E23: Tomasz Tunguz on the Theory Behind Theory Ventures

0:10Welcome back to Turpentine VC, a podcast where we discuss the art and science of building successful venture firms, VC to VC. Today's episode is a special one. CJ Gustafsson, host of Run the Numbers, a Terpentine Network podcast, joins me in interviewing Tomas Tongas. Tom is a general partner at Theory Ventures, an early stage firm he founded in 2022 after 14 years at Redpoint Ventures. We discuss the Theory Ventures model, the coming contraction in investable VC dollars, Tom's contrarian view on blockchain, the parallels between PE and VC, and much more. Here's our conversation.

0:56Tom, welcome to the podcast or podcasts. Thanks so much for joining. Thrilled to be here. Thanks for inviting me. So Tom, let's start with the theory behind theory. uh what you know given that you're entering a very crowded uh venture market how did you think about hey where do you want to fit in within the within the ecosystem where did you want to differentiate walk us through kind of the idea maze of how you thought about you know you're getting your fund off the ground and where you wanted to play when you thought about fund size when you thought about portfolio construction where you thought about where you wanted founders to see you in market compared to all these other amazing firms have been around for a long time Yeah, great question.

1:34I mean, there are a lot of wonderful firms out there. The way that we think about ourselves is we initially started by using math to create portfolio construction. And so we started running Monte Carlo analysis on historical venture returns to come up with basically a thesis-driven, highly concentrated portfolio where we research spaces for long periods of time, try to understand the entire landscape, and then invest at the early stage and continue to invest as companies grow. So that's the core idea. And today we're a team of six people all pursuing that vision. Yeah. And you have unique insights on portfolio construction in terms of you want it to be more concentrated, higher ownership.

2:14Talk about how you got to conviction on that being the right opportunity. Yeah, I think there's probably two dominant models in venture capital. I think there's the, you would call it like the Y Combinator Build-In Index, very, very broad. and then ideally have a growth fund or a mid-stage fund to be able to pick up the winners and then concentrate. And then the other is what you might call the classical venture model, which is very early bets with significant ownership, where over time you're able to drive a lot of returns for investors. And we think the U.S. venture capital asset class grew from about$800 to $300 billion over the last 12 years.

2:4981 % of those dollars, according to PitchBook, are from non-traditional VCs. And as we start our firm, one of the things that we need to do, and this is the advice that we give to startups, is we need to focus to be able to win. And so our goal is to be able to focus on categories that we really care about, go really deep. And then once we have a lot of conviction in those businesses, help them grow as quickly as we can, both with advice, but also capital. And that$300 billion number, where do you expect that to be in the next uh in the next few years yeah i think it settles to somewhere around like 150 to 180 would be my guess um so pretty significant correction uh obviously the vast majority of those dollars are kind of in the mid to late stage sort of definitionally um so i i think and where will it be in five or ten years is much harder to predict but i do think over the last 10 years the venture capital asset class has become institutionalized if you think like dave what David Swenson did at Yale, 30 to 50 % in privates.

3:47That's here to stay. And one big part of it is index investing has dominated the public markets. And the second is the total number of publicly traded companies in the US is significantly less. And the last part is 25 years ago, a company with 25 million in trailing could go public. Today, you probably need 150 million in trailing. So if you're an institutional investor and you want exposure to some of these mid-stage, early mid-stage companies, the only way to do that is through privates. Yeah. And so if we go down from 300 to 180 or 150, almost a cut in half, where does that get cut from? Is it multi-stage firms just lower their fund size significantly?

4:26Is it just that there's no new entrants? Is it that a lot go out of business? Is it all of the above? How does that happen? Yeah, I think, well, one, you have hedge funds who are public and private. They look at, I would imagine they're looking at public multiples on a relative basis and thinking that the public markets are more attractive. You look at if you want to play the AI trend as an institutional investor, Microsoft is actually a really phenomenal way of doing that. And so I wonder if some of the hedge fund money moves out. The second is the really big growth funds will be much harder to raise than they have been in the past.

