective board member, I think you need to learn and change along with them. And then the third learning curve, which I only discovered much later on, was the industry itself is changing. So it's gone from 8 to 300 billion. There's a huge amount of retail capital that will come into the asset class. And so maybe by 2027 or 2028, it'll be a $500 billion asset class. And so how do you manage a business within an ecosystem that is changing a lot? And anyway, so that's why I think the business of venture capital has three infinite learning curves.
With founders, I think the, we were talking about this internally earlier this week, there's, you know, advice is a tricky thing, right? Because advice is one person's experience generalized, and can, if you apply it too broadly, especially when the underlying dynamics of an ecosystem are changing as quickly as they are, doesn't always work, right?
You talk to, I mean, certain leaders and recruiters today of executives, they're looking for a different kind of, like the classic sales executive or the classic marketing executive who hasn't thought through how to operate a department within the world of AI is in a challenged place until they can make that transition. And so you can't just take what worked 10 years ago and apply it now.
So I think the most important thing as a board member is to try to learn all the time and provide a series of different examples or a collection of different experts who themselves are tinkering and who are active.
Right, beautiful. Like amongst founders, right, we have this chatter about all the time about scaling yourself as a founder with the company, and because the company scales faster than an individual. First time I'm actually hearing that even the VCs actually have that. Like, is it like a common conversation that happens amongst VCs to say like about scaling with the company? Like this is first time I'm hearing that it's...
Yeah, well, I think, I mean, we, yeah, I mean, I think that's the purpose of a board is to figure out, okay, what does the company need now, and who do we have in our collective, who is within our collective networks that might have a perspective here? How do we find, identify the rate-limiting factor and then improve it, right?
So yeah, I think it's, I think it's a different, the nature of the conversation with founders, I think, is a little bit different because the pain is so much, it's a daily pain where board meetings might happen three, four, five, six, eight times a year. So I think the acuteness of it might be a little different, but I think the sentiment is still there.
I have one more question related to it, right? If a board member actually interacts a lot more with the company closely, even in the interest of learning, sometimes it can feel a little disorienting for people, like, why are you working on my company, right? Like, I'm giving you updates on a regular basis can be a reaction from some people. How do you find a way to then, it requires a certain personality to be able to just be like, work, be able to work very closely and then be able to still learn, but while still giving the space to let people operate.
Oh, yeah. Yeah. Yeah. Yeah. We're not operators. I'm not an operator. And so I think it can be dangerous or overbearing, I think, if we imagine ourselves in that role. I do think so. The most effective board members I've seen are the ones who ask simplifying questions.
That takes, running a business is incredibly complex, right? You think about all the different dimensions that there are: the competitive dynamics, the product, the engineering, the financials, the people management component of it. And there's, you know, there probably one to two priorities that really matter for a business. And there's probably one or two mental models that are useful for some short period of time in managing those priorities.
And exceptional board members, I think, are the ones that are able to just focus on those things and every once in a while provide a mental model for how to think about it or orient a company or a process. And that's why trying to learn from as many different companies about the way that they work, I think, is an essential part of the job because, I mean, like Munger, right? Like Poor Charlie's Almanac is just a collection of different frameworks, right? And yeah, you know, even Dalio, you know, his Principles book is just a collection of different frameworks and all of these...
Yeah, no way, that's awesome. It's somewhere around here. I think it's on my, that side table.
And so I think the more you can accurately know the underlying frameworks and then accurately map them to the business at the right time, the more effective you can be.
And is there a particular mental model that comes to your mind today that you recently talked about with a founder, especially for companies that are making the transition from pre-A to the kind of changes that are needed, especially?
Yeah, I remember, so Mike Viki, who used to be at Index and is now at Habi, I remember he asked this question that I use from time to time now, which is, is a company product limited or go-to-market limited?
