How I Invest with David Weisburd - E4: Dr. Abe Othman of AngelList | Data, Research, and Quantitative Investing in Venture Capital

Episode Date: August 7, 2023

David Weisburd and Erik Torenberg sit down with Dr. Abe Othman, the Head of Data Science at AngelList and the Head of the Investment Committee at the AngelList Quant Fund. Dr. Abe has one of the largest and most granular private market data sets in the world of nearly 15,000 startups. In this discussion we dive deep into the data: power laws, what Dr. Abe’s research tells us about which companies to avoid at all costs, and whether alpha truly exists in venture capital. If you’re ready to level-up your startup or fund with AngelList, visit www.angellist.com/tlp to get started.

We did a study on founders alma maters with the performance of their startups. What we found is that those founders actually, their markup rates were higher than baseline. It is a positive signal to invest in someone who went to GSB, but that it's not as positive as you think, because what's happening is that the startup founders are aware that they possess this positive signal and it's translated into higher valuations for their deals, which means which lowers the chance of future market. Welcome to the Limited Partner Podcast, where we talk about venture capital through the lens of limited partners. I'm your host, David Weisberg, co-founder and head of venture capital at 10X Capital.

The world of LPs can be notoriously private and discreet, but on this podcast, we speak candidly with limited partners about their true feelings on venture, the ecosystem, and 2023 and beyond. Good afternoon, Abe. Abe Offman from AngelList, head of research. Pleasure to meet you and welcome to the podcast. Thank you so much. It's great to be here.

So let's get started on something fun. So one of your most famous slash infamous blog posts is what AngelList data says about power law returns and venture capital. And in this article, you claim or you claim that the data says that only top quartile or top 25% of VCs beat the entire basket of venture capital assets. And explain what you mean by this and how this could possibly be true. Sure. So this was one of the earliest things that I published at AngelList. I would encourage folks that are interested in seeing a little more detail on the research of power laws within venture capital to take a look at something that we published at the end of 2019 called Startup Growth and Venture Returns, which is a little bit more,

I would say, rigorously grounded. But the conclusions are relatively the same, which is that early stage venture, we're talking pre-series B and ideally seed round is governed by a very extreme power law in return multiples, meaning that there are many ways of characterizing what a power law means. One is, I think maybe one of the easiest ones is the divergence between mean and median or the divergence between average and typical outcome. And I think for really wild power laws like what you observe in seed investing, it's completely sort of unintuitive what happens there. So let me just give you sort of an illustration of that. The AngelList platform as a whole has returned something around 26% net LPs on a yearly basis.

What that doesn't mean is it doesn't mean that like a typical AngelList SPV investment will earn somewhere between 20% and 30% a year. That's absolutely 100% not what that means. What that means is that the typical AngelList SPV investment will more or less be flat and that there's an incredibly small minority of investments that are responsible for producing virtually all of those returns. Consequently, if you are a venture capital firm and you are picking 10 companies to seed investments in, your returns are really going to be driven by how many of those companies end up returning in the top decile. And the market return is to have one-tenth of the companies trying to top decile, including all of the most outstanding companies are going to be in the market basket portfolio.

And so you as a venture firm, in order to beat that market basket, really need to have two of your 10 picks be in the top decile and hopefully have one of those picks be relatively far into that top decile in order to beat the market. So that's kind of the suggestion there is that the typical venture firm will not be diversified, will not have enough bets to achieve anything other than maybe a small lift off of the typical investment return. And given the wild power play here, the divergence between average and typical outcome, the average return, which is the market basket, will tend to be much, much higher than the median venture firm. There are some shenanigans that happen in terms of venture capital funds. We had always an Angelus and always closely looked at SVB balance sheets and a huge source of SVB is profitability or capital call lines of credit, which, of course, have a completely innocuous explanation of time shifting and making sure that funds can pay for quarter's investments or whatever, but also have a

very not innocuous explanation, which is that they are used to juice funds, IRRs for reporting. Particularly if you're a fund manager and your goal is to have as high as possible IRRs two and a half years in when you're raising your next fund, capital call lines of credit are a great way to do that. So I think actually some of the skepticism around this result is more a function of the steps that GPs take to get their fund to have an artificially high IRR, at least in terms of reporting metrics where they want to talk about it. Obviously, using those capital call lines of credit the end result is a lower tvpi for your lps but in the short term it can be used to juice irrs and i've heard a lot of people criticize that capital line of credit but at the same time that capital is not coming from

lps why is that an artificial metric why is that a vanity metric when when lps are not actually funding the that capital it depends on what you think the point of what the implication is when you report an IRR to a prospective LP. If you are reporting those IRRs with the intent of making a suggestion, even if there's a bullet point on your slide that past results are not indicative of future returns, I'm saying we're going to return this much in the future on an IRR basis. It's very misleading to have a juiced irr from capital and credit obviously by a financial

