E453: Quantifying Venture Capital, Finding the Next Anthropic, and the Science of Picking Winners

What if you could identify the next billion-dollar startup before its founders even approach investors? In this episode, I sit down with Jerry Ye, Partner at BITKRAFT Ventures, to explore how data science is transforming venture capital from an intuition-driven business into something increasingly measurable. Drawing on his background in machine learning, large-scale data mining, and quantitative investing, Jerry explains how investors can identify promising startups through employee movements, credit card transactions, customer retention, and alternative data.

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00David Weisburd?

  This anthropic signal before Anthropic became anthropic today, what exactly were you seeing in the data?

• 0:05Speaker 01

  Anthropic was hiring people with more in terms of like a pedigree type signals, like better schools, better experience and on average that actually had them surpassing OpenAI.

• 0:16David Weisburd?

  A lot of venture capitalists have been trying to quantify which founders are going to be most successful at the very early stage. What do we know today about what kind of founders succeed?

• 0:26Speaker 01

  This really depends on the sector. I remember like an early study that showed PhDs in enterprise tended to have better outcomes. So domain experience, if your particular niche or your sector requires that domain expertise, but that doesn't really matter at all for consumer companies, where you might have a very strong pedigree founder and team, but you don't have traction and traction in that case is actually super important.

• 0:49David Weisburd?

  Let's say that you're advising a CEO of one of the top multi-stage venture capital firms, one of these top 15 firms. And the CEO was bought in of the firm and they wanted to institute data in the deepest level within their firm on every part of the staff. How would you advise them to execute on that strategy?

• 1:08Speaker 01

  You need KPIs, like what a true north star is. And it has to be objective that's in a way that's grounded by data.

• 1:14David Weisburd?

  One of the interesting factors that we see in the startup ecosystem today is these tranch rounds. Sequoia will come in at a billion dollar valuation and then startups will raise another round at $2 billion almost a couple of weeks later. What do you think about these tranch rounds?

• 1:30Speaker 01

  It's pretty interesting.

• 1:35David Weisburd?

  Jerry, you've spent most of your career trying to make venture capital quantifiable. What have you learned?

• 1:40Speaker 01

  Early in my career, we spent a lot of time looking at the talent ecosystem. So trying to understand where the most interesting founders were coming from, and who they were able to attract in terms of talent. So starting off there was just mostly looking at it from a who perspective. And later on in the career, I spent a lot more time in terms of trying to understand traction data, mostly focused on consumer, prosumer type companies. And there is mostly trying to understand product market fit from the outside in. So buying alternative data, credit card data, whatever panels that was actually pretty familiar to the hedge fund side, but net new to the venture side.

• 2:22Speaker 01

  And that gave us a lens into what was essentially trending and what was seeing inflection, was seeing high retention from the outside in before we even reached out to the companies.

• 2:34Speaker 01

  And then the more recent part was kind of like, you know, bringing in all these other signals that other firms were working on in terms of like quantifying growth. Like social capital did a bunch of work there and being able to do that on data room. And once the company was actually in front of us, trying to make better decisions or get faster to time the decision.

• 2:55David Weisburd?

  Start with talent. A lot of venture capitalists have been trying to quantify which founders are going to be most successful at the very early stage, which is where you focused on. What do we know today about what kind of founders succeed?

• 3:10Speaker 01

  This really depends on the sector, right? And I remember like an early study that showed PhDs in enterprise tended to have better outcomes. So domain experience, if your particular niche or your sector requires that domain expertise.

• 3:28David Weisburd?

  Something like a very hard and technical problem.

• 3:30Speaker 01

  Yeah. And this is a reason why if you look at NERUP, right? I think from 15 years ago, it was like a small niche, little academic research conference for 200, 300 people. Now it's talking about 20, 30,000 people, right? And VCs are constantly there on the conference floor trying to source and identify the next deal. But that doesn't really matter at all for consumer companies, right? Where you might have a very strong pedigree founder and team, but you don't have traction, right? And traction in that case is actually super important. So it really depends, right? And it's not just the founders, it's also who you're able to attract after you start the company.

• 4:12Speaker 01

  So your founding team, the people you're able to pull away from the Mac seven, right? How you're getting these additive people? Because one way that I like looking at it is similar to a venture round, right? Each engineer or each product person you're adding that could have made an extra 100K, 200K per year at a Mac seven could be seen as like a mini round, right? And in aggregate, that's actually a very strong signal in terms of conviction or product market, or in terms of conviction for the company.

• 4:44David Weisburd?

  So another way, they could have gotten a higher salary at a Mac seven firm. Instead, they're foregoing that salary for stock. So they're betting with their own career and with their own time into that startup. And that's a very strong signal.

• 4:58Speaker 01

  Yeah, I mean, that's exactly what venture capitalists are doing as well, right? They are, you know, investing their capital for carry or stock going forward. And this is the same thing.

• 5:09David Weisburd?

  The difference there, you could argue, is that venture capitalists maybe have two, three, four meetings with the team. And somebody that's quitting Google to go join a startup is probably doing a lot more diligence. It's probably talking to a lot more friends that work at this company. There's much higher level of granular diligence being done by the employees joining than by the actual investors.

• 5:30Speaker 01

  Yeah, and especially the ones that are more connected or the ones that have already had multiple outcomes. And that's, you know, like going back to the talent signal, right? That's probably what you want to consider as well as you're identifying, you know, who's joining the team and kind of quantifying each individual that joins. Because someone that joins fresh out of college versus someone with multiple unicorn outcomes is a very different signal.

• 5:57David Weisburd?

  This episode is presented by Juniper Square, the operations partner for private markets.

• 6:04David Weisburd?

  Perhaps it's obvious, but why the difference?

• 6:07Speaker 01

  They have a lot more choice, right? So it's also similar to venture, like who's on your cap table, right? If you have a top tier venture fund, that's, there's a signal there. And it's also the same as if you're able to attract someone that, you know, has already had an IPO at Facebook and then decided to go to OpenAI and now they join your startup, right? That's saying something.

• 6:29David Weisburd?

  This concept of this talent magnet, Elon Musk famously is the greatest talent magnet. He's able to just recruit the very best. At some point, I think he had five of the top rocket engineers at the very early stages of SpaceX. And this has become this consensus view that if you could attract talent, arguably even more important than venture capital. Do you think that it's the fact that these engineers are making the best choices on the companies that's driving that? Or do you think that the engineers themselves, the top engineers are the ones that are building these amazing companies?