So it's a very simple question, right? And obviously like hugely reductionist, but it's a really good, like, and so for a given company, it gets you thinking like, is it, do I have the right product and my challenge is top of the funnel or my sales effectiveness or marketing and positioning, or do I need to build something else? And if you have, those are two different modes, probably a pendulum swinging between the two.
Good. And you don't want to get into the builder trap. It's not like you'll never stop building more and more things, but it's just that are you getting into a builder's trap and think you, unless you build it, you will not be able to activate a go-to-market motion to be able to succeed. Like it's a huge unlock. Like it's a beautiful question.
It's a good one, right? And so I would argue like after what GPT-3.5 came out, most companies were product limited.
Even if they had huge go-to-market teams, because now all of a sudden they needed an AI feature or some positioning, and then when that was released, okay, now we're go-to-market limited because now we need to learn how to sell this new product and educate the customers. And then, okay, now you've done that, great, the rest of the competition has caught up. Now we need to go back and innovate again. And so you have this sort of classic pendulum swing.
Right, wonderful. So, switching topics about writing. So, you are a prolific writer, right? Almost every other day. So, Intercom's co-founder Des Traynor once mentioned that you put us all to shame, and I wholeheartedly agree with him. So, you write almost every day and have done so for a decade and a half, which is phenomenal, and what you published has been incredibly influential for a lot of us. So, in what ways does this discipline show up internally at Theory Ventures, and is there a bit of writing advice that has stuck with you over the years that helps you?
Yeah, great question. So, writing, if you ever need to explain something to somebody else, you really have to understand it. And I think writing in public is maybe the greatest forcing function for that because you really have to, you know, you really have to be able to communicate in a particular way. And so I'm really grateful for the opportunity to write all the time. I think it's grounding and it's simplifying and anyway, I love it and I'm really grateful for the opportunity.
I think the one piece of advice, especially for business writing. Okay, so when I remember graduating college and I really wanted to work at McKinsey, like I desperately wanted to work at McKinsey, and there was this book that I had been told by a couple of current McKinsey consultants that there was this legendary book that was written by one of the founders of McKinsey that every McKinsey consultant reads. And, you know, I found a copy on eBay for like three or four hundred bucks. Couldn't afford it at the time.
And then I got a job at a startup and then I went to work at Google and my manager, a new manager came in. His name was Scott and he was a PhD in aerospace engineering and then had worked at McKinsey. And so he taught me a lot of the concepts from this book. And one of them is actually, it's called situation, complication, question, and answer. And it's a form, it's a way of making an argument that is very structured and consistent. And if you read my blog posts underneath the hood, they're all SCQA, all of them, because it naturally really sets you up to have a hook.
Like, what's yesterday's post? Let's hold on. So we'll break it down, right? So, oh, maybe we'll take the second-order effects of AI, right? So the first sentence, AI revenue will double classic software. So that's the situation. And then that's the situation. The complication is the trend is so large it's starting to have second-order effects. What question does that raise? And then you say, well, where do you see it? Well, the answer is in MongoDB's earnings. And then you do two or three Q&As and a conclusion and then you're done. And so it's not that everything is that rote and structured, but the...
Yeah. Like the, anyway, the idea is like it gets you to a so what, like, why are you telling me this, right?
Beautiful. So that makes me remember another framework that we used to have. This is TIPS, which is the trends, implications, possibilities, and solutions. I don't know if it's derived from somewhere like this, but it's one of those ways in which we would prepare if you meet someone and if you have to explain this, how would you explain the trend and what is the implication of that? And then what are the possibilities for them? And this is in consulting. And then what are the solutions that are relevant for them?
Oh, I need to check that out. Yeah, that's awesome.
Thank you. I didn't realize that there is a simple framework that you use, but it's brilliant, right? It's a simple way to think about it, but it actually enables you to be able to structure it. And the more you write, I'm sure you would have figured out your own editorial way to actually make it better and better. It's awesome. Thanks for sharing that.