4:58And the top firms will continue to do that. But, you know, they called, and I just learned this term during the fundraising, staple funds. And that term means raising the early stage fund. And then I have a growth fund. And they're called stapled because in order to invest$1 in the early fund, you must invest$2 or$3 in the growth fund. I think that dynamic may change. And so I think you'll see a lot less later stage capital for this reason. Yeah. And so what's your advice to emerging managers? You are an emerging manager, although, you know, at quite a big fund. a lot of emerging managers listening to this thinking about, hey, what's the market like?

5:33How big of a fund can I raise? But also, what is an approach that will resonate with LPs in this market? When you're advising emerging managers, what is some of the non-common advice that you're giving? I think the most important thing, well, one, a strong track record with DPI, I think, is the golden ticket in this ecosystem. The other, so I would say like six months ago or nine months ago, the fundraising market was quite difficult. And that's just because many LPs were re-evaluating their investments. And there was obviously the US public markets crashed, geopolitical tensions with China, the real estate implications or the implications to real estate of 400 basis points of increase in the Fed funds rate, what that means for the value of those assets.

6:18I think at this point, a lot of the repricings of the private portfolios have come through and limited partners or institutional investors have a better sense of where they stand, which is a good thing. I think what I've been hearing from the institutional investor base is that smaller funds is where many institutional investors would like to be because the multiples are higher. And so the most important thing is to be able to find LPs who are interested in potentially moving from bigger multi-stage down to smaller funds and asking qualifying questions about how many emerging managers have you backed in the last 18 to 24 months is a really good way of figuring out whether that LP could be a fit.

7:06And what types of strategies do you think are most differentiating? People are asking, should I differentiate on sector-specific? Should I differentiate on geo-specific? Should I try to do something different with portfolio construction? What do you think are the types of strategies that really resonate in the market outside of sort of, hey, we've got some DPI, it's great references, et cetera? I think that the really important question when you have a startup is, I have a product idea, I have a product and I have go-to-market strategy and they all need to align. And the companies that work really well are the ones that are able to draw a line through those three points.

7:39I think it's the same for early stage fund managers where you can focus on AI, super crowded category. If you have a proprietary network or access, you can make it work. You can come up with a different sort of fund strategy. There are a handful of friends who are starting SaaS roll-up funds where they buy companies that are not, they're growing at like 10 to 20%, 10 to 30 % profitable. And they're rolling them up as a way of creating a next generation constellation, which is a publicly traded company that does this. And so I think it's just about like the consistency of vision and some specialization that's defensible.

8:15And it's not that there's like a unique specialization, but just as long as it's like a very consistent story, you can be in a good place. Totally. I want to transition a little bit into, you mentioned AI earlier and how Microsoft could play. How are you approaching AI as a venture capitalist in terms of where do you think are opportunities uniquely suited for venture investors to play and invest and make money from? Yeah, great question. I mean, I think, so at the highest level, who's earning the most money from AI right now? It's OpenAI, like 1.3, 1.5 billion in run rate. That's what they've said.

8:51Microsoft has two businesses that, when you total them, are roughly the same size. And then there's sort of everyone else. The foundation model is a really difficult place, I think, for small venture firms like ours to play just because of the capital intensity of those businesses. Where we're spending a lot of time is trying to understand And so one, the tooling, we interviewed a bunch of data leaders, about 25 of them over the last couple of months, and only one had actually deployed an LLM inside of production. And the major challenge there is just understanding the risks associated with moving data into and out of LLMs and containing it.

9:29So one of our themes is just enterprise readiness around large language models. If you're a Goldman Sachs, if you're a Fidelity, if you're a Nike, how do you become comfortable with these kinds of issues? I went to dinner with a friend yesterday who works at a publicly traded company, and she raised this issue around terms and conditions. What indemnification exists if I'm using an image generation software? And so there's just a lot of these questions that have yet to be asked and answered around the use of LLM. So we're spending a lot of time around the developer tooling infrastructure. And then the second part is at the application layer.

10:07So it's clear you can't beat Salesforce with an LLM-enabled CRM. You can't beat Zendesk with an LLM-enabled customer support tool. And our current working hypothesis is that there's an opportunity to redefine categories. So there's a CRM category and there's a marketing category. and those have been sort of etched into the fabric of how we sell and build and buy software and we're starting to wonder whether you know these large language models you can use english as an api you can you can write something and it'll translate it to what a computer understands and bring you the data back but what if you could what if you could combine crm and a marketing software in one or the entire go-to-market functions in a single stack what would that look like?