perspective you know that it presumably you're not committing fraud like that actually is the irr refund but it is misleading for lps to think oh wow that's like a you know 70 plus percent irr and you know it does lead to i think in general skepticism about those numbers, discounting. I think the market is fairly efficient from that perspective, but it also tends to hurt funds that don't sort of play this game, that essentially you sort of have to, because if you go out there and sort of you've done well first few years and you're honestly reporting a 25% IRR or something, you're going to look worse you're honestly reporting a 25% IRR or something, you're going to look worse than the person reporting a 45% IRR that may have made the same

or even worse bets, but just been much more clever about how they've handled their first set of capital calls. You mentioned the median and the mean. You, of course, reported on the whole market, the 26% across all of AngelList. You have other niche funds like the Access Fund, which invests alongside the top VCs. What is produced better? And what are the learnings from that comparison? I think one of the learnings, and we still don't really have a great sense of this, is that there is kind of a minimum size or a minimum number of startups that you sort of need to be exposed to get sort of the benefits of being broadly indexed.

And that if you're not there, if you're only making 10 investments, it's sort of a different game than if you're making, say, more than 100. The Axis Fund has in general, in terms of the slice of investments that it has had the opportunity to invest in, Parfer and the Axis Fund has shown alpha in selecting those investments. The actual returns of the Axis Fund because of some investment sizing issues are not as high as they should be, but in terms of the perspective of just like Parfer, you're going to vote yes or no, the subset of things that end up getting Axis Fund investment, do those end up performing better than the universe as a whole the answer is yes uh that said you know the access fund is investing in a lot of stuff um you know something like 20 or 30 percent of deals um so it is pretty broad as opposed to like a narrow selection of a small group of startups uh you know there's a question does

alpha exist in venture capital and i think actually this is one of the lines of research that we have. I believe the answer is yes. So in a couple of ways. So first you have the existence of the Axis Fund through whatever means that the investment committee is making their decisions. We have a very clean historical, it's a really nice experiment, right? Because we know the investments the Axis Fund had the opportunity to invest in. We can see their performance relative to a random similar size subset or just a blanket

kind of yes to everything. And the Axis Fund does appear to have alpha consistently across years. So it does appear that who knows how they do it. Maybe they do a deep dive in slide decks. Maybe they invest in companies that, you know, their logo is the color blue. I don't know. And I don't think it matters from the perspective

of the question of like, is there alpha? Does it exist? Yes, I think the Axis Fund is certainly a positive example. We also looked at this and I don't think we've published this research. I think it's somewhere in our queue. We also turned the question around, is there a subset of syndicated deals

on the AngelList platform that were sort of a priori identifiable as being worse quality than other deals? And what we did for that identification based on discussion with the investment committee was look at list of co-investors. So if you have a deal, an early stage syndicated deal where the only co-investors that have listed co-investors, so the GP has entered this information, and the only

co-investors they've listed only ever appear in that deal, which is about 10% of the deals of the early stage syndicates, that subset is noticeably and consistently worse performing than the market as a whole. So these characteristics suggest to me that there is in fact, despite some appearances the contrary, there is in fact a rank order list of seed investment opportunities. We can dive in on what our shape of that universe looks like. You know, the question of like, I don't think it's accurate.

I don't think it's supported by our data to have the view of early stage investment as being throwing darts at a dartboard or being, you know, completely random or some undifferentiated mess. I don't think that's supported in the data. And just to double click on that, you're saying that only 10% of all deals on AngelList would not fit the quality bar.

Can you explain that? I think that they are a priori identifiable as worse than deals that do have a recognizable. By recognizable, I mean they do multiple deals, co-investors. The interesting thing is these investments take on a season. If you go back to the startup growth and venture returns paper, you see that what happens is that as companies compound returns, the return multiple alpha parameter gets lower, i.e. the return multiples get wilder over time. And so right now, the return multiple that we fit to these like bad, or maybe

it's better described as like marginal subset of deals is above two, but there is a reasonable likelihood it will fall to below even this marginal subset of deals on Angel's platform will fall to below two. I think realistically what that means is probably the sort of the double gate of the Angel's platform, which is like first your deal has to be approved to be on the platform. And then there have to be enough LP. You have to get enough LP interest to actually close the deal. It probably puts the bar higher than what we identified in the startup growth and venture returns paper, which is like the credible deal threshold, which is worth diving into, I think.

Let's step back for a minute and define power laws. How would you explain power laws and venture capital to a lay person or to finance professionals and other asset classes? The easiest way to describe power laws is that they have the opposite intuition of everything you've been told about in terms of the typical world. So I'll just give you an illustration. I've been working with this stuff for years and still this surprised me. We had an inquiry from an AngelList GP whose startup had

just had been made in early stage investment. They just raised series A. Their investment after two and a half years was now marked at 2X. And they asked like, hey, would it be possible to find out how long does it take for the typical AngelList? Is this fast? Is this slow out how long does it take for the typical angel list to invest? Is this fast? Is this slow? How long does it take for the typical angel list investment to get to 2X and then get to 3X in terms of return multiple? I think there's a vision of this world again, where it's like, well, the typical investment is going to return about 15% or 20% a year.