• 7:00Speaker 01

  A little bit of both, right? So the team definitely makes a huge difference. But the company itself, culture wise, also matters too, right? Like you could have a company that's attracting top engineers or top product people and they're churning employees like crazy, right? So that's actually another signal that you might want to look at as well. And another example of this is, I think, if you just looked at LinkedIn data or talent data, that a couple of years ago, you saw that Anthropic was probably on average attracting more interesting talent than OpenAI, right? And if you look at it from a churn perspective, and it might be a reflection of the valuation increases now,

• 7:44Speaker 01

  but if you look at churn, Anthropic is churning engineers at a lot lower rate than OpenAI. And so it is a combination of the people and who they're able to attract and what they're able to build, but it's also a reflection of the company and the culture, what they're able to accomplish.

• 8:02David Weisburd?

  Take me back to this because you noticed a signal, this anthropic signal, before Anthropic became anthropic today. Tell me about that. What exactly were you seeing in the data?

• 8:13Speaker 01

  So I think anyone that was looking at talent signal a couple of years ago would have seen that on average Anthropic was hiring people with more, in terms of like a pedigree type signals, like better schools, better experience and on average that actually had them surpassing OpenAI at some point, right? So I have to caveat that this is kind of like an art versus science here. Because, you know, like how much weight do you give to like a senior manager versus a junior person, right?

• 8:49David Weisburd?

  So you're aware of this inflection with Anthropic having better talent before OpenAI? Maybe you could double click on that. Quantify, what exactly does that mean?

• 8:59Speaker 01

  In terms of what you would see on a profile of a person in terms of resume, it's, you know, what schools they went to, that's a signal for selectivity. What other companies they worked at, where they went to, right? How, how many people went there, what churn looked like? And by essentially doing that as well as, you know, understanding.

• 9:19Speaker 01

  How much capitalization those companies had was the outcome, where were the founders, et cetera. You can start putting together a formula for how to rank people and one that's more technically focused in terms of engineers will show that Anthropic was leading OpenAI three, four years ago.

• 9:38David Weisburd?

  Obviously, talent is arguably the most important factor when it comes to the success of a startup. What's right up there with talent?

• 9:46Speaker 01

  Talent is definitely super important because it spans all sectors and especially for like enterprise, that might be the only thing that you have. But for a lot of other sectors like consumer or gaming, fintech, prosumer, there's actually a lot of signals from alternative data panels, especially the credit card data, web traffic, app traction that you can leverage to identify traction. And some of them are actually so granular that you can actually figure out what retention rates look like as well.

• 10:16David Weisburd?

  What exactly are you looking for when it comes to revenue?

• 10:19Speaker 01

  It might be helpful to take a step back in terms of what I'm talking about for, let's say, like card panel data, right? So there are data providers out there that basically act as like the interconnect between your banking account and your brokerage. And they, as a side business, resell this and a lot of hedge funds buy it, right? Sometimes this stuff is actually real time and hedge funds will pay seven, eight figures a year for this. Venture on the other end, because the time horizon is 10 years, the ROI is not the same, will be happy with like a delayed panel, right? But you're literally talking about a three to 5% sample of all US credit card and debit card spend, right? And with

• 11:03Speaker 01

  that, not only are you able to figure out, you know, how the corner Starbucks is doing, like who's coming back, how, what the retention rates are, you're also able to figure out what, you know, OpenAI is looking like in terms of revenue on the consumer side. These are all generally consumer type card panels, but there are some that kind of focuses on business transactions as well. And for the consumer side, what you can do is try to identify the companies based on the credit card descriptors, which are similar to what you see in your credit card bill at the end of the month. And for these panels, they generally match the stuff with companies with stock tickers or late-stage venture

• 11:48Speaker 01

  companies.

• 11:49David Weisburd?

  So hedge funds are using this to figure out earnings data before they're announced by the companies?

• 11:54Speaker 01

  Yeah. So interestingly enough, I remember going to a meeting here in New York that this was almost like a decade ago and it was, it was like a dinner with a bunch of hedge funds and they basically concluded there was no more alpha from credit card data. That's just like how far advanced they were, right? Whereas still on the venture side, just because a lot of this stuff wasn't tagged, right? Because it was only tagged for the long tail, I'm sorry, the, the head revenue, the long tail was not tagged. So there's actually a lot that goes into it that, you know, you can do any resolution to figure out, does this square, is this square, square cafe? Or is it like square? Or is like, is there actually like a sub merchant under there? And it's still largely

• 12:38Speaker 01

  untapped. You can't feed this into your LOM, like you'll blow up your context window immediately, right? And

• 12:45David Weisburd?

  And that's because there's so much, just, just so, so much data.

• 12:49Speaker 01

  Billions or trillions of data points, right? And so there's a little bit of work there. But if you do the, the grunt work there, you can identify, you know, like up and coming companies, like what is inflecting, what, what has high retention, right? And a lot of times you could figure out these companies before it becomes like a cultural phenomenon or before the media picks up on it or other VCs pick up on it.

• 13:14David Weisburd?

  And the majority of venture firms, the large firms are using this data?

• 13:18Speaker 01

  I think a handful are, right? Especially the larger platforms. The crossover funds are all definitely doing it, right? They had eight-figure data budgets with 20, 30 people, engineering teams. A lot of the other ones, even on the venture side, large platforms may not be doing it. And all the smaller ones, like it's, it's a data set that's actually pretty expensive. So they might work with, if they're like a partner on this, where, for example, there was Second Measure before and they tagged a lot of companies. Most of their clients were venture-backed, were, were venture funds, but most of their revenue

• 13:59Speaker 01

  was actually from the hedge fund side, right? So they ended up being acquired and kind of deprioritized the long tail tagging. So it's still something that a lot of firms would have to do on their own.

• 14:10David Weisburd?

  I love the disciplinary nature of the show because I get to talk to public investors, private investors, and sometimes these things are so obvious to public investors, you said, with hedge fund managers. It's so obvious that there's no alpha there. There's, there's no more juice to squeeze in the public markets. Most venture capitals haven't even heard of this.