10:54So that's a theme that we're currently researching is how do you redefine the way that people think about these categories? That's a helpful overview. What are other areas that you're particularly interested in investing in or sort of where you think there's lots of opportunities for startups? You'd like to see more startups pursue companies in this area or companies of this type? Yeah, we're keen on the modern data stack. So if you look at the vendor q3 data bi databases that's a single fastest growing category for spend the second is security and so we spend a lot of time in the modern data stack next generation databases streaming that kind of thing and then the last is blockchains as databases so we look at ethereum it's worth about 250 billion fastest growing database company of all time worth five snowflakes put together and one of our long-term theses is that many future software products were built with web three components embedded inside of them.

11:50And so those are the three areas where we spend time in addition to AI. It's contrarian to be thinking about blockchain applications in 2024. It is. It is. I mean, I think we've gone through this bear market and it's been a really healthy development because what we would have called like the projects of the 2020 and 2021 era, where a pitch deck didn't have any notion of like revenue or cost of customer acquisition or any of the economics of a business. Now that's changing. Many web three companies, and I think it's a very healthy development are now starting to think about what kind of businesses can we build?

12:25Which buyers do we sell to? How do we package up the core technologies that are fundamental innovations and really meaningful and solve an end user problem that doesn't have to do with a stable coin movement or decentralized finance, but is solving a real enterprise need. Yeah, that's well said. I want to ask one more question, then I'll pass it over to my co-host cj um you gave a talk at angel list confidential uh a year ago where you kind of uh talked about some of the biggest data trends of the of the year in in the space if you were giving that same talk this year and it just happened so maybe you did give give that talk um what would be some of the biggest uh points that you'd uh you you would emphasize in in that talk today yeah i think the first thing is historically data teams have been um well, I guess, but let me put it this way.

13:13Data teams are becoming software engineering teams. Historically, they've lived sort of downstream of the analysis, right? You build a product, the product produces data, your customers do stuff that produces data, and then you ship it to the data team and they do a bunch of post hoc analysis. Now data is becoming fundamental to the building of applications, personalization or risk scoring or chatbots. And so the data teams and the software engineering teams are converging. And that's, there's a, there's a lot of opportunity to build software to enable that to happen because you've, you know, that hasn't existed before.

13:51The second one are the applications of machine learning within data. So there's two big problems with these, these large language models that are relevant to data. The first is they're non-deterministic. That's a fancy word that means if I ask it the same question twice, it will give me two different answers, which is really hard. If you're the CEO of a business and you ask a BI system, what is my revenue by region for a particular product? It will give you one answer. But if you ask it, what is my revenue by region and product? And then if I flip it, and if I say, what is the revenue by product and region?

14:22I'll get two different answers. That's a big problem for data. And then the second is the hallucinations. So I think there's a significant amount of opportunity to solve those problems and then also bring more sophisticated data analysis to people who may not speak SQL. Text to SQL is a huge movement and many companies are spending a lot of time. And then the last is what I would call the small data movement. So the laptop that is in front of me today is more powerful than the server, the third most powerful server that Snowflake used to build Snowflake in 2012. So I can manage exactly the same volumes of data on this MacBook as Snowflake could in that era.

15:06And so there's this big drive now to take advantage of the fact that I can develop locally. I can process huge reams of data. I'm running some large language models on these machines. And as a way of both reducing costs and improving the developer experience for people manipulating data, you have next generation databases that operate and data analysis systems that operate and take advantage of the fact that have this incredibly beefy machine at home. Those would be the top three. Well said. CJ, take it over. I just want to say up front that I'm pumped to be here and included in this jam session.

15:43The lineup already had 88 Jordan Pippen vibes, and then someone accidentally invited the Horace Grant of Sassmetrics. So excited to be here. Horace Grant, even with the glasses. Remember those days? I thought about wearing glasses to this to make the joke really stick, but I didn't know if it would land. All right, on to more important things here. You know, I wanted to say that your writing is super well respected amongst operators, of which I am one. And usually VC content, it's really heavy on theory, but light on practice. And it actually inspired me to start writing a long time ago. And that's how Eric actually found me.