And so you're going to get a noisy distribution around that. And so it'll be four years for that investment to double. And in reality, when we looked at the actual returns of these deals, the median was flat, and even the 75th percentile took years to get to a two X return. And that really it was even the question itself begged a structure of returns that don't exist for venture capital. It begged this idea that I think is common in the public markets of like, well,

the market's returning about 10% a year. So what I'm going to get is some noisy distribution around this 10% a year. And that is not what I'm saying for venture is it doesn't look like that at all. In a normal distribution, the average outcome and the typical outcome are the same. And we have become very, very used to thinking about things in that context. Like so much of our mental kind of framework for the world is based around thinking about things like that. And in power laws, that's just wildly false. Like especially for the most extreme power laws, you know, they have a well-defined median and the mean doesn't exist.

It's, in some sense, it is unbounded, right? I'm not gonna say incident, but so they're like a direct challenge to I think a lot of our implicit assumptions about how things work. Hey, we'll continue our interview in a moment after a word from our sponsors. The Limited Partner Podcast

is proudly sponsored by AngelList. If you're a founder or investor, you'll know AngelList builds software that powers the startup economy. AngelList has recently rolled out a suite of new software products for venture capital and private equity that are truly game-changing. They digitize and automate all the manual processes that you struggle with in traditional fundraising and operating workflows, while providing real-time insights for funds at any stage, connecting seamlessly with any back office provider. If you're in private markets, you'll love AngelList's new suite of software products. And for private companies, thousands of startups from $4 million to $4 billion in valuation have switched to AngelList for cap table management. It's a modern, intelligent

equity management platform that offers equity issuance, employee stock management, 409A valuations, and more. I've been a happy investor in AngelList for many years, and I'm so excited to have them as a presenting sponsor. So if you're ready to level up your startup or fund with AngelList, visit www.angellist.com slash TLP. That's AngelList slash TLP to get started. Back to the show. I think from the perspective of finance and perspective asset classes, I think what I like to say about early stage venture capital is that it is probably best contextualized as the opposite of a bond market.

Um, so in bond, like traditional bonds, not like weird high yield bonds or anything, typically for bonds, like there are bond index funds or whatever. And they, you know, they measure, you know, thousands of thousands of bonds and you have bond index funds that you can invest in that have like a few hundred of those bonds in it. And you know what? They track almost perfectly. Whereas if you have a few hundred startups versus the 15,000 startup universe, there's a very high risk you won't get anything close to tracking. In bond funds, it's actually really easy to beat the market because the whole index is exposed to the random couple bonds that end up defaulting. If you just have a bond portfolio that just doesn't have any

of that rare default event bonds, congratulations, you've outperformed the market bonds. Their price movements over time are just insanely correlated and have virtually no idiosyncratic you know there can be ratings changes perhaps but there's virtually no idiosyncratic component to the movement of bond prices over time whereas like for startups it's all idiosyncratic uh there there's virtually no market effect at all it's's all, you know, did this startup hit the milestones that it needed to, to get to the next fundraising? And you said something there.

You said a 200 portfolio startup portfolio would not get close to a 1500 or 15,000 startup portfolio. Can you explain that? It would not reliably track the market. It might outperform. It is more likely that it will underperform, but it will definitely not track the performance of the market as whole. I think it's one of the most challenging things about, especially one of the foundational things that my research has tried to show is that early stage investing is its own asset class.

When we take ideas that are borrowed or otherwise co-opted from other asset classes like public markets investing, it just doesn't work. So again, to give you an illustration, there's like, oh, I invest in a certain kind of company, kind of like I'm a sector specific investor and so I should only be compared to like, sure. If you look on AngelList, B or one of the highest performing sectors on AngelList is product called productivity tools.

And it has an IRR of I don't want to quote this, but very high IRR for productivity tools, and you might think, wow, that's like a really growing market with work from home or whatever that's like super exciting, like got to make some product, too. Really, that's Notion. It's because Notion has just blown up. I love Notion, by the way, use it all the time. Great product. It's not the productivity tool sector that's done well.

It's the investment notice notion that's done really well. That is kind of a challenge to people who are used to thinking about like, oh, I want like sector overweight or sector underweight. That's not at all what's happening here. It's these individual idiosyncratic companies. Again, this is an example of a model where we go from public markets investing and it just maps very badly onto what's actually happening in the early stage

universe. Let's assume that in synchronicity and the power laws, I think that's something that's widely at least quoted or stated in venture capital, whether it's implicitly understood or not. In terms of signals, what percentage of those power law, what percentage of those notions or Facebooks and Googles are identifiable as top 10, top five, top 1% opportunities at the time of investment at the early stage? Has your research done any on that? And also, what signals, more importantly, do you believe? And does your research show our predictors of this?

I think this has been one of the most surprising results. Because I tend to not have the highest view of venture capitalists. My research has suggested is that certainly not on the individual investment level, but on one level zoomed up, the venture market is actually relatively efficient. And I think that's very so like by that, I mean, price tends to have a meaning and like a deal that's priced at 2x. Another deal will tend to have a structure where it will return 2x the amount of money. That's what a pricing efficiency kind of means in this context. That's kind of surprising.

