Jun 17, 2019 · 30m · 20vc
20VC: Why Historical Loss Ratios Are Simply Too High, Why Data Is The #1 Most Important Piece When Evaluating Effective Reserve Allocation & Why Nothing Is Truly Defensible Today with Jonathan Hsu, Co-Founder and General Partner @ Tribe Capital
0:00 / 0:00
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of The 20 Minute VC, host Harry Stebbings interviews Jonathan Hsu, Co-Founder and General Partner at Tribe Capital, exploring how data science transforms venture capital sourcing, investment evaluation, and portfolio construction. Hsu shares frameworks for quantifying product-market fit, lowering portfolio loss ratios, and navigating competitive early-stage markets.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Harry holds 37.9% of the talking time here. How this is scored →
Harry as informed peer 3.9 Guest teaching 4.6 Guest disagreement 1.9 Harry pushing back 2.1
speaking balance: gold is Harry, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 17:17 Dismissing Pricing Precision in Early VC
Jonathan directly rejects Harry's premise that data enables precise asset pricing at Series A/B, stating 'Not really. Not at all' and explaining how VC power-law dynamics render pricing precision irrelevant.
Hardest push from Harry ▶ 22:50 Challenging Defensibility ClaimsHarry directly challenges the standard VC thesis of defensibility, arguing forcefully that no startup is truly defensible today given the overwhelming talent and resources of tech giants.
Biggest teaching moment ▶ 7:41 Network Supremacy Over Pure Machine SourcingJonathan educates Harry on why machine-only sourcing fails in early-stage VC, explaining that traditional top-tier firms succeed through brand and network effects that algorithms cannot replicate.
Harry holds his own ▶ 7:24 Calling Out Pseudo Data SourcingHarry demonstrates industry knowledge by cutting through marketing buzzwords, asking whether most 'data-first' VC firms are merely using LinkedIn Sales Navigator smartly rather than deploying true machine intelligence.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Harry as informed peer | Guest teaching | Guest disagreement | Harry pushing back | Why |
|---|---|---|---|---|---|---|
| Interview Opening and Guest Welcome | 2 | 3 | 0 | 0 | Harry introduces Jonathan warmly and asks standard background questions regarding his string theory PhD and building Facebook's data science team. Jonathan details his career path smoothly in an agreeable tone. | |
| Applying Data to Venture Capital Sourcing | 4 | 4 | 2 | 3 | Harry asks an insightful question probing whether 'data-first' VC firms are actually doing transformative sourcing or just using LinkedIn Sales Navigator intelligently. Jonathan reframes the dynamic, noting that machine sourcing alone puts firms at a massive disadvantage compared to traditional networking. | |
| Using Quantitative Data for Investment Evaluation | 3 | 5 | 2 | 2 | Harry brings up potential downfalls of quantitative data during investment picking. Jonathan educates on the spectrum across startup stages, pointing out that seed relies on founders, growth equity relies on financial data, and Series A/B sits in the middle. | |
| Winning Competitive Deals Through Data Transparency | 3 | 4 | 1 | 1 | Harry asks how data helps win competitive deal battles against other term sheets. Jonathan explains how presenting objective benchmarking data (such as quintile rankings) creates high-context alignment with founders. | |
| Post-Investment Value Add and Growth Philosophy | 4 | 4 | 1 | 1 | Harry frames Tribe's approach as a 'freemium VC model' where upfront data insights transition into post-investment help. Jonathan outlines how Facebook's growth discipline is applied to portfolio company product flywheels. | |
| Startup Execution Gaps and Timing Data Discipline | 3 | 5 | 2 | 1 | Harry asks about execution gaps and when data discipline should be inserted. Jonathan notes early-stage startups are never great at execution and cites Slack and Carta as companies that didn't build formal data science practices until Series C. | |
| Asset Pricing Dynamics in Early-Stage VC | 5 | 6 | 4 | 3 | Harry asks if data helps price early-stage Series A/B assets intelligently. Jonathan bluntly rejects the premise with 'Not really. Not at all,' explaining that binary power-law outcomes render pricing precision irrelevant compared to quant hedge funds. | |
| Reserve Allocation and Co-Investor Alignment | 5 | 5 | 2 | 2 | Harry asks an insider question on reserve allocation and Henry Ward's concept of 'N of 1' markets. Jonathan clarifies 'N of 1' versus '1 of N' using airport car rental companies as a counterexample. | |
| Evaluating Defensibility and Network Effects | 6 | 5 | 3 | 6 | Harry forcefully pushes back on traditional VC defensibility logic, asserting that nothing is truly defensible against FAANG resources. Jonathan validates Harry's long-term view while reframing defensibility to medium-term network effects. |
Statements from this episode (16)
Hsu: A PhD in string theory is the most useless degree
“It's really sort of the most useless topic you can probably get your PhD in.”