• 14:28Speaker 01

  When I first went to my, the first alternative data conference in New York, and I looked around, I was literally the only one representing a venture fund there, right? And you saw everyone else on the hedge fund PE side. And it is, it is fascinating, but a lot of it is also in terms of what the profile of the teams look like, right? If you look at venture, most people can, you know, handle looking at, you know, their data rooms in Excel, but being able to, to figure out signal from a data set with billions of data points is actually difficult, right?

• 15:05David Weisburd?

  Perhaps it's obvious, but these venture capital platforms, the large platforms and the crossover investors, they're getting this data because it's essentially ground truth. Because if you know what the consumer and the enterprise is spending on, on their credit cards, you know exactly what the next top company is.

• 15:21Speaker 01

  So it really depends on the type of company, but if it is a consumer-facing company, let's say like an online D2C subscription company, you should have really good visibility into it. And actually similar to polling, what's going on right now, right? There's only a percentage of the population you actually need to poll. And we're talking about a three to 5% sample of US credit card, debit card spend, right? So depending on the size of the company, you actually have a really good idea of some companies wants to hit a certain, you know, like run rate or a certain stage.

• 15:59David Weisburd?

  And who's selling the status? Are these data brokers buying from Visa and Mastercard?

• 16:04Speaker 01

  So these are generally interconnects between like your bank and your like brokerage. If you ever sign on and it says like connect your other account for like external transfers, et cetera. Plaid is definitely not.

• 16:16David Weisburd?

  Shout out, shout out to Plaid and non-sponsor.

• 16:18Speaker 01

  Yeah, it's actually not Plaid. But there's, it can't be Visa or Mastercard because their agreements are actually with the merchants, right? And the merchants definitely do not want, you know, them selling this kind of data. There are panels like Yodlee or Consumer Edge. This is all, you know, public information.

• 16:38David Weisburd?

  And where do they get their data?

• 16:39Speaker 01

  Yeah. So they, they form the interconnects for a lot of these companies, right? Some of them actually buy smaller banks, smaller card issuers, et cetera.

• 16:47David Weisburd?

  So using this card data, Signalfire, invested in Grammarly. Tell me about that.

• 16:52Speaker 01

  So as part of the, the panel, right? And just keep in mind that you can figure out what retention rates look like as well. But we saw a company that was inflecting and not only was it inflecting, you saw really high retention rates, which were similar to what you would see in a SaaS company. And that led to investments there. But what was interesting there was actually, and this kind of ties into what I'm working on now too, is before, from the outside in, that led to a investment in the company early on based on cohort performance. And now, Grammarly actually, a couple of years ago,

• 17:32Speaker 01

  ended up receiving a billion dollar user acquisition financing facility from another firm. And they're essentially leveraging those same cohorts, but for user acquisition, that's non-dilutive, right? So it's not an equity investment. And for them, it's actually a short-term problem. And this is a short-term solution for it because the revenue comes back in 12, 18 months, right? And they don't have to sell a chunk of their company for it.

• 18:02David Weisburd?

  This touches on One of the most underrated and least talked about forms of funding for startups, user acquisition financing. Tell me about that. What is user acquisition financing?

• 18:13Speaker 01

  For companies with post-product market fit, there's actually a solution now where it's non-dilutive. It's not equity. It's not venture debt. And for these companies, they know that every dollar they spend on UA turns into $2, $3, $4 down the road. But it takes six, 12, 18 months for that capital to come back. So even for the most well-capitalized companies, they quickly become cash flow constrained. What this does in place of equity is because this is a short-term problem, is a short-term solution where we fund a large portion of their monthly spend in exchange for a capped revenue share. And that's paid back as the cohorts

• 18:57Speaker 01

  generate revenue. And once the principal plus the fees are paid back, then all the residual revenue actually belongs to the company. So you have high ROI for the company. It actually makes a ton of sense because it's essentially just pooling growth forward. And I mean, there's obvious downsides as well. It's like if we mispredict the paybacks or whatnot, there is risk to this asset as well. But as long as the company is hitting performance thresholds and thresholds are set when we make a monthly investment, like they have to have 40% ROI by the first month, 50% by the second, et cetera. If it ends up underperforming, then we actually

• 19:39Speaker 01

  pause funding for the next month. So instead of like a 10-year time horizon, we're talking about investing each month separately, but on average, they pay back seven, 12 months.

• 19:52David Weisburd?

  One thing I've learned from talking to hundreds of investors is that great investment firms aren't built on investment returns alone. The firms under are great at the things most people don't see, their operations, their relationship with LPs, and the quality of information they use to make decisions. And here's what AI has changed. Every firm now has access to the very same models, so the intelligence isn't the edge anymore. The edge is what you could feed it. A firm with its fund operations and data in one connected record can actually put AI to work. A firm running on disconnected systems simply can't. That's why thousands of GPs run their funds on Juniper Square. Juniper Square puts your fund operations, data, and administration together in one connected record. That means

• 20:33David Weisburd?

  less time managing disconnected systems and more time investing, working with LPs, and building your firm. This episode is brought to you by Juniper Square, the operations partner for private market GPs. Learn more at junipersquare.com/HowIInvest. That's junipersquare.com/HowIInvest. Give me a sense for how many startups are using user acquisition financing today? How big of a market is this?

• 20:59Speaker 01

  It's still very nascent, right? So there's another large venture fund that has raised $4 to $5 billion just for this asset.

• 21:07David Weisburd?

  Is that General Catalyst?

• 21:08Speaker 01

  Yeah, General Catalyst's Customer Value Fund. And they not only work for startups, they also work with public companies, right? So like Lemonade for the longest time, Fiverr. They've invested $150 million into Lemonade. It was like $250 into Function, I think, and a billion to Grammarly. So these are actually really large checks for this product.

• 21:36David Weisburd?

  Have you heard of Henry Shi and the Startup Leaderboard?

• 21:40Speaker 01

  I have not.

• 21:40David Weisburd?

  So Henry Shi, previous guest, he started this scoreboard, which is the Lean Startup, basically scoreboard, where he tracks the top companies that have raised very little capital and have grown. Famously, you had Mid Journey. I think they had something like a dozen employees and $10 billion valuation at some point.

• 22:04David Weisburd?

  I'm wondering how AI changes the fundraising game for companies. In other words, what are the downstream consequences of having smaller companies with high revenue? Does user acquisition financing fit into this new way of funding startups?