16:19And one of his pitches to get me to sign up with Turpentine was that I'd have hopefully a chance to talk with you here. And so I've been working on these questions, not to freak you out here for, you know, a couple of months now. And so the first one I wanted to hit you with, Tom, was, you know, VCs are in the game of providing capital, right? But they often have a skew or an area of expertise that they offer to help them differentiate themselves to founders. And for some that I've worked with before, it's pricing, others, it's market analysis, and some, it's go-to-market strategy. In your writing, you cover a lot of ground.

16:59What would you say your skew is at Theory Ventures to differentiate yourself amongst others? I think the thing that we want to be known for is the depth of our research and our analysis in these domains. Success is hearing a founder say, you understand the space better than any investor that we've spoken with. And the reason we want to be known for that is we think about launching a rocket, right? If you launch a rocket and it's two degrees one way versus two degrees another, today that may not matter that much. But in 10 years time, over long distances, the net result is pretty meaningfully different.

17:35And so if we can be users of the technology or have a deeper appreciation for how to build these companies, Do you need a customer success team for an enterprise deployment or is it better suited to a PLG? We hope that we can save companies or we can more accurately point the rocket and help more accurately point the rocket at the early stages so that as the businesses pick up steam, there are fewer corrections and it's a more efficient path to that ultimate moon landing. I like that. I like that. I think there's some formula out there for speed. It's like velocity times, I don't know, something.

18:11or other like that. But Tom, you seem to stick with companies and support them for a long time. And that inevitably means there are times when you have to replace a founder or level a founder and bring someone in who's more experienced. I wanted to know, how do you know when it's time to suggest that? And second, is there a preferred way of messaging that that's worked in the past than that you often lean on with operators? So this question, I think the best way of doing it is having a consistent open dialogue. And the way that I've gone through this, we've been through this as boards a few times is we're all strong in certain areas and we're all weak in different areas.

18:57And the business needs and demands different things from its leadership, depending on the journey that it takes. What often happens when you need, well, when the board needs to replace a leader is that a founder or the current CEO is strong in an area that the business or the business needs the CEO to be strong in an area that person isn't very strong in. And the CEO will often seem like exhausted or stressed about it because they're having to learn a skill in real time and it's a challenge. And so in, I'm thinking of at least two or three different scenarios over the course of a few months, you can pretty quickly see that happening.

19:35And I think those conversations by and large, if they can start from a place of empathy, I see you struggling. It's clear that the business needs this. Let's figure out the right way of introducing that skillset into the business and finding people who are experts. That natural transition is ideally the best outcome for a business because then you can bring in a seasoned operator who's strong. They can work together for a while, whether it's like as an advisor or board member. You know, and these processes often take like three to nine months and there needs to be a lot of trust when that transition happens, both with the outgoing CEO and then also, and maybe even more importantly with the management team that's there in place.

20:13And so it's not just hiring. It's not a very, very quick hire. So that's the way that it's really done, that I've seen it done well, or at least that's the, those are the better processes that I've seen. that's a super empathetic way of going about it and i wanted to ask do you think there's in your very data driven do you think there's a pattern when this starts to bubble up most commonly whether that be you know a revenue size or number of employees where you know the founding team starts to have those struggles so there's teams who from zora taught me about this which is and there's a blog post somewhere out there that i wrote which is there are these different if you think about great teams are structured with a span of control of about seven so typically a manager's best is working best when they have no more than seven people and so you know first early days of a business there's a founder and then seven people reporting to that person then you have to add another layer of management to the cake and then another layer of management to the cake and the way that it's described is there's a skill set of managing people there's a skill set of managing managers of people, there's another skillset, which is managing managers of managers of people and so on.

21:24And so there's this like, I think the number is like eight, 37, 144. And it's basically just how many managers do you have and how, if they have seven people, what does that layer cake? So that's one dynamic where it can be, I mean, like learning to delegate, I'm stunned that it's not taught in business schools. I have needed it and I'm finally getting some training in it. But that's a very, very difficult skill to learn how to manage through other people. So that's one driver of the needs of companies. The second is there's a strategic shift in the business. So let's assume the company historically has been product-led and now all of a sudden the market dynamics require that the company move into the enterprise because a competitor has come in and commoditize the low end of the market.