Hsu: Data science is no silver bullet for venture sourcing
“By and large, that there's no one silver bullet in sourcing, right? Sourcing is one of these things where good old ground game, just raw networking and being out there is a really important piece of it. Building brand is an important piece of it and having an …”
Hsu: Seed stage focuses on founders, while late stage focuses on business metrics
“If you think about the seed stage, right, they're really buying the founder when the investor invests. And when you think about the late stage, they're buying mostly a business, right? And then it's definitely some founder, but mostly a business.”
Jonathan Hsu: Data is unbiased; humans introduce bias through inference
“Data is actually totally unbiased. When you infer something from data, then you do something biased, right? That's where you introduce bias.”
Hsu: Growth is fast, data-informed iteration, not growth hacking
“And when you use the word growth, we don't really mean growth hacking, but it's more like product development through the philosophy of growth, which is really about being fast, being very data, being highly data informed. You know, it's not necessarily data d…”
Hsu: Startups need quantitative data once reaching hundreds of users
“I think anecdote works really well, you know, when you're dealing with the first couple of customers, when you're dealing with only maybe a dozen users, maybe two dozen users. But once things start pushing into dozens of customers, possibly hundreds of users, …”
Hsu: Slack and Carta didn't build dedicated data teams until Series C
“I just want to think back to Slack and Carter, two companies that we've worked with, you know, in the last couple of years, you know, both of them didn't really start investing in this Full bore, you know, well into about the series C, if I remember correctly.…”
Hsu: Valuation variances up to 50% do not alter early-stage venture returns
“The probability distributions that we're dealing with at the very earliest stage are such that if you're plus or -20, 30, even 40, 50%, it's not going to make or break the result. Either you're going to lose all your money, or you're going to make multiple tim…”
Hsu: Historical Loss Ratios in Venture Capital Are Too High
“We believe that historical loss ratios are too high in venture to some extent.”
Hsu: Industry Venture Capital Loss Ratios Sit at 30% to 40%
“So historical loss ratios for venture at large tends to be in this 30 to 40% range”
Hsu: Data becomes the primary evaluation factor as startups mature beyond Series A
“When you're in the earliest stage, Data is one of many things. You know, the interesting thing about a Series A is that there is no one most important thing. There are many, many important things. And then the further and further you get along, the more and mo…”
Hsu: Tribe Capital will sacrifice ownership percentage to recruit top growth co-investors
“We don't necessarily think that we're the best growth investors out there. So part of our job is to help our portfolios get the best possible growth investors in their later rounds. So even if that means that we have to give up some ownership for it, because, …”
Jonathan Hsu: Tech giants dominate by creating superior products, not choking supply
“These companies are dominant, not because they choke off supply, but because they actually provide the best product for customers. That's true of Google. It's true for Facebook.”
Hsu: Nothing in tech is defensible in the long term
“I think in the long term, you're absolutely right. Nothing is defensible in the long term, right? The question is like, are you in an era where you can defend it for a while and give yourself enough breathing room to possibly innovate something else, build som…”
Hsu: Network effects are the primary source of medium-term defensibility
“You know, when we think about defensibility right now, the most obvious pattern that we've seen in the last 15 years now is really the concept of a network effect, right? The concept of a network effect seems to really be able to put in a chunk of defensibilit…”
Hsu: Accounting was the original form of data science
“I think the biggest misconception here is that the only way to use data is like the way that machine learning or AI works. You know, people overlook things like good old-fashioned accounting. You know, accounting was sort of the very first data science.”