• 22:21Speaker 01

  I mean, if you think about a company, right, a lot of their, especially for consumer companies, once they have product market fit, a lot of, if not their largest line item is UA spend. So, and as I mentioned, right, for those types of companies, it's a highly predictable problem in the short term. So does it make sense to fund that with like 10-year capital at a cost of a 10-year capital structure? Or is it better to relegate the cost of your data center or your token spend, et cetera, for that side? But for UA, this is a lot cheaper capital. So you can potentially have companies, especially on the consumer side, that leverage

• 23:04Speaker 01

  and lean into this type of structure and are able to increase valuations quite significantly, especially in like, let's say, the gaming space where fundraising has been depressed over the last couple of years. For example, in the gaming space, where fundraising has been difficult, and there are companies that have leveraged UA spending to scale their revenues 3, 4, 5, 10x, right, over the course of a year or two. And that now has opened significantly, opened a lot of doors in terms of being able to raise the next round.

• 23:37David Weisburd?

  Yeah, I was going to ask you about that because the downside of having companies that I just read that pre-seed rounds are typically now at $100 million valuation. And I guess seed rounds, that means, are $500 million, $1 billion. Why even do user acquisition financing if you could go and raise equity? I guess your answer would be that not every sector of the economy is this red hot sector today.

• 24:00Speaker 01

  Yeah, not every sector is that red hot. But the other side is, it is still coming at a cost, right? Even at that, at those valuations as well, right? So if you just price it in terms of, well, what if you use UA versus not losing that 20% chunk of your company? And you do hit your revenue goals, that's actually pretty expensive.

• 24:21David Weisburd?

  And talk to me about venture debt. So venture debt has been around for many decades, since really the earliest history of venture capital. Why is venture debt not better than user acquisition financing?

• 24:35Speaker 01

  A, it's not necessarily available for all companies, right? Especially as you go earlier stage. For example, if companies that have $10 million AR, you're not going to get venture debt offers generally. But B, it's usually on a fixed repayment schedule. And user acquisition financing is actually where the revenue is coming from revenue share. So we're not getting paid unless those users actually generate the revenue as well. So it actually matches when revenue is actually generated and when the company actually have the funds to repay, right? There's also, as we discussed earlier, it's non-recourse as

• 25:15Speaker 01

  well, right? So it is a lot more founder-friendly. So we're not going to end up with a sizable chunk of the company because you underperformed for a quarter or so.

• 25:25David Weisburd?

  We talked earlier in the podcast how you spent your entire career making these very art-like aspects of venture capital, quantifiable. What percentage of being a great venture capitalist can you quantify today?

• 25:39Speaker 01

  It's a very interesting question. And I think initially when I started, I would have said that it's difficult to quantify the largest portion of that equation, and that is access and the ability to be able to get into the room of a company. But what I've seen over the past few years, especially talking to other peers at large firms or even small firms, there's a common exercise that all venture funds do, or I hope they do, is you're actually looking at deals that are done, right? Deals that are announced on TechCrunch, which shows up on Crunchbase PitchBook. And you are triaging whether you were actually

• 26:22Speaker 01

  in the room or not with the company, right? And at what point in time, right? So if you look at it from funding announcements, like where they actually connected with the founder at some point? Was there emails? How many emails? Was there a meeting? How many meetings? Was there a term sheet? Did it go to IC, partner meeting? What did they decide, right? And all of that stuff is actually in the CRM, right? The thing is, most firms kind of see it as like a to-do list of, you know, this is who to follow up on. But it's actually a goldmine in terms of understanding and even quantifying a firm's ability to access

• 27:07Speaker 01

  deals, right? Because if you join what deals are done with your CRM, you can start adding a score essentially to how much access does this firm have to a particular sector, particular GO, right? So let's say you covered 85% of all AI deals done in the Bay Area, but you only covered 15% in New York, right? That is a reason to add a partner in New York. If you ended up doing only 2% of the deals from the best papers at Europe's, you have

• 27:47Speaker 01

  like a research and academic conference hole in your sourcing, right? So it's a reason to add a partner there, right? So what's actually interesting is that if you look at this full circle from like an LP perspective, right, is just imagine if someone's of another fund is fundraising and you ask for the API key to their affinity or their CRM, you can actually quantify this, right? Like if a new firm says, you know, we have access to all the AI vertical deals in Europe, well, you can see from the meeting, right? You can see that, well, if this partner just left a large firm and say that they brought all these relationships, right? Is that true? And that's actually like a very interesting

• 28:31Speaker 01

  part that you can quantify as well.

• 28:34David Weisburd?

  It's fascinating because this concept of track record, people talk about it in venture capital as if it's the 10 commandments. It's these tablets that everybody goes around, everybody looks at the same 10 commandments and says, these are the commandments. In reality, even the most sophisticated LPs in the world, they've gone on the podcast, they all say the same thing. It's very difficult to understand attribution of track record. There's this old saying that success has of a thousand fathers and failure is an orphan. This is a real problem venture capital on every level. VCs trying to hire VCs from outside the firm. How do they know that they're actually did these deals?

• 29:14David Weisburd?

  LPs trying to invest in VCs and across and even startups trying to pick the GPs that actually did the deal versus somebody else in the firm that kind of actually did the hard work and did the value add for the startup.

• 29:29Speaker 01

  Yeah, definitely. And all that data is there, right? It's in the email inbox or the CRM. The big question is, you know, like, is anyone actually going to be willing to share this? Right? I think even the top LPs have problem getting into funds as well. Right? And getting them to actually share this data is probably not going to fly for a lot of a lot of large firms.

• 29:53David Weisburd?

  Yeah. Venture capital is this access class where it's kind of reverse selling. The LPs are trying to get into the top funds, the top 15 funds or so. And, and they're very cautious about they, they don't want to anger the GPs or then they get in the back of the line for, for allocation to the fund.

• 30:11David Weisburd?

  If an LP did want to do that diligence, and let's say the GP was willing to share that diligence, talk to me about what you would ask for from a GP and specifically what you would ask for.

• 30:21Speaker 01

  It's a lot easier to do now with Cloud.

• 30:25Speaker 01

  You need an API key to their CRM, right? And this will pull in all their emails, calendars, how many meetings, how many emails, et cetera. And you might want something about their network data as well. And so then you kind of know if they're actually connected or not, if there wasn't.

