👋 Hi, I’m Andre and welcome to my newsletter Data Driven VC which is all about becoming a better investor with data and AI.
ICYMI, check out some of our most read episodes:
Brought to you by Vessel - Agentic fund operations for VC and PE firms
How are GPs actually implementing AI in their own operations? Most firms are sitting on fragmented data, inconsistent CRMs, and big ambitions, with no clear path forward.
On June 23rd, Thomas sits down with Anthony, Managing Partner at Antler, to discuss:
How the best fundraisers think about LP relationships differently
Why clean data is the only real prerequisite to AI leverage
What it actually looks like to build agentic workflows across an 18-city team
Welcome to another Data Driven VC “Insights” episode where we cover the most interesting research and reports about startups, VCs, LPs, AI & automation.
Your First Customer Defines Your Ceiling
Kyle Poyar's data analysis across 1,043 SaaS and AI companies shows that the customer segment a startup targets on the way to its first $10k MRR shapes its retention and growth trajectory for years.
Deer Hunters Grow 3-5x Faster: Companies targeting $300-$2,999/month sustained 22% average YoY growth, versus 4% for rabbit hunters and 2% for mouse hunters, with the gap widening continuously after $10k MRR.
Early Churn Is Catastrophic at Low Price Points: Median annualized GRR at $10k MRR is 24.6% for mouse hunters and 31.6% for rabbit hunters, versus 70.5% for deer hunters, meaning low-ACV founders are refilling a leaky bucket from day one.
70% of Startups Never Change Their Target Animal: Upmarket migration is rare and slow, rabbit to deer is the most common shift at just 13% of the dataset, and moving downmarket averaged -11% growth.
✈️ KEY TAKEAWAYS
The deal size decision at $10k MRR is a structural commitment, not a provisional one. Deer hunting ($300-$3k/month) produces the best risk-adjusted growth, with enough budget to fund real GTM and retention that compounds. Investors should treat a founder's initial ICP as a leading indicator of long-term unit economics.
The True State of DPI
Peter Walker, Head of Insights at Carta, analysed DPI distributions across 2,689 US venture funds ranging from $10M to $1B+ to assess how capital is actually being returned to limited partners.
2017 Vintage Leads, but Capital Is Still Locked Up: 84% of 2017 vintage funds have returned at least $1 of DPI after 33 quarters, yet only 57% exceed 0.25x and just 25% exceed 0.5x, indicating distributions remain thin even in the most mature cohort.
2019 and 2020 Vintages Are Structurally Behind: Only 49% of funds from those vintages have returned a single dollar, with 16% of 2019 funds and 11% of 2020 funds clearing 0.25x DPI, reflecting both the compressed exit environment and the shift to a 15-year fund lifecycle.
Early DPI Is Essentially Uncorrelated to Final DPI: Where a fund sits in its early quarters has no meaningful relationship to where it ends up, which undermines using current DPI as a performance signal for re-ups in younger vintages.
✈️ KEY TAKEAWAYS
LPs who entered venture in 2019 or 2020 are operating inside a 15-year asset class, and the data validates that framing. The near-term pressure falls on GPs: low DPI makes re-ups harder regardless of TVPI, pushing managers toward secondaries and structured exits rather than waiting for the IPO window. DPI is becoming the primary trust signal in LP relationships.
Why Meeting Notes Are Your True Moat
David Haber at a16z argues in Everything is Recorded Now that workplace recording has become the default for AI-native companies, creating a structural gap between firms that embrace it and those that resist.
Meeting Context Is How AI Actually Learns a Company: AI onboarded through recorded meetings picks up a company's actual operating logic, its edge cases, cultural norms, and real decision-making patterns, in ways that static wikis and CRM data cannot capture. Haber's framing: you don't hand a new employee the documentation and expect them to get up to speed, you invite them to meetings.
Verbal Cultures Gain a Compounding Advantage: Companies like Shopify and OpenAI that default to verbal communication benefit disproportionately, since AI can now persist and structure what was previously ephemeral, while written-culture companies like Stripe already capture most context by construction.
The Default Is Flipping from Opt-In to Opt-Out: Haber predicts the recording assumption shifts within six months to "assume you're being recorded unless explicitly designated otherwise," with sensitive meetings carved out under something like an "AC Priv" designation.
