How to Actually Measure a Venture Studio

A venture studio quoting one headline number is telling you almost nothing until you know which number it is. Three metrics carry the freight in private markets, and they answer three different questions. IRR answers how fast. TVPI answers how much on paper. DPI answers how much in cash. Confuse them and you will mistake a marketing slide for a track record.

This is a guide to venture studio performance metrics for readers who have to allocate real capital. We will define IRR vs TVPI vs DPI, show what each one hides, treat survivorship bias head on, and explain why the GSSN studio benchmark of ~50% IRR versus ~19% for traditional VC still holds up as a directional fact. Avante Ventures is a venture studio building AI-native companies in Brazil and Latin America, and a young one, so this is also how we think about reporting on ourselves.

IRR, TVPI, and DPI are not the same claim

Each metric answers a different question, and a studio that blurs them is usually hiding behind the friendliest one. IRR is the annualized, time-weighted return on cash flows. It rewards early distributions disproportionately. TVPI, total value to paid-in, is realized plus unrealized value over invested capital, the total created on paper. DPI, distributions to paid-in, is the cash actually returned to investors over what they put in.

The cleanest framing comes from a [Value Add VC breakdown of fund metrics](https://valueaddvc.com/blog/how-venture-capital-fund-performance-is-measured-irr-tvpi-dpi-and-rvpi-explained). IRR is a marketing metric. TVPI is a progress metric. DPI is the only one that settles the debate. A 1.0x DPI means investors got their money back in cash. A 0.3x DPI means 70% of the fund is still paper and a story.

• IRR. How fast capital compounded, on paper, with timing assumptions baked in.
• TVPI. How much total value exists, realized plus unrealized, before anyone is paid.
• DPI. How much cash actually landed in an investor account. The only one that cannot be marked up.

Why IRR alone is a yellow flag

IRR is the most flattered number in private markets, and the academic record is blunt about it. Ludovic Phalippou, the Oxford finance professor, has spent years documenting how since-inception IRR overstates returns. In his November 2024 piece for the [CFA Institute on the tyranny of IRR](https://rpc.cfainstitute.org/blogs/enterprising-investor/2024/the-tyranny-of-irr-a-reality-check-on-private-market-returns), he shows that across 12,306 private capital funds holding $10.5 trillion, the median IRR was 9.1% and the public-market-equivalent implied just 1.4% annual outperformance over the S&P 500.

The flattery is mechanical, not mysterious. IRR assumes interim cash can be reinvested at the same rate and it overweights early cash flows. A manager can lock in one early exit, hold the strugglers at cost, and post a number that has drifted away from anything an investor will ever bank. Phalippou put it sharply. IRR has become the theatre of private equity performance. It delivers a beautiful illusion until someone looks at the maths.

So when a studio leads with IRR alone, especially in its early years, read it as a yellow flag. It is showing you the metric that is easiest to flatter and furthest from cash.

TVPI is paper, DPI is cash

TVPI tells you the trajectory. DPI tells you the truth. The gap between them is where most self-deception lives. The J-curve means early TVPI sits below 1.0x for years on fees and timing alone, and the metric mixes real cash with speculative paper valuations that depend entirely on the last round's price.

Markups prove the point. In 2021 paper values ran hot. In 2023 and 2024 many funds quietly cut residual values by 30 to 50 percent, per the same Value Add VC analysis. A TVPI built on stale 2021 marks is a number waiting to be revised down.

This is also why DPI lags for years and why an honest young studio cannot show a meaningful one yet. Distributions need exits, and exits need a liquid market. The metric that cannot be gamed is also the one that takes the longest to arrive. Patience is not a weakness in the number. It is the number working correctly.

Survivorship bias and how benchmarks correct for it

Every studio benchmark you read is built on a survivor's sample, and that quietly lifts the headline. Survivorship bias, as [VC Beast defines it](https://vcbeast.com/venture-capital-glossary/survivorship-bias), is the tendency to study only the funds and companies that lived while ignoring the far more numerous failures. The structural problem for any benchmark is plain. Failed funds stop reporting, so the database keeps the winners and loses the losers.