No, no, of course. Yeah. The other book that I'll put in a plug for, there's a guy named Verlyn Klinkenborg who is the executive editor of the New York Times, and he wrote a book called Several Short Sentences, which is an awesome read because each sentence is on its own line in the book. And yeah, so there are no paragraphs. Every paragraph is a sentence. It's a one-sentence paragraph. And there's two key lessons from that book. The first is the responsibility of each sentence is to ferry and excite the reader to go to the next one.
Makes sense. Like Copyblogger used to say this, like the purpose of each line is, the first line is to make the reader read the next line.
Yeah, exactly right. Yeah. And then the second point of putting each sentence on each line is to look at the cadence and the rhythm because it should feel, there should be a rhythm to it. It shouldn't all be long sentences. Nobody wants to read that. It shouldn't all be staccato short sentences. Anyway, that's a fun read.
Awesome. Awesome. So, thank you. This is about writing, right? The part I also want to understand from you is, you said how you write to think. You also said that how you write to think and figure out what you believe in. And can you tell us about one of the recent pieces that helped you discover something important about a market you are evaluating or an internal challenge that you are grappling with in today's context?
Yeah. I mean, I think one of the things that we're all learning is how to work with AI. And writing about how to work with AI, I mean, I think is one of the things that I didn't realize. I mean, I started writing about agents and then readers started to email me and say, I don't know what an agent is, right? And it's hard. Like I was sitting with a relative, we were having lunch and he asked me, can you show me what you talk, you write all the time about AI, show me, like I know how to ask questions, but make it do work, right? Like make it work for me.
And that blew my mind, right? Like, and I guess, you know, whatever, living in the bubble and all that stuff. But that was one example. And so I've been trying to, in the writing, like whenever, every once in a while I write a piece just for fun, but to show more and more of those workflows in ways that are much more tangible.
And I think in a sense, as leaders of organizations need to do this for their people. We were, you know, we had a dinner with ten executives and they were all talking about having demo hours or hackathons or some way of exposing within the organization how different people are using AI. You have one of those thirty-minute sessions once a week. It's amazing the creativity. So that's, I think, an example of exposing or learning something about the ecosystem.
Brilliant. This is awesome.
Yeah. What's your, I mean, do you have a fun AI workflow?
Uh, sorry. Can you repeat the question?
Yeah. Do you have an AI workflow that's one of your favorites you use all the time that just gives you a sense of satisfaction of leverage?
Right now? Um, that's an interesting question. Satisfaction and leverage. Mostly the ones that I use for summarization, the research and summary to help me think through on various topics is the one that I use a lot more, right? Especially the one where I can use as a sparring partner to understand newer subjects and then apply that in our context in the business. That's the one where I have a GPT that actually enables me to think better.
Yeah, I agree. Summarization is amazing. The volume of information that you can consume.
Right. And then use it as a thinking partner to say, how do I apply it in this context? And then the comparative dynamics, then you can actually like do a lot more together at the same time. And then it just moves things forward at a pace that I couldn't imagine earlier. That's the one where I find it most productive.
Yeah, agreed. Yeah, sorry, that question caught me off guard.
No, no, no. I'm trying to keep you on your toes here.
All right. So GTM, so you have spoken a lot about why GTM is perennially harder. So whether it's a Theory Ventures thesis that states that you invest in companies that leverage tech discontinuities into GTM advantages, or the scores of S-1 analyses that you have done, you've always had a keen emphasis on emphasizing the outsized gains of figuring out go-to-market. And I would love for you to unpack that thesis and tell us how you come to believe in it so strongly over the years, especially about GTM.
Yeah. Okay. So I had this Belgian professor when I went to grad school who on the blackboard first day of class, he wrote innovation equals invention plus distribution. And the core idea there was in order to change the world in some way, you need to invent something awesome and then you need to convince people to buy or use it. You need a combination of the two. If you invent something amazing and don't tell anything about it, good for you, right? But you know, and if you just market vaporware, well, okay, not a whole lot there. So you need a combination of the two.