One reason that's surprising is that what you're buying when you're an early stage investor in eight years, when you look back on that investment, it will almost absolutely be like there's virtually no early stage investment where you're like, man, I got a solid 10% a year return out of that investment. And it was like, pretty good. It's like, it's either going to zero or it's, you know, it's going to make at least 3x, 10x, you know, 20x, 2000x, right? Like, you are either very much overpaying or very much underpaying for every single one of those investments. And yet, when you zoom up a level, it does look fairly efficient. What that means that uh it actually you know and as much

as i would like this to not be the case it actually minimizes the values of signals significantly so again to give you an example we did we did a study on founders alma maters with the performance of their startups and what we found was that actually so you know okay i'm gonna invest in like someone went to gsb or someone went to harvard or someone went to Harvard or someone went to MIT. What we found is that those founders, actually, their markup rates were higher than baseline. So it is a positive signal to invest in someone who went to GSB, but that it's not as positive as you think. Because what's happening is that the startup founders are aware that they possess this positive signal. And it's translated into higher valuations for their deals, which means which lowers the chance of future market.

And we actually think that's probably more than half of the like boon of this signal is actually captured by the founders. And so as a result, if there is alpha in deal selection venture cap, like there's a sort of naive thing, which is like, you know, the signal is still positive. It's not like GSB founder deals are so overpriced that it's actually a negative signal and you should never do a GSB deal. It's more that it's a lot less positive than you might think. This is also the case for say, repeat founders of successful companies,

a positive signal, but you're paying for that signal. That payment is what really reduces that substantially reduces investor returns. It makes it less of like a clever arb and more of like an appropriate value that you're paying for a deal. We think that if there is alpha from like signals, it comes from like, it comes from unconventional places.

So like, I believe the top three schools for markups by founders were good schools that are a little bit unconventional. They're University of Washington, Waterloo, and Brown, I believe were the top three. So it's like very good schools, but not Harvard, Stanford, MIT. So it's like a second order is where you'd find any investment opportunities. and even then, the opportunities are still relatively small, and you are still paying a little bit for that signal. But the benefit of having a Waterloo founder is a little bit better than the markup that they might charge for that signal.

The way this fits into the rank order list is that we do think this rank order list of deals exists. And from best seed deal all the way down to like worst possible deal. As you go down that list, what happens is that the signals become worse and the price gets lower. But that we think the exit possibilities also like actually fall faster than the price falls. And so at a certain point, because what we're looking at is this alpha less than two power law as a threshold, at a certain point, the exits no longer justify the price, and you actually get sort of a different asset class. And so somewhere on that list, there's a bar that says, like, everywhere above this is going to pull from an alpha two or lower power law. And those are credible deals. And they are early stage startups that can

grow into huge markets and produce huge returns if they do well. And you should have exposure to all of us. Now you could have differential exposure. You could wait more towards the highest signal ones. That's fine. But in some sense, you need to have exposure to all of those deals.

Below that threshold, the rules are very different. There's a lot more benefit of being more selective. But that's the universe as we see it. And I think the most surprising thing is that actually the startup pricing is actually at least relatively within deals, it's quite efficient. And I think just given my dim view of V vcs i think a lot of that pricing efficiency probably comes from founders who are probably very able to say like well that guy's company got a 30 million pre-money valuation but like my company is doing better so i should have a higher value

like i think a lot of that pricing efficiency probably comes from founders and not from vcs but i you know i'm also willing to say like i came into my research with a very, very dim view of VCs. And this is something where I've actually been like, VCs are not leaving $100 bills on a sidewalk. They might be leaving like $10 bills on the sidewalk, but there's no one weird trick that will make you an obvious investor that all great startups have. Instead, what you get are a bunch of signals that are sort of mutual. Everyone kind of sees them. The founders know what they mean. They're reasonably efficiently priced. They're still positive. The reason that I think this model is correct, it has kind of one implication, which is that there is one weird trick, which is there's one positive signal that

you don't have to pay for. And that's the participation of a top tier co-investor in the deal. And so if you know my model of the world is correct, what you would see is you would see a lot of venture firms whose only kind of job or only thing they do is try to sneak into deals that are being led by or priced by a top tier venture. And that is, I think the exact world we live in. I'm not going to name names, but there are, I would say, dozens, possibly even hundreds of venture firms whose first question in diligence is who else is investing. I think that's the world we live in. And I think what this research suggests is at least a mathematical justification for why. And

that's because it's the positive signal you don't have to pay for, as opposed to every other positive signal about a company. Let's say Sequoia is investing in a series A, a very high signal, I think you would agree. What are some ways that that signal is dimmed? What are other criteria that makes it less strong of a signal when a strong signal from the lead is present? I think every startup is so wildly idiosyncratic

that I mean, how big is that data set? Every startup in it is unique. So I actually think that's probably a fool's game to try to think about that. One famous fund, and there's even an article comparing Peter Thiel and Founders Fund to AngelList. In many ways, Peter Thiel takes the opposite philosophical approach and has done quite well. Are there cases when a concentrated portfolio will beat this market approach? A couple of things with that. So the first is that Founders Fund, we have them as one of the most prevalent co-investors in AngelList data.