• 30:44David Weisburd?

  Something outside of LinkedIn?

• 30:46Speaker 01

  Yeah, so probably, I think generally probably around like getting their network dump or something, right? But then it's just joining that along with what deals were actually done, right? And then you could do different cuts of it by geo, stage, sector, you name it, right?

• 31:07David Weisburd?

  Yeah, it's, the calendar meetings are very obvious in retrospect and strong indicator because founders, again, it's reverse selling. The LPs are pitching to get into the top GPs and the GPs are pitching to get into top founders. What are the odds that a founder will meet with a partner five times, 10 times and not give them allocation to next round? I would argue it's probably close to zero, depending on how, how big their check size is. So you could use this calendar and these meetings as this upstream indicator of a GP's access.

• 31:39Speaker 01

  Yeah. And it's even, even a proxy for, you know, like decision making before it even gets to an IC or.

• 31:46David Weisburd?

  Support for today's episode comes from Square, the all in one way for business owners to take payments, book appointments, manage staff and keep everything running in one place. Whether you're selling lattes, cutting hair, running a boutique, or managing a service business, Square helps you run your business without running yourself into the ground. I was actually thinking about this the other day when I stopped by a local cafe here. They use Square and everything just works. Checkout is fast, receipts are instant, and sometimes they even get loyalty rewards automatically. There's something about businesses that use Square. They just feel more put together. The experience is smoother for them and it's smoother for me as a customer. Square makes it easy to sell wherever your customers are, in store, online, on your phone, or even at pop ups, and everything stays synced in real time.

• 32:29David Weisburd?

  You could track sales, manage inventory, book appointments, and see reports instantly whether you're in your shop or on the go. And when you make a sale, you don't have to wait days to get paid. Square gives you fast access to your earnings through Square Checking. They also have built in tools like loyalty and marketing so your best customers keep coming back. And right now you could get up to $200 off Square hardware when you sign up at square.com/go/HowIInvest. That's S Q U A R E dot com slash go slash How I Invest. With Square, you get all the tools to run your business with none of the contracts or complexity. Run your business smarter with Square. Get started today. Support for today's episode comes from Square, the all in one way for business owners to take payments, book appointments, manage staff, and keep everything running in one place.

• 33:12David Weisburd?

  Whether you're selling lattes, cutting hair, running a boutique, or managing a service business, Square helps you run your business without running yourself into the ground. I was actually thinking about this the other day when I stopped by a local cafe here. They use Square and everything just works. Checkout is fast, receipts are instant, and sometimes they even get loyalty rewards automatically. There's something about businesses that use Square. They just feel more put together. The experience is smoother for them and it's smoother for me as a customer. Square makes it easy to sell wherever your customers are, in store, online, on your phone, or even at pop ups, and everything stays synced in real time. You could track sales, manage inventory, book appointments, and see reports instantly whether you're in your shop or on the go. And when you make a sale, you don't have to wait days to get paid. Square gives you fast access to your earnings through Square Checking.

• 33:57David Weisburd?

  They also have built in tools like loyalty and marketing so your best customers keep coming back. And right now you could get up to $200 off Square hardware when you sign up at square.com/go/HowIInvest. That's S Q U A R E dot com slash go slash How I Invest. With Square, you get all the tools to run your business with none of the contracts or complexity. Run your business smarter with Square. Get started today. In investing, we talk a lot about execution. The same principle applies to running a home service business. Doing great work matters, but so does everything behind the scenes. Housecall Pro is the go to software solution built for HVAC, plumbing, electrical, and other home service professionals. Housecall Pro brings scheduling, customer communication, dispatching, invoicing, and

• 34:41David Weisburd?

  payments into one place so owners could spend less time on admin and more time growing their businesses. On average, businesses using Housecall Pro saves more than eight hours per week. Ready to save time and win more jobs? For a limited time only, our listeners get 50% off base plans for their first three months when you head to housecallpro.com/HH. That's housecallpro.com/HH for 50% off. After signing up, let them know HowIInvest sent you. I want to go back to what you said about evaluating a firm overall and its strengths and weaknesses, not from an LP perspective, but from a GP's perspective. They want to get ahead of the ball. They want to make sure they don't miss the next OpenAI, Anthropic, name your top disruptive company.

• 35:26David Weisburd?

  Talk to me about that. How should GPs use data in order to get a better picture for what's going on inside the firm?

• 35:32Speaker 01

  At least on the data side, right? There's actually like a parallel to this, right? So it's not just, did the GP talk to this company, the email? You can also do the same analogies in terms of, given the data sets that we have and the systems that we have, did we identify this deal beforehand as well? Right? So that's kind of the North Star KPI.

• 35:51David Weisburd?

  In other words, let's take instinct.

• 35:53Speaker 01

  Yeah.

• 35:54David Weisburd?

  There's two mistakes you could have made. One was not seeing it, which you could argue probably is a bigger mistake, and two is seeing it and passing.

• 36:01Speaker 01

  Exactly. Right? So is tying in the data side of, did you see it from like an inflection point or market mapping or, you know, like, was there a news article on it? Was there a GitHub repo on it or something? And then tying that in with, you know, like your investor meetings, their emails, right? Like, did they see that? So combined, those two give a pretty good holistic view of that. In terms of more broadly as a firm, I mean, there's probably efficiency things that you can quantify as well, right? It's like, what is the hit rate? Not just a term sheet or 10 years later, what the valuation is, but what's the hit rate of this new principal that joined the team in terms of meeting everyone that in the sector, right?

• 36:46Speaker 01

  Where the deals were done a couple of months later, right? So you can quantify this in terms of hit rate. And what's actually been interesting that other firms have done is just comparing that in terms of like a data hit rate versus like a junior associate hit rate versus like a senior partner hit rate.

• 37:03David Weisburd?

  Talk to me about that.

• 37:04Speaker 01

  You have data that might source for things outside of the network, right? And generally that might actually be what you want is to tune it. You don't want to see the same companies that you would have otherwise gone through your network anyway. So you do want to be able to quantify like, how many of these companies could we have met with earlier because this flagged it, right? But that's just a list of companies, right? That's your numerator, right? The denominator in this case is how many of these companies are actually done, right? By you or other top tier firms. You could do the same analysis for a principal or, you know, like a GP as well, is of all the meetings, of all the emails that they actually interacted with, that's the same denominator, right?