And I think being a technologist or, you know, because anyway, starting my career as an engineer, I always thought it was at least equal, fifty-fifty invention distribution, if not seventy-five percent invention, twenty-five percent or more. And I read the HBS case studies on Betamax VHS, and you read the history of like different forms of electrical current, AC versus DC, or early electric cars versus...
You read that lesson enough and you realize actually distribution a lot of the time is a bigger lever than the technology. Every once in a while that technology can be so amazing, so compelling, would be an example, right? ChatGPT, it's a magic moment for many of us. Technology is so amazing that distribution took care of itself, right? But, you know, even Google, one of the, I think, untold stories about Google was they were very aggressive on distribution and they focused a lot on it.
And so anyway, so with that framework, I entered the working world and one argument, and I don't know if this is broadly true, but I'm starting to wonder if it is, the cost of invention, particularly in software, is less and less, right? Vibe coding or AI should ultimately reduce the cost to make a new invention. And so in that equation, the relative value of distribution increases.
The other dynamic here that's really important, Gartner ran a survey, I think two or three years ago, where they looked at cost of customer acquisition for software companies, increased sixty percent monotonically. Every year cost of customer acquisition will increase for the industry. There's an inflation, a CAC inflation that is inexorable. And if that's the case, well, a company that has a five or a ten percent cost of customer acquisition advantage compounded over seven years will need to raise like half the capital to achieve the same ARR.
And so if you can find a company whose technology allows them or something inside of the company that allows them to five or ten percent better cost of customer acquisition, they'll be the winner if they can sustain it.
Beautiful. And that advantage could be attained through like any number of ways, which could be a pricing lever or the cost of producing the product or like using that as an advantage to subsidize the product distribution, any number of these combinations. Is that how you would think about it?
Exactly right. So open source, PLG, sales-driven channel, right? PR, I mean, there's so many different, social media, brand of an individual. So there are a lot of these. Who was it? There's a publicly traded project management company that started advertising on YouTube because there was nobody else.
ClickUp. Oh, Monday. It was Monday. That's right. Yeah. And so they found an arbitrage opportunity in paid media on YouTube. Well, phenomenal, right? So there's some wedge, there's some angle where there's a cost of customer acquisition advantage that's sustainable. And so imagine you have five companies, all raise the same amount of money, all found at exactly the same time. The one that has that small efficiency gain, it's like a Formula 1 race or sailing race, they'll always be at an advantage and then over time it compounds.
Mm-hmm. Brilliant. It makes me reflect. I want to go back and then do that homework.
Right. So earlier this week, I spoke to Reforge's Brian Balfour, and he spoke about the rise of a new distribution channel in ChatGPT. How do you see the evolution or decline of some of the past distribution channels coming about over the next couple of years?
So I think we made this prediction in 2023 that fifty percent of consumer search would be AI within twelve months. It took us eighteen months to be right. So we're already at a place where classic search is a minority of search broadly. And then in February or January of this year, I wrote, and there was a lot of talk about, you look at HubSpot's traffic falling seventy percent in a month.
Right. So clearly all of a sudden the ways that we're educating ourselves about enterprise software have changed in a pretty meaningful way. And now the question is, okay, well, whenever you have a platform shift, the teams that figure out how to manipulate the new platform to their advantage will win. And this was true with Google search engine optimization. It was true with mobile app store rankings, and it will be true with AI.
That part I think is now, I mean, many people are trying to figure this out, and the people who figure it out will have distribution. I mean, Neon, right? Seventy-five percent, seventy or seventy-five percent of the new databases created in the quarter before they were acquired by Databricks were all created from coding agents. So it's not necessarily that all of this discovery for enterprise software will happen through the dot-coms, right? The web pages or the mobile apps. Much of it might just happen through agents.
And so now, great, if that's the premise, well, what kinds of content and which formats need to exist where to be able to supply agents the information to choose my product? That's all new, right?