✈️ KEY TAKEAWAYS
The enterprise software category being built around voice and conversational context is still early and largely undefined. The real competitive moat belongs to whoever owns the context layer inside a company, as the AI trained on two years of internal meetings is a fundamentally different product than one that read the documentation. For investors, this is the wedge for a new system of record that could displace CRMs and wikis over a 5-10 year horizon.
Join 1757+ investors in our free Slack group as we automate our VC job end-to-end with AI. Live experiment. Full transparency.
AI Already Replaced Your Deal Screener
Dan Gray at Odin argues in Encyclopedia Galactica that AI and venture capital are complementary at the margins of the market but substitutive at its center, and that the "2 and 20" fee structure has not kept pace with either dynamic.
LLM-Assisted Deal Screening Is 537x Faster with No Quality Loss: AI-based opportunity screening matched human analyst quality at 537 times the speed, making automated deal flow viable for scaled platform funds pursuing a smart beta index strategy rather than outlier hunting.
The Solo GP Is AI's Primary Beneficiary: For small funds whose edge comes from proprietary sourcing and conviction, AI functions as a permanent associate handling LP reporting, memo synthesis, and inbound triage, without competing with the human judgment that generates alpha.
Management Fees Are Structurally Miscalibrated: Fee drag on gross-to-net returns runs 5%-8% annualised, with VC funds at the higher end at 8.5%, suggesting scaled firms running AI-augmented operations could deliver materially better LP outcomes by moving to 1%-1.5% management fees.
✈️ KEY TAKEAWAYS
AI is accelerating the bifurcation of the VC industry that was already underway. Scaled platforms should automate and reduce fees to compete on access and efficiency, while boutique funds should raise their floor by proving idiosyncratic alpha that AI cannot replicate. The funds most at risk are mid-sized generalists with neither the distribution to index nor the differentiation to justify high fees.
How Waterfalls Actually Work
Ilya Strebulaev, Stanford GSB Professor, continues his VC 101 series with a detailed breakdown of liquidation preference seniority and pari passu structures, using real cap tables from Dropbox and Kabbage to show how exit proceeds are actually distributed.
A 25% Stake Does Not Mean 25% of Exit Proceeds: The Kabbage case study shows common shareholders holding the largest fully diluted stake (~28%) received zero at a $430M exit, because six preferred series with combined preferences exceeding $435M cleared first.
Early-Round Terms Compound Forward: New investors almost never accept worse terms than existing ones, so a 2x liquidation multiple in Series A becomes the Series B floor, and in Strebulaev's example this single decision shifts total liquidation preference from $25M to $40M, eliminating founder payout at most moderate exit scenarios.
Hybrid Structures Are the Norm: Real contracts mix seniority and pari passu, with later series holding priority over earlier ones while the earliest series share proceeds proportionally among themselves, making it essential to model the full waterfall rather than using ownership as an exit value proxy.
✈️ KEY TAKEAWAYS
Ownership percentage is a misleading headline in multi-round companies. Investors and founders both need to model the full waterfall at each exit scenario, and founders should pressure-test every early-round term against its effect on future negotiations. The Kabbage example is the clearest illustration: $430M exit, common shareholders get nothing.
Stop Collecting Information. Start Acting on What You Already Know.
Tim Denning argues in a widely circulated X essay that the real barrier to progress is not a lack of information but an addiction to gathering it, rooted in the dopamine reward of learning without doing.
Information Gathering Mimics Progress Without Producing It: Dopamine from reading and planning activates the brain's reward system the same way action does, allowing people to feel like they are moving forward while remaining stuck in a planning loop that life's unpredictability will eventually invalidate.
Real Learning Requires Negative Emotion: Denning frames genuine skill acquisition as a function of high-stakes failure and consequence, citing a $1.2M crypto hack, business collapse, and repeated personal setbacks as the experiences that produced lasting change rather than any amount of information gathering.
Minimum Viable Information Is the Correct Starting Point: Deploy what you already know, learn from the output, and add personalised help iteratively. Denning credits this approach, not pre-action research, with reaching the top 1% of Medium writers and generating $70k/month.
✈️ KEY TAKEAWAYS
Venture runs on information advantage, yet the founders generating the best returns often acted before the information was complete. Denning's minimum viable information framing maps directly onto lean startup principles, and the practical signal for evaluating founders is whether they are learning from action or endlessly refining a plan. Bias toward action is a competitive differentiator when everyone has access to the same data.
That’s it for today!
Stay driven,
Andre
PS: Don’t forget to join Vessel’s session on “building the agentic VC firm” with Antler on 23rd June here