• 37:47Speaker 01

  But how many of these are actually done by, you know, the firm or other firms that you respect?

• 37:52David Weisburd?

  The one thing that I disagree with probably 90% of smart investors, a Peter Thiel question, what do you disagree with that, that smart people agree with in your industry? And that's this, to me, absurd idea that you don't really know whether you're a great venture capitalist for 10 years. So here's the logic. Well, companies take 14 years to exit from pre-seed to IPO. So when you're 10 years in, you haven't even gotten your first exit. And it's this very alluring narrative that people say, there's some usefulness to it in that you don't want to be arrogant, you want to be tracking, you want to make sure, you know, just because a company has gotten momentum and gotten marked up. Doesn't mean that's a good company. There's some lessons there, but it's absurd because

• 38:37David Weisburd?

  on the path of that pre-seed company becoming an IPO company, there's other milestones. Not just funding rounds, seed A, B, C, D, E, F, G, but also revenue. 1 million, 10 million, 100 million, 200 million, also secondaries, company tenders, all these other liquidity events. I think more so today, there's this underrated meme by Marc Andreessen, which is today, there's this underrated meme by Marc Andreessen, which is there are no more diamonds in the rough, only diamonds. Which explains these $100 million pre-seed evaluations, these billion dollar seed rounds is the consensus is coming earlier.

• 39:17David Weisburd?

  So what's second order effect of that? If you know that now who the best companies are, you could more quickly as a firm determine whether you're going to have a good portfolio and whether or not you're likely to be one of the top firms moving forward.

• 39:32Speaker 01

  I think at least in terms of being able to identify companies earlier, that still hasn't changed, right? A lot of this stuff has honestly just become table stakes as well, right? Like me working on talent data back in 2013 is now something that you can get off the shelf from a lot of vendors as well, right? So the, the alpha there in terms of identifying signals is diminished a bit. There's still a lot of interesting data sets that you can figure out what's happening sooner. But there's just a lot more that you can continuously measure as well, right? Like as you said, funding data, looking at tender offers, right, et cetera. But that also presents like other opportunities for exits as well, right?

• 40:17Speaker 01

  So it's selling secondary shares, it's selling into a continuation vehicle or whatnot. So it is kind of more in terms of portfolio management, right? Where the bet that you made for 10 years generally still applies, but there's just more options in terms of how you can exit that position as well if things turn south.

• 40:38David Weisburd?

  I want to go back to this concept of firms learning and firms actually growing. For probably up till about five years ago, venture capital was seen as this cottage industry where it's just the partners. And the venture capital firms didn't even have enterprise value. GP stakes famously, very few have happened in the venture space for that reason. There wasn't seen to have terminal value for venture firms. Do you think in an age of AI that might be changing?

• 41:10Speaker 01

  We've already seen that happen a lot, right? So yeah, there's a number of firms focused specifically for GP stakes in the venture industry. And especially now where just structurally with companies staying private much longer. There's just a lot more capitals allocated to the sector and the space. And it's just stuck there otherwise. And GP stakes that you kind of tap into that, right? So there are definitely more options like for growing a firm as well, right? And the ability to exit positions or share that upside with other investors too.

• 41:48David Weisburd?

  How do you look at the value in the venture capital firm between the, the house? The data and the venture capital firm and this learning organism like a company versus the individual general partners? How do you quantify those today?

• 42:02Speaker 01

  At the end of the day, right? Venture capital, unlike a hedge fund where you could just make a trade immediately. It's an interpersonal relationship business. It is about access. So the data side is helpful for identifying opportunities or whatnot. But getting onto the cap table is the most important part, right? And I think that's why you see a lot of firms focused on what I call like more of a token budget versus a data budget. Because you're essentially working on ways of making your existing investor base more efficient, right? Being able to check in on the founders that you're tracking, right? Saying congratulations on a new baby, et cetera.

• 42:43Speaker 01

  That helps in terms of building relationships with the founders. And that's what matters, right? At the end of the day is getting the founders to actually sign that term sheet.

• 42:53David Weisburd?

  Do you see that play out in junior staff in terms of it's becoming much more top heavy of venture capital firms with these star GPs? With, with AI and not as much room for associates and analysts?

• 43:04Speaker 01

  We're seeing that across all industries, right? So the partners at a legal firm, they're hiring significantly less associates.

• 43:14Speaker 01

  Your marketing, right? You're seeing the same thing. So yeah, I think there's going to be a lot more efficiency gains for the top GPs. There's going to be a lot more marketing budget that is going toward, you know, like campaigning for GPs and getting their names out.

• 43:33David Weisburd?

  Talk to me about value add as well. You've, you've quantified this going back to 2013. What does that even mean? How do you go about quantifying something like value add?

• 43:41Speaker 01

  There's MPS scores, obviously, right? Which is helpful. So I think a lot of venture funds started targeting this from a talent perspective, right? The value add being, the team being super important. So let me help you identify how to grow, like engineers to grow your team, like finding the right product guy when you need it. And one thing that we realized is sometimes this value add is important. And sometimes it's actually pretty significant in terms of hiring the right person. But the founders may or may not actually remember who to attribute that hire from, right? I've actually heard cases where they thanked another fund for this hire. And you're like, wait a minute, we're the ones that made that introduction.

• 44:24Speaker 01

  So it's, there's a question in terms of feeling like how much of value add that actually is. But there's other areas of this that are interesting, especially on the consumer side, where there's just a lot more data points, right? Like being able to figure out what market share looks like trending over time, right? So the, the early days, I think we actually did a lot of consulting for companies. And you could apply this to other things, you know, like food delivery companies, like in this geography. Is it actually trending or not? How many of your customers actually churned, right? Like this one that canceled on your service, are they showing up on another one, right?

• 45:07Speaker 01

  So there's a lot of value from that as well. I think the other part is, is also just like figuring out where from the other venture ecosystem, you could fill out the round as well, right? Like what other venture funds are not conflicted out that are investing in the sector and making the right introductions to that as well.

• 45:26David Weisburd?

  You're in the midst of seeing hundreds of these rounds come together today.

• 45:32David Weisburd?

  What would you say founders are prioritizing from their venture capital equity partners?