So they're not backing a handful of founders. They're backing a lot of folks and well into the three figures annually of starting investors. We see a lot of founder funds deal. I think they're either number one or number two, depending on how we have Y Combinator classified in terms of like co-investors. They're a very, very frequent co-investor. So now that said, there is a very interesting site, I think, that I want to talk about from Ben Evans, who I think used to be Andreessen, but is not anymore. But he had this study where he looked at the performance of venture funds and found that

in the top decile of performing funds, there were actually more goes to zero than in kind of the next decile. And the idea being that a higher fraction of goes to zero. And the idea being that like the real best venture funds really swing for the fences. And I've actually heard that as justification for GPs making certain investments because it's like, oh, this is like a big bet, but we're swinging for the fences. And I disagree with that. What I think you're seeing are that top-down funds are successful early-stage investors where you have a higher risk of failure. But because they are successful, their successes look like giant home runs.

I don't think they were intended to be swing for the fences. I just think what you're observing in that top-down style are the most successful real estate investors. It is the case that if you look at the best performing funds, they will have relatively concentrated portfolios of several companies that have done exceptionally well. But that doesn't necessarily mean the best way to produce good returns for your LPs is to take a concentrated portfolio. We joke that the best portfolio is to put all your money into the next Uber. And the second best portfolio is to invest in everything, or at least have exposure to everything. So yeah, I would push back on that in a few ways. It is absolutely the best thing in the world to have a concentrated portfolio and to get lucky.

You're going to do that. Congratulations. You've done exceptionally well. But for those of us who are not necessarily banking on luck, a concentrated portfolio will tend to just increase volatility and lower returns as opposed to being a magic ticket to doing exceptionally well. You mentioned something there that's really interesting in terms of top decile and loss ratio. Have you seen some kind of almost power law distribution in terms of

loss ratio and performance? Have you seen any meaningful correlation there? Yes, but it's lower than you think. It's loss, especially given the timeline here, it of loss ratio, which is more just markup rates. So like, did a deal you made three years ago get marked up? You can assume that if the answer is no, like there's a very, very good chance that investment is going to certainly be a, maybe not a total write off, but at least on the negative end of the spectrum and not. If your investment hasn't been marked up in three years, it's probably not the next Uber. So that said, we think there is actually a surprisingly light correlation, put it maybe around 0.4, 0.5 between market break over

what would be expected and actual kind of IRR and TBPI of that fund. The differences are like GPs who are very lucky or very unlucky. They've gotten a lot of markups. So for every reason, those markups seem fairly small. And when we were doing this research originally, I actually called up on the phone. I knew the guy, like the GP that we identified on the Angelus platform as being the most unlucky GP, so having the most excess markups, like a lot of deals that were getting to series A or series B, but pretty bad returns. Not still positive because it's hard to have a lot of markups and not have

positive returns, but IRR of I think it was like 10% a year net LPs. I actually called the guy up and I'm like, hey, so you were identifying your data. Are you trying to hit singles instead of swinging for the fences? He was like, I'm incredibly surprised to hear that. All the companies I'm investing in, I think could be 1,000X returns. I haven't been super happy with the returns we've been getting. I guess the markups have been nice. Obviously, this is an anecdote, but I think it was telling that I do think the structure exists. There is a correlation. It's hard to have a lot of markups and not make money. And it's hard to have a lot of write-offs and at the same time make money. So there is a correlation, but it's weaker than you'd think. And I actually think both of these

are informative for future returns, both current returns, as well as how frequently your deals are marked up or written off. So what is the correlation between TVPI and DPI? In bear markets, DPI has become popular again. How would you characterize that? I can't talk about it with Angelus. We have such a small DPI, I think. For what it's worth, the platform is huge.

So we have I mean, I feel like I should give a plug. AngelList has distributed more than a billion dollars to its LPs, but that's a small fraction of the money that's been invested on the platform. Like, we don't have a long enough data set. We don't have enough exits. One of the challenges with AngelList data, I'm just telling you, it's a weakness of our data, AngelList really started humming humming maybe 2018, 2019 in terms of just being able to say, oh, this is actually the venture market. And so when you talk about exits, it's like you're looking at companies from the 2015 era of AngelList, where AngelList was capturing maybe 10% of the market.

So 90% of those exits are not reflected in AngelList data. My answer to that is check back in a decade, and we should have really good data on distributions and stuff. It's funny actually, for what it's worth, it's not first question I've gotten about DPI. I think one of the access fund faces joked like, hey, if you're looking for cash on cash returns, early-stage venture is not the place for you.

