👋 Hi, I’m Andre and welcome to my newsletter Data Driven VC which is all about becoming a better investor with data and AI.
👉Join our DDVC x Slush Breakfast in Helsinki Nov 18 with Goodwin & Harmonic
Brought to you by Affinity – The AI-first private capital CRM
Deal Stage Diagnostic: Find the Hidden Costs on Your Deal Team
Most deal teams assume time is lost at one particular stage. In reality, time leaks out of every stage: logging notes and updating CRM records, piecing together context the firm already has, and tracking down connections a teammate already made. It’s easy to overlook, but the hours add up.
Use Affinity’s ROI Calculator to answer six quick questions about how your team works and see which stage costs you the most and what that time is worth.
Welcome to another Data Driven VC “Insights” episode where we cover the most interesting research and reports about startups, GPs, LPs, AI & automation.
The Death of Venture Capital's Short Kings
Kyle Harrison at Contrary argues that USV and Benchmark, the last firms that stayed small by choice, have now scaled up, and asks what that shift means for the companies venture will fund.
The Last Holdouts Scale Up: USV announced $900M across its newest funds, the largest cycle in its history, about four months after Benchmark raised $2B, its first meaningful step up since 1999, including a first-ever $1.25B growth fund.
Capital Concentrates at the Top: In Q1 2026, roughly 73% of LP capital raised went to just five firms. a16z alone held over $40B in AUM, against roughly $4.8B at Benchmark before its new raise.
Small Funds Lost Their Seat: Harrison argues the rounds that decide a company's fate now come later and are led by megafunds, so the only way to protect founders was to grow big enough to join them.
✈️ KEY TAKEAWAYS
For founders, a lead investor's fund size decides which exits move its needle, so ask early what outcome it needs to return the fund. For LPs, Fred Wilson still expects USV to produce strong returns if its biggest winners land in the single-digit billions, a segment fewer and fewer firms are built to pursue.
The Optimal Portfolio Size
Mackey Leventis models fund returns by number of portfolio companies, using Correlation Ventures data on 21,640 US financings.
More Companies Lift the Median: The modeled average fund returns 2.06x gross at every portfolio size, but the median rises from 1.57x with 10 companies to 1.92x at 40 and 2.01x at 100, because each added company is another shot at the roughly 4% of deals that return 10x or more.
One in Four 10-Company Funds Loses Money: A 10-company fund has a 34% chance of owning a 10x and a 26% chance of losing money. At 40 companies, those odds move to 80% and 4%.
Diversification Caps the Upside: The top 5% of 10-company funds return 5.5x or more gross, against 3.5x at 40 companies, assuming the 40th deal is as good as the 10th. Setting fund size aside, Leventis puts his sweet spot at 40-60 companies.
✈️ KEY TAKEAWAYS
Emerging managers should pick their portfolio size based on the results they want: about 10 companies means a shot at a standout fund but a 1 in 4 chance of losing money, while 40-60 companies almost always return the money invested. LPs should note these are pre-fee figures assuming equal checks and no follow-on funding, so actual returns will come in lower.
Upgrade your subscription to access our premium content & join the Data Driven VC community
The Future of AI Labs
Tomasz Tunguz at Theory Ventures argues that AI models have reached commodity status, which leaves labs competing for share through partnerships, interfaces, and routing data.
Prices Fall as Usage Climbs: Token usage has risen 50% since July while prices fell 41%, according to Ramp's AI Index. OpenAI cut prices on Luna by more than 80% to win share.
Buyers Move to Cheaper Tiers: The share of tokens going to frontier models, the most capable and priciest tier, slipped from 53% in August into the mid-40s, according to Ramp data, as companies defaulted to smaller, cheaper models.
Distribution Becomes the Moat: Tunguz’s playbook for labs has three parts: (1) resell other providers' models, as OpenAI has announced with Baseten, (2) own the interface customers use, and (3) use data on which model handles each request to lower future training costs.
✈️ KEY TAKEAWAYS
If models become interchangeable, value moves to whoever owns the customer, so investors should weight retention and switching costs heavily when assessing AI application startups. Founders can turn falling token prices into margin by sending each task to the cheapest model that handles it well.
Solo Founders Lead 38% of New Startups
Hamza Shad at Carta tracks the share of newly founded US startups on Carta led by a single founder, and links its recent acceleration to AI coding tools.
Up From 23% in 2019: Solo founders led 37.9% of US companies founded on Carta in H1 2026, up from 23.4% in H1 2019, based on half-year founding cohorts.
Most of the Rise Came Recently: The share climbed 5 points over five and a half years, from 23.4% in H1 2019 to 28.4% in H2 2024, then another 9.5 points in the 18 months to H1 2026.
Vibe Coding Lowers the Bar: Shad credits vibe coding tools with making a minimum viable product easier to build, though he doubts they can fully replace technical expertise for securing that software and keeping it robust long term.
✈️ KEY TAKEAWAYS
Investors meeting solo founders can probe how they plan to cover security and long-term engineering, the gaps Shad himself flags in vibe-coded products. The data counts newly founded companies only and does not show whether solo-led startups raise or exit at similar rates, so treat it as a sourcing signal for now.
Join 2,002+ investors in our free Slack group as we automate our VC job end-to-end with AI. Live experiment. Full transparency.
You Need a New CFO
Seema Amble, Ivan Makarov, and David Borecky at a16z argue that AI-native tools are making the CFO an architect of the company's operating system, run by a smaller, higher-leverage finance team.
Each Era Hired a Different CFO: Companies first hired CFOs from Big 4 accounting firms, then from investment banking, then for financial planning and analysis (FP&A). The authors see today's scarce skill as designing the data, workflows, and agents a company runs on.
CFOs Become Builders: The authors see CFOs increasingly acting as "finance engineers" who build their own tools with AI coding assistants. At Anthropic, one such tool produces a 90-95% complete monthly financial review in about thirty minutes.
Lean Teams Target 2% of Headcount: The old rule of thumb put finance at about 5% of headcount. The leanest teams a16z sees now target 2%, hiring people who write SQL and prototype automations.
✈️ KEY TAKEAWAYS
Founders hiring a first CFO should look for someone who decides which finance work software should handle and builds those tools, since the a16z team expects this to free up the CFO’s time for bigger decisions like pricing and compute spend. Investment firms can apply the same playbook to fund accounting and LP reporting, as long as every AI-generated number can be traced back to its source.
References Are the Most Underrated Tool
Brett Berson at First Round explains how he runs reference calls as a core hiring tool, a method he calls composite referencing.
Most Reference Calls Waste the Signal: Berson says the typical reference check in tech and venture is a short call with generic questions and glowing answers from people the candidate picked, which adds little and can simply confirm what the team already believes.
Scale References to the Stakes: He suggests at least 3 references for junior hires, at least 5 for mid-level hires, and 12 or more for executives, since a wrong executive hire costs far more than a wrong intern hire.
Start Early and Widen the Circle: He runs about a third of references early, before the team is committed, mixes candidate-named references with backchannel calls, and ends every call by asking who else knows the candidate.
✈️ KEY TAKEAWAYS
Berson's critique covers venture firms hiring their own team, and the same playbook carries over to founder diligence: start calls early, prepare about 10 questions built around the role and your open doubts, and ask each reference who else to call. When praise sounds hesitant, dig deeper with follow-ups.
That’s it for today!
Stay driven,
Andre