• 45:37Speaker 01

  I think it depends on the stage, right? So for, and there's also signaling risk and all these kinds of things with larger platforms as well. But there is, I believe someone did a study a while ago where there is actually like an impact on your ability to hire. If you, you know, get funding from a top tier firm, right? So that's definitely like top of mind, but as

• 46:03David Weisburd?

  That's really brand, this brand affiliation.

• 46:06Speaker 01

  Right. And as you grow as a company, then it probably becomes less important, right? Then it's just

• 46:13David Weisburd?

  The company starts to build its own brand.

• 46:14Speaker 01

  It's just the terms in that case, right? Then capital from one company or like a large bank probably doesn't differ that much from a venture fund.

• 46:24David Weisburd?

  Andreessen Horowitz just reportedly fired a lot of their recruiting team. What do you make of that, given how important talent is?

• 46:34Speaker 01

  I think that goes back to the example I had a little bit, right? Where sometimes it's just hard to even attribute where that came from. So I'm not sure if they saw that from their end in terms of like MPS or surveys that they sent out to their portfolio. But at a certain stage as well, it's just, it's just hard to beat a, a company or a firm that's actually just purely focused on talent. And for, for large companies, right? They have their own in-house teams. I think we touched a bit on terms of AI making teams much smaller as well. So maybe hiring itself is also just like not as needed.

• 47:14David Weisburd?

  The distinction between value add and attribution is a fascinating one. It tracks to some of the other conversations going on that I've had with others. Sam Altman's coach, Joe Hudson, talks about this, which the top founders that he's ever coached, and he's coached not just Sam Altman, but some of the top founders in the world. They all have pretty much goldfish memories. They have to make thousands of decisions a day, so they don't second guess themselves. They make a decision, they move on, they just compartmentalize on an S tier level. So they don't remember. If somebody made an introduction three months ago, they don't remember. And that's not a bug, that's actually a feature. That's what allows them to scale and manage these teams of tens of thousands of people.

• 47:55David Weisburd?

  So that tracks to why these firms might not be getting the, the credit for, for some of the value add that they're providing.

• 48:00Speaker 01

  Yeah, definitely.

• 48:01David Weisburd?

  One of the interesting factors that we see in the startup ecosystem today is these tranch rounds. Sequoia will come in at a billion dollar valuation and then startups will raise another round at $2 billion. Almost, you know, a couple of weeks later.

• 48:20David Weisburd?

  What do you think about these tranch rounds?

• 48:22Speaker 01

  It's pretty interesting to see those, especially on the AI side. A lot of these are a reflection of revenue growth. They're just scaling so quickly that a round that was priced a couple of months ago might be completely mispriced already, just five, six months down the road. So I think that makes sense from that perspective. And venture itself is still sitting on the most dry powder it's had ever, right? So there's a lot of money as well slashing around. It's just about getting into the right positions.

• 48:53David Weisburd?

  You've been quantifying venture capital, this art of venture capital, arguably longer than almost anybody, going back to 2013 and really built your career on this. If you go back and give yourself one timeless piece of advice when you first started doing this, what would that be?

• 49:08Speaker 01

  A lot of firms take the if it ain't broke, don't fix it mentality. And it is hard to impact using data from a structural standpoint, right? You really need top down support from the GP. Unless there's actual KPIs around targets and whether there's outreach, et cetera. It's hard to, to actually motivate the team to actually go and do the outreach or to, to follow up on flags from the data side. It's actually much easier to target more of the operational side in terms of making the firm more efficient. There's a lot of uncomfortable conversations that come with, you know,

• 49:52Speaker 01

  just quantifying access or understanding the efficiency of a venture firm. And especially as like an outsider, like which most data people in venture are. Coming from like a outsider perspective is hard to, especially when you deal with attribution as well. And it's an uncomfortable conversation and it kind of makes it harder than what I initially thought it would be in terms of, of, of getting adoption through a firm.

• 50:19David Weisburd?

  Let's say that you're advising a CEO, one of the top multi-stage venture capital firms, one of these top 15 firms. And the CEO was bought in of the firm and they wanted to institute data in the deepest level within their firm on every part of the stack. How would you advise them to execute on that strategy?

• 50:42Speaker 01

  I think you need KPIs, like what a true North Star is.

• 50:45David Weisburd?

  You need to make it objective.

• 50:47Speaker 01

  Exactly. And it has to be objective. That's in a way that's grounded by data. And by that, I mean, going back to the early example as well of what deals are actually done, right? And measuring that against, you know, what data flagged. What is the baseline for the firm in terms of, you know, what your associates, principals, GPs did in terms of deal sourcing, et cetera, decision making. But keeping it at that level in terms of what actually happened, what, what ended up being funding, right? A mistake I've seen a lot of times is trying to figure out, well, let's have the investment team rank each of these, right? And what I've seen a lot of times is those decisions actually change over time, right?

• 51:29Speaker 01

  So this decision, let's say airtable in 2016, right? It's genius move, but more recently after the acquisition, the outcome was not as good as, you know, a lot of people thought, right? So those kinds of decisions change over time and actually, and actually a lot of investors are churning now too. So you may or may not have the same ground truth label that you're trying to fit your model to. So I'm a hundred percent advocate of having a North Star or a KPI that is quantifiable in terms of data value add. And then you need that top down support to actually institute this throughout the firm.

• 52:07David Weisburd?

  There's an old saying from Jason Tref from 37 Signals, which is, don't build a half-assed product, build half a product. And I think this is really critical for AI, for anyone building and growing an AI-first organization. You need to have, I agree with your point, which needs to be quantitatively. But second to that, you need to have a successful implementation. So a lot of people will come in and say, well, we have sourcing, we have picking, we have winning, we have liquidity, all this stuff. Let's create this whole system and let's do a nine week offsite and then let's have a two-year implementation. That's not going to work. What's going to work is taking something that people care about, which is, are we getting a look at the very top companies?

• 52:52David Weisburd?