You should go buy a multi-family apartment building outside of Houston. if you're looking for cash on cash returns, early-stage venture is not the place for you. You should go buy a multifamily apartment building outside of Houston. But I don't know if you've seen the news lately. Actually, it seems like multifamily apartment buildings outside of Houston are not doing particularly well right now. So I do think there's no sure thing, and what the world looks like has changed a lot

over the past few years. So tell me, a lot of people asked me to ask you about the Quant Fund. What could you say about that? Yeah, I think this is a good illustration of kind of the shape of the way that we're doing investing. So the Quant Fund is using most of the money is going towards investing in the companies that are hiring the best on what is well found, what used to be AngelList talent in terms of tracking applicants, like top applicants. And when we were backtesting this, like one of the cool things is that we can do like backtests, which is very rare for venture. So who are the companies that we would have wanted to invest in in the past? And we identified like Bolt, Calm, Chime, Divi Homes, DoorDash were all companies that were at or near the top of this list years and years ago when they were early stage companies.

And so we're like, hey, this seems like a pretty auspicious signal. I also think, you know, how does this fit in general with the model? Well, a few things. The model of the world is that as an LP, you should want exposure to every credible seed stage deal. Where credibility, I can tie it back to that alpha-lesson-two-power-law thing on the rank order list. The challenge that you could give to that credibility threshold is that word credible is doing a lot of work.

What it essentially means, does that mean that GPs should just broadly spray and pray? No, they should invest within their area of expertise to be able to know what a credible deal is. And that I think is what distinguishes Seed as an asset class is that it's very easy to know what a credible series A is, right? Is there a professional venture capitalist taking a board seat as part of an equity financing? That's a credible series A.

For Seed, it's like two dudes, a slide deck, they've kind of been working at it for six months, company might not even exist yet. It's like, is that credible? Is what they're talking about credible? Is this credibly going to be a billion-dollar company? That requires specific expertise. I think what the results suggest is that LPs should invest broadly among a number of GPs, each of which have their own kind of specialized area of expertise to be able to evaluate this credibility threshold. And that's the model of this world. So how does this play in the quant fund? Well, we think that we have this interesting credibility filter, which is looking at this kind of inbound hiring signal,

and we are investing broadly within it. We write really small checks, 100, 150K, and we're trying to invest in as many companies as possible. 100 plus companies with these small checks. That is the game plan of the quant fund. We think we have this little edge from our proprietary data, and we are investing in such a way that is compatible with the way that I think that the math speaks about what the world looks like.

And you mentioned credibility. Is it partner versus fund? Would you go with a top decile firm or a top decile partner that has left to found his own firm? I can answer this from a data perspective and then kind of from a startup founder perspective. So from a data perspective, I think the preponderance of evidence leans towards partners. Just to give a good illustration, I want to give a shout out to Jonathan Shue of Tribe Capital, who's very helpful actually when we were talking about the early steps of the Quant Fund. It was very encouraging. Jonathan is a great guy.

He left Social Capital to start Tribe. We have Tribe as one of our best co-investors. In terms of our own internal rankings of co-investors, we have Tribe as one of the best ones. I think there are some other examples that might be able to less closely correct that I don't want to mention. I think the preponderance of evidence is towards partners and not firms on the data side.

That said, as you know, wearing a startup founder hat, one of the, for, especially for an early stage company, one of the best values that a VC brings to a financing is that you can put their logo on your webpage and prospective employees will view your company as much more credible because you're backed by a venture firm that they may have heard of. And so from that perspective, maybe it's firms. I think it's a little bit of a balance, but I would encourage startup founders to not be afraid of going for founders. I do think it's the sense with going with partners and that there is a sense in general that I think, in general, AngelList of course

is a little bit taught here about here, but we think solo GPs or small partnerships tend to outperform larger partnership megafunds. You said something very interesting there about solo GPs versus traditional incumbent funds. Is that going back to the signal? Is it that solo GPs versus traditional incumbent funds. Is that going back to the signal? Is it that solo GPs have less capital to deploy so they invest at a lower valuation? What is some potential rationale for that phenomenon? Part of it might be that solo GPs

tend to hustle harder. They may also have, again, I think it goes back to this idea of the ideas that an early stagegp needs to be about be able to evaluate credibility within their like specific domain of expertise i think it's probably also the case that for a solo gp versus a larger partnership especially like a more junior partner like they might know that a company is a good company but not be able to get it through the partnership and similarly they may you know assent to deals where they don't really this is not my space, but if this other partner says it's good, I guess I have to go along with it. Or there could be horse trading within the partnership for various investments.

I think the solo GP structure tends to just clarify it. It's like, do you have an area of expertise where you can adequately assess credibility? And I think one of the reasons you see the solo GPs tend to overperform is that the answer to that is like, yes, because chances are, if you don't know what you're doing, but also think you know what you're doing, you're not going to go out and raise your own fund and do that one thing that you think you're doing when you can have a sort of a safer partnership structure in place. I also think there's probably something with early stage investing where I mean yeah it is smaller

checks but it could just be because it's earlier stage. I think there's something with earlier stage where in general like a partnership structure is hard to make work well because you are looking for fairly disruptive companies. And so anything that maybe tends towards consensus, our consensus view could be really limiting to that. So I tend to really like solo GPUs or small partnerships. I think their performance is generally quite good. I think it's a good fit for the model of,

can you assess credibility of a deal that might not even be a company yet. Zooming out, you've been published research for a long time. How have you seen the LP community or broader ecosystem evolve on these topics? At the time, it was a very contrarian view. Have you seen them kind of accepted a bit more or how have you seen it evolve?