The distortion is measurable in neighboring asset classes. A 2025 Wedge Capital Management study found survivors beat drop-outs by about 0.86% a year, and hedge fund indices have historically been inflated by roughly 3 to 4.5 percent annually from these biases. Numbers built only on who is left will always read high.

Credible benchmarks fight back in three ways, and a reader should check for all three. Fix the time window so one boom vintage cannot dominate. Widen the sample so no single survivor swings the average. Disclose the sample size so the reader can judge it. The venture studio dataset compiled by 9point8 Collective does the last one openly, stating its performance figures rest on 20 studios with 2015 to 2022 vintages and that only 13% of studios keep firm-level track records.

Why the GSSN gap still holds up

The studio outperformance figure survives a skeptical read, as long as you frame it honestly. The Global Startup Studio Network benchmark puts studio IRR at roughly 50% against an industry-standard roughly 19% for traditional VC, about 2.5x the IRR of traditional VC over realistic time horizons. State it precisely. That is the GSSN studio-model benchmark, never any one studio's realized return, and a young studio has no meaningful DPI behind it.

What makes the gap credible is that independent samples point the same way even when the decimals differ. The [9point8 Collective venture studio data](https://9point8collective.com/research) reports an average net studio IRR of 60% against 33% for top-quartile traditional VC, studios putting 2.3 to 2.6x more value into each dollar deployed, and studio-built companies reaching Series A in 25 months versus 56 for traditionally founded startups. Different number, same direction.

The mechanism is why it is structural rather than luck. The studio compresses time-to-traction and routes capital into product instead of company-building overhead, repeated across every venture. Datasets argue about the exact figure. None of them argue about the sign. A directional benchmark that holds across independent samples is exactly the kind worth trusting, and exactly the kind worth stating with its caveats attached.

Metrics that mislead

Some numbers a studio shows you are not performance at all. They are activity wearing performance's clothes. Learn to name them on sight.

• Logo counts. The number of companies launched says nothing about whether any returned cash. Vanity at portfolio scale.
• Total capital raised by the portfolio. This measures other investors' enthusiasm, not the studio's return. A portfolio can raise a fortune in follow-on and distribute nothing to the studio's own backers.
• Unrealized markups in a frozen market. With most 2019-2022 vintages below 0.5x DPI, a TVPI resting on 2021 marks is a haircut waiting to happen.
• Since-inception IRR with no horizon context, the exact metric Phalippou argues should be banned in favor of horizon IRRs.

How Avante reports on itself

The honest position for a young studio is to name what it cannot yet prove. Avante Ventures is a venture studio building AI-native companies in Brazil and Latin America. We launch 3-4 ventures per year through a six-stage system, Research, Partner, Build, Traction, Revenue, Compound, deploying $500K-1.5M per venture and retaining co-founder economics. Avante is young, so our own DPI is not yet meaningful. Saying that plainly is the point. The ~50% IRR versus ~19% IRR gap is the GSSN benchmark for the model, not a claim about Avante's realized return.

What we can be measured on today are leading indicators with teeth. Time-to-traction, where a studio venture launches 6-9 months ahead of a comparably funded standalone team. Capital efficiency, where solving company plumbing once routes roughly $300K-500K of effective capital per venture into product rather than overhead. And the copilot to data to fund flywheel, where an AI copilot generates proprietary data that earns the right to raise.

The regional reading matters here. LATAM venture capital ran near $4.5 billion across 751 deals in 2024, a thinner and more cyclical exit market than the US, which makes DPI honesty more important in Brazil, not less, where services are roughly 70% of GDP with low software penetration. The edge is operators with 10+ years of Brazilian-market scar tissue paired with first-ticket capital on day one. A studio that earns the right to quote ~50% IRR is the one willing to tell you its DPI is still zero. Read the thesis at [/why-avante](/why-avante) and the operating model at [/principles](/principles).