  These are top 150 companies. How do you know they're all worth over a hundred, over a billion dollars? These are the unicorns. This is what the market has determined at 150. How many of these did we look at? Okay, we looked at, we only got to look at 40 of these 150. In doing a project with AI on that specific thing,

• 53:13David Weisburd?

  a problem that no one could say is not an important problem, a problem that's very clear, a problem that's definable, and then executing on that, and then starting to expand the scope. I think this is just such an underrated aspect as people kind of go between doing these grand projects, which never works. It's like the high-speed train in California, which I think they, they did like one, one stop for like $200 billion versus these, almost like these startups, these lean startup approach to it, which is, get something that works, expand, make that work, expand further on. That's what I've seen as work day in and day out, whether it's a venture capital firm or just a startup implementing AI.

• 53:54Speaker 01

  Yeah, a hundred percent. And it's, it's critical to do that quickly as well. So being able to nail down an MVP and having a quick win is also important from, from what I've seen at other firms as well. But I think a lot of firms did do exactly what you said, right? Like I know a number of them, top firms that focus purely on data room diligence, right? And that's what they spent their time on for years. And then they wrote on, added on sourcing, right? Other firms took the other approach where they focused purely on sourcing and then they got to data room diligence. And I think a lot of that actually depends like where to focus, depends on your firm, your size, where you are at.

• 54:38Speaker 01

  And like a firm that gets all the inbound anyway, because you're a top brand, probably doesn't need as much on data sourcing, right? To begin with, and they're just trying to manage the avalanche of data rooms that they have to deal with. And whereas a smaller upstart venture fund, getting into the room and identifying those opportunities early may be your top priority and sourcing is the, the biggest thing, right?

• 55:05David Weisburd?

  So where's your bottleneck? There's this book that every MBA has to read. It's called The Goal. I think it's the only piece of management advice that Elon Musk has ever taken from, from MBAs, which is find the bottleneck in the business and focus all of your resources on that bottleneck. And then once you un-bottleneck the bottleneck, whatever the term is, then a new bottleneck forms. So you solve sourcing, now you have so many companies, now you actually have to deal with, oh, well, not dropping the ball, having an execution, oh, you solved that, okay, great. Now you have to focus on financing. How do you find the LPs to fund all this? So every business, no matter how much it scales, you know, on an infinite basis, there's always one bottleneck and just focusing the organization on those bottlenecks is a very underrated way, way to manage a company.

• 55:50Speaker 01

  It's definitely iterative, iterative process that can take a while to.

• 55:55David Weisburd?

  So tell me what you're working on today and how has that evolved in the last couple of years?

• 56:00Speaker 01

  The thing I, I found during my experience was that just venture itself is just inherently, you know, you're, you're making a decision at one point in time, but then it takes 10 years for that to play out. From a data scientist's perspective, that's a very long feedback loop, especially for someone. My background was in ads before, so you're literally picking the ad and seeing if someone clicked on it within.

• 56:24David Weisburd?

  Second by second feedback.

• 56:26Speaker 01

  It's milliseconds even sometimes, right? So, uh, so for that is actually, you know, way more intellectually interesting, right? And there's things along the line that, you know, along the lines where, you know, we looked at revenue data, which is a little bit faster, right? Looking at talent data and that kind of helps you track the digital breadcrumbs of companies as they grow. But what I wanted was something even faster, right? And for user acquisition financing, what we're working on right now, we are literally working with large scale data where companies are sending us their entire transaction ledger, their daily ad spend as part of the underwriting process. And we benchmark this against other companies.

• 57:08Speaker 01

  We see, we build predictive models around, is this actually going to pay off or not, right? Like if we fund a new cohort, is it going to pay back and within what timeframe? And after investment, we actually, and then like after investment, we're still tracking this on a daily basis, right? We're making API pulls into their Stripe accounts. We are looking at their meta, their Google ad APIs to figure out how much they're spending. And all this is to figure out, you know, like, is there ROI for every dollar that they spend, right? From a technical perspective, that's super interesting. And what we're providing on, to the founders, right, is non-dilutive financing.

• 57:51Speaker 01

  It's, we're not taking equity in any of the deals that we do. And for them, because they have post-product market fit, they know that every dollar they spend turns into two, three, four dollars down the road. And it's a short-term problem, but historically they've been financing on a 10-year capital structure and paying the costs of that. Whereas this is actually in line where, once they pay back the principal plus fees, they keep all the residual revenue, right? So imagine if you have Forex ROI for your ad spend, it's amazing for the company because quickly they become profitable if you just fund a couple of cohorts with those kinds of metrics, right?

• 58:33Speaker 01

  So from a business perspective, like I think this has been resonating as an industry for founders. They're seeing this and they are raising, you know, user acquisition facilities alongside equity. From data science perspective, it's very fast clip. So it's very intellectually interesting as well. And as an industry, it provides DPI in a environment that's pretty low DPI otherwise.

• 58:58David Weisburd?

  And we talked about this, General Callis has like this $5 billion fund going, going after a similar strategy.

• 59:03Speaker 01

  Exactly. So they definitely pioneered this and they've been targeting much larger, you know, companies, public ones even with billion dollar lines, 150, 250 lines. And for us, we think there's an opportunity in the mid-market. Big Craft itself is a gaming entertainment fund. So we've been seeing this a lot for our gaming founders as well and companies. And especially in that sector where it's been difficult to fundraise over the last couple of years, that something like this has been able to help companies scale three, four, five, 10x in terms of revenue, all in a way that's non-dilutive. And it actually creates opportunities for them to raise their next round as well.

• 59:48David Weisburd?

  I'm curious, you guys are processing quite a bit of data. How has your token usage evolved?

• 59:55Speaker 01

  Interestingly enough, uh, we are spending a lot more on, you know, traditional AWS infrastructure. So we're not just dumping everything that we have into a LAM and say, come up with a decision, right? We're using that to build infrastructure around it. So going back to that example of credit card data or whatnot, right? You're not going to just pass your entire credit card panel and say, identify the companies and, you know, attribute the revenue to it. You just blow up the context menu in your budget. The, the way to do it is, you know, build me entity resolution algorithms, right? For tagging transactions and then, so you have your traditional engineering workflows.

• 1:00:40Speaker 01

  It's just accelerated and you need a lot less people to build that.

• 1:00:44David Weisburd?

  Jerry, this has been absolute masterclass. Thanks so much for stopping by.

• 1:00:47Speaker 01

  Thanks so much for having me, David.

• 1:00:48David Weisburd?

  Congrats on getting to that in my interview with Jerry. If you enjoyed that episode, you're going to enjoy the episode right here.