I'd like to hope that the idea of going broad for early investing has been more accepted for LPs. That said, for the quant fund, it was hard to raise that fund. And actually, we struck out with some, but not many conventional venture LPs. And actually, our largest investors and frankly, the easiest checks to get were from Quant Hedge Funds or principals at Quant Hedge Funds, where the idea of you pitch them and we're like, we invest based on this proprietary signal. And they're like, that's what investing is.

Of course you do. That's the nature of what an investment means. And they're not traditional venture LPs. So I don't know if my research work has resonated enough to make that fundraise easy. And certainly, I do think there's also a sense of, even with saying like spray and pray, right? The idea, and I brought this up in the startup growth and venture returns paper,

it sounds a lot better, right? To say, hey, I'm going to use my network to find five to 10 startup founders who are like really, they're going to change the world with their narrow set of businesses that I'm super excited about. And I'm going to work closely with them. It sounds like you're being a better steward of LP money than if you're just like, we're going to blast off so many investments, I'm not going to

know what these companies do. One of these things sounds like a bespoke artisanal, like really kind of thoughtful way of managing somebody's money that frankly justifies carry. And the other way sounds flippant and unserious in some sense. And I think what's still provocative is that I actually think the latter way is a better way to do real estate investing. And so that is, you know, I think the biggest mistake that LPs can make is not being exposed to enough companies. I know one of the big changes I've had as an LP putting, you know, a third or fourth hat on for this interview since I started AngelList is I am now an LP in several

venture funds. Whereas before I was like, well, I do my own early stage investing. Like I don't need to pay someone carry for deals. Like I don't want to do that. I was now just like, if there's differentiated, credible deal flow that someone will give me, I will definitely be an LP in their fund. So, for instance, one of the funds I'm an LP in is Ryan Hoover's weekend fund. Ryan Hoover is the product founder, and he is an exceptional high end brand within this space of kind of of fun consumery companies. That's something I don't know anything about those companies. I don't see opportunities to invest in them.

Then the ones that I do, I can't assess at all. I have no problem paying Ryan Carey to do a good job of finding and investing in those deals because it was literally not something I could do myself. At the same time, I think that those deals, that's playing in the deep end of the pool, that's playing in the credible seed deal threshold area. I think it certainly changed my behavior, this idea of trying to be broadly exposed.

I think it will be a matter of time before it changes behavior more broadly, but I do think it's the way things are going. There is one hack with the broad investing as well, which is that typically when someone discusses their prowess as an investor, the way you see that prowess is by listing their number one or top three investments that they've made. And if you've invested in a ton of companies, your top three will tend to look quite better than someone who's invested in 15.

And so there's like if you want kind of a life hack way that this kind of broad investing is likely to possibly take over as a modality, it's that it's the hack around this like thing where you don't see the 97 failures. You just see the three huge unicorns. And therefore, you're an awesome investor. And therefore, people want you on their cap table. Another variation on that hack that's well known is investing right before the IPO and saying, I was a private investor in this company. That's a common one in BC.

You mentioned Ryan Hoover and his consumer expertise. How much of a signal and what has the data showed you in terms of somebody's domain of expertise with Ryan or the hypothetical GP that's really good in one sector, be really bad in another sector? That's a great question for future research. I think that's actually a blog post that we should do on AngelList by the end of the year is to see if there's sector effects. I can tell you that the early stage investing that my friends and I do is through a partnership called Indicator.

At our portfolio reviews, after years of investing together, we now joke that we make three kinds of investments. We make deep tech investments, we make f FinTech investments, and we make bad investments. And the funny thing is that we continue to make bad investments. They look really good. And we're like, this one's going to be different. And then almost inevitably, they end up in the loser column. And so it's kind of been like a honing of our area of expertise.

So I mean, just based on my personal lived experience suggests that sector effects are 100% real. Whether or not we have good enough, fine-grained enough data to tease those out on AngelList is another question. Because I think it's like the sector labels can be hard. It can be fuzzy. Startups can change what they do. Are you enterprise? Are you B2B? Are you SaaS? I think we all have those different tag labels. That can be tough. But I do think it's a super interesting question. And my guess is that, yeah, I will say that I think that is something from the public market that does carry a relation to hedge fund investing that carries over to GPs, which is uh style drift is a known real negative thing and i would imagine in hedge fund investing and i imagine that's a negative thing yeah venture and i mean you see those effects as

well right like are most hedge fund managers style drift because they do really well in their narrow area of expertise and suddenly they have they have to manage all this huge money and they're like they become macro investors making bets on the price of oil whereas before they were investing in stage two biotech companies or something. For GPs, I think you see the same thing. You have a fund that's a small fund that's only investing in this narrow area and maybe they scale up and raise a ton more money and then suddenly they're doing all sorts of wacky stuff. I would not be surprised if that analog to the public markets did hold up in venture investing the data.