22:14That can be another change where you need a more sales-oriented leader who has experience managing outbound or large field sales teams. That's also sort of like a big strategic shift that can be another pretty important driver. And then the last is, and probably the most difficult and emotionally fraught are unresolvable disputes amongst the management team about strategy. So if you have two founders who fight or don't see eye to eye, who for some reason or another have grown apart, which happens, that's another pickle that leads to these kinds of situations. How rare in your experience do you think it is that a founder makes it all the way through IPO?

23:01I see stats sometimes on arguing, are founder CEOs inherently better because they know the business from its inception and maybe they're more leaning into taking risks versus professional CEOs. Well, what are your thoughts on that? Yeah, I think it's really difficult to paint in broad brushstrokes one path. I mean, the founder imperatives, like the gravitas a founder has within his or her own business is unquestionable. There's just a respect that the employee population affords to that person for having created this incredible machine.

23:46And the hard part, I think one of the hardest parts about being a leader of an organization, particularly a hyper growth organization, is that the business evolves very quickly. And the very best leaders are the ones who anticipate what the business will need from them and ensure they either have the skill set or the people around them to meet that need. and so you can do that in a bunch of different ways you can learn it yourself or hire advisors or build out a management team and some people are better at doing that than others and consistent you know i had this great manager at google kim scott malone who wrote the book a radical candor and she taught me the best leaders always manage themselves out of a job and so the people who are better at doing that i think are the ones that scale but there are success stories and what i've learned, I think maybe if I, one maxim to kind of make this point is there are many ways of getting to the top of the hill.

24:37That's what I've learned in venture. Like you can go to, you don't have to go to college. You can go to Harvard or Stanford and you can still be really successful. You can be non-technical and be really successful. You can be technical and really successful. And so, and I love that part of the Valley. Like that's the ethos for me for startup land is that you can come from anywhere and achieve what it is that you want. It's the American dream. I think is very much alive. So for me, it's hard to say there's a hard and fast rule. And I hope that remains the case for a very long time. I love how you emphasize the forethought to think what the business will need in the future.

25:12And maybe it differs based on how fast the business is growing. But as an operator, I always tend to think higher for the person you need 18 months out from now. What's your take on that? Is that too long of a runway to be thinking forward? No, I think that's right. So the way I think about it is like, if you're a seed stage company, your horizon is three months. The business should be planning for three months. At the series A, you should be thinking maybe like four and a half, five months out. At the series B, the leadership team is thinking six to nine months out. At the series C and beyond, it's 15 to 18 months.

25:48So what do I mean by that? Well, your product roadmaps need to be longer. When you're seed stage and you're trying to find product market fit planning beyond three months or even six weeks doesn't make sense because you won't know what necessarily will hit but by the time you get to be a sales leader or a marketing leader of a series b or series c company you're not thinking about the pipe you shouldn't be thinking about the pipeline for next quarter you're thinking about how do i develop the pipeline for three quarters from now and how do i know that i'll be able to hit that number and so that sort of instead of looking at your feet you're kind of looking more and more to the horizon.

26:22So at the point where the business becomes publicly traded, this is awesome quote I just found from Bezos. I think it was two days ago where he said, he would often laugh when analysts, public company analysts would tell him congrats on a great quarter. And what he was thinking is that quarter was baked three years ago, right? So the work that he had done in order to get that quarter had happened three years before. So he's thinking three years from now. And so I think that that's a very sort of like natural transition, like looking from your feet to the horizon. I guess to piggyback a bit on the question I just asked, I'm a startup CFO.

27:00When do you usually counsel your portfolio companies to start thinking about hiring either a VP of finance or a CFO? VP of finance is, I think the best leading, it's not an ARR metric. It's a contract signed metric. So how many sales contracts are coming in within a given period? When does the deal desk need to be established to create some kind of consistency, both within the contracts and then also inform the plan? Because what you're really looking for is a pipeline to quota ratio that justifies the hiring of a marginal rep or two. And so that needs to be a combination between the sales team and the finance team.