Speaking of style drift, AUM drift is a big issue in venture capital. How do you look at that and how do you expect that to change the signal and the nature of returns in terms of power laws, also in terms of valuation and stage? Maybe you could talk a little bit about that, the growing AUM and the growing valuations. It has been, just to maybe slip it around a bit, it's been one of the, and it's hard because there's real people with real dreams and real

money being invested, but it's been frustrating to see in the data how I would describe it as stubbornly, persistently high early stage valu evaluation has been even over the past kind of crash. And I think we have a piece on the blog that's going to be coming out soon. But up until second quarter of last year, the median price equity seed round was down to 20 million pre money. And we think that and now it's at, like, 17 million pre money. And we think that and now it's at like 17 million pre money. And we think that needs to go down to like 12, 12 and a half for it to be like a sustainable recovery and to return.

Like, I think that's where we're going to end up. And I my sense is that the glut of money that is present is what is support. Like I think there's a couple of things like whether it's structural or not, a glut of money that is present is what is support. I think there's a couple of things, whether it's structural or not, a glut of money that's going earlier stages might be supporting these valuations and keeping them artificially high. But that said, we also think that there's

going to be over the next year a ton of pricing pressure, at least seed, now that Series A investments have been moved in price a decent amount. If you do a lagged, what were seed prices a year ago versus series A prices now, that like ratio peaked a year ago at like 6x. And I think that encouraged a lot of price inflation at seed because, you know, investors could just be like, well, yeah, we can double the price of this and we'll still make 3x. So it's not a problem. Now that ratio is like at 2X and still falling. I think that will encourage pricing discipline or may encourage pricing discipline on the

part of earlier stage investors. What happened with SoftBank? Maybe that era of mega fund is over and a scientific experiment was run and it came up with a failure. That to me is like, I can't be completely sure what the future looks like, but my sense is that the prices still need to come down, that the presence of folks who are keeping these prices up is just serving to delay

the inevitable pricing recognition because I do think it's like prices just need to be a certain pricing recognition because I do think prices just need to be a certain amount to accommodate risk appropriately. I think the idea of SoftBank, the SoftBank vision fund style things is probably done for an investment generation. I'm sure in 10 years, Masa will be back with future vision fund that will be even bigger. But I think that's done for the near term. I think a lot of people misunderstand the SoftBank thesis, which was capital as a moat, which is an interesting thesis. In retrospect, every VC is a genius and could predict every potential outcome. But I thought it was bold and interesting, although obviously, ultimately not successful.

As a concluding question, what do you wish people knew about AngelList? And what are some misconceptions? What would you like to resolve and for people to know about your platform? What's out there tends to be pretty bad. I happened to look at the AngelList Wikipedia page recently and I was like, this is not, this does not at all look like what AngelList is. I've gotten that it's a, most people are familiar with it because the jobs platform, obviously not anymore. And then if they're familiar with it beyond that,

they're familiar with it as an LP where they had to click through a bunch of screens and maybe it pulled an ACH and they're banking out. I think AngelList's actual clientele are GPs and founders. I think that's probably the best analog to what AngelList actually is, is like a financial services company that happens to have this startup thing attached because for whatever reason, like real companies, when they need to raise debt or equity, they talk to bankers, but startups, the founders do their own banking.

And so it is like a financial services company that has a very vigorous founders and startups arm. And I think that's probably the best analog to what AngelList is right now. And yeah, I think it's really hard to see that. I think certainly from the outside, because those products are like, most people don't know what a prime brokerage for a hedge fund does.

Like that's just, the people who need to know that know exactly what a prime brokerage is and who they use for a prime broker. But it doesn't affect people's day-to-day lives on the street to know what that is. And so I think what's weird about AngelList, I think one of the gaps is that like for a long time, you know, most people only saw this as like the thing that touched the most people was the jobs website. And the thing that touched the second most people was the investment kind of LP

closing flow slash investment dashboard. And then what AngelList actually is, is something kind of different that virtually nobody actually sees. And that I think is the cause of kind of the maybe misperception of AngelList in the market. Great. I'd also add, you guys have a great GP fund admin platform in many ways,

streamlining a lot of the pains of some of the most famous and biggest services platforms out there. So Abe, it's been an absolute pleasure uh it's great to catch up and always chat with another quant nerd such as myself thank you for elucidating a lot of interesting things a lot of things that keep me up at night uh and that i i've always wanted to know and i look forward to catching up live soon thank you so much david great questions enjoy the interview a lot. Thanks for listening to the Limited Partner Podcast.

If you enjoyed this conversation, please like, subscribe, and share with your VC and LP friends and anyone interested in learning more about the venture ecosystem. Please follow Eric and I on Twitter at Eric Tornberg and D. Weisberg, linked in the show notes below.