Venture Capital Fund Returns: Vintage, Quartiles and Realized Outcomes
A fund can rank in the top quartile and have distributed little capital. An index can rise while LPs continue contributing more than they receive. A portfolio can include an outstanding manager yet capture little of that manager’s performance because the allocation available was small.
Interpreting venture capital fund returns requires connecting four things: the peer group, the calculation method, the opportunities actually available and the investor’s cash flows. “What does venture capital return?” is an incomplete question until we specify whose investment experience we want to measure.
Our TVPI vs DPI guide explains how to read an individual fund’s metrics. This article takes the next step: turning those numbers into a useful comparison for manager selection and allocation.
1. The index return is not the return of the fund you can subscribe to
In its commentary published on July 31, 2026, Cambridge Associates reports a 21.1% return for its US Venture Capital Index in 2025. This is a pooled IRR over the annual horizon, net of fees, expenses and carried interest. During the same year, the sample called $61 billion and distributed $42 billion. Improving performance and LPs’ continuing net liquidity needs can coexist. Source: Cambridge Associates, data as of December 31, 2025.
That 21.1% describes one year for a collection of funds. It is neither the since-inception return of a 2025 vintage fund nor a forecast for a new commitment. Investing today buys exposure to future decisions, prices and timelines; it does not purchase the index’s historical results.
Benchmark misuse begins when a valid measure is asked to answer a different question. An annual return can describe the performance of an existing portfolio. Assessing a new manager requires comparisons between suitable peers; deciding how much to allocate also requires understanding liquidity across the LP’s entire investment program.
2. Vintage: the same cohort does not mean the same entry prices
Vintage is the year used to assign a fund to a cohort. Check the convention: it may refer to legal inception or the first capital call. A fund’s inception year and the years in which it invests in individual companies are also different pieces of information. Cambridge Associates makes this distinction explicit in its work on investment-level benchmarks. Methodology source.
Imagine two funds from the same vintage. One invests much of its capital quickly in a fiercely competitive market; the other spreads its initial investments across several years. Vintage provides an initial basis for comparison without making their entry-price exposure identical.
The useful question for the manager is: how much of the result came from selection, and how much from when you invested? To investigate, I would request the distribution of investment cost by initial investment year, alongside stage, geography and ownership. These are descriptive data: they cannot independently establish skill or luck, but they make the GP’s explanation testable.
Comparing funds at the same age has a limitation too. Looking at every fund in its fifth year improves comparability of maturity, but does not remove differences in the financing and exit markets each experienced. A single adjustment cannot resolve both questions.
3. Mean, median and top quartile answer different questions
A benchmark should make clear what each column represents. The glossary in Cambridge Associates’ VC benchmark distinguishes pooled cash-flow calculations from statistics on individual funds and defines the upper quartile as a threshold. Source: US VC benchmark glossary, June 2019, used here for definitions.
Measure | What it describes | Misinterpretation to avoid |
|---|---|---|
Pooled return | Performance calculated by aggregating the sample’s cash flows and values | Treating it as a simple average of fund IRRs |
Median | The middle value among individual fund results | Treating it as the return on all capital invested |
Top-quartile threshold | The entry point to the highest 25%, for a given metric and cohort | Confusing it with the average of the best-performing funds |
LP portfolio return | The result of the investor’s actual allocations and cash flows | Assuming it equals the benchmark or the average of selected managers |
Consider a purely illustrative example, using no Atlas data: five fully liquidated funds, each with the same paid-in capital, produce final net multiples of 0.5x, 0.8x, 1.0x, 1.5x and 6.0x. The median is 1.0x; the aggregate multiple is 1.96x. An LP investing equal amounts only in the first four would have earned 0.95x.
The exceptional return is real, but it is unevenly distributed. This example does not establish that adding more funds is enough to produce a good outcome: it specifies no selection probabilities. It shows that the performance of the whole sample and an investor’s experience can differ substantially. Without cash-flow dates, these multiples cannot tell us the annualized return.
4. Define the comparison before seeing the result
A serious comparison starts with criteria fixed before checking the fund’s ranking: source, strategy, vintage, currency, observation date and metric. If the universe changes whenever performance looks less favorable, the ranking loses its usefulness for decisions.
Specialization deserves consideration, but can become an excuse. Comparing a specialist pre-seed fund with all global venture capital may reveal little; narrowing the sample until only a few almost identical funds remain may create false precision. I would ask for two views: a relevant peer group to assess execution of the strategy, and a broader reference to assess its opportunity cost.
The sample matters. Carta’s Q1 2026 report covers 2,775 funds closed since 2017; approximately 89% are smaller than $100 million. That helps explain the dataset’s composition. It is not a neutral representation of every segment of global venture capital. Source: Carta.
Finally, a quartile is a relative result: a fund’s position can change even if its own performance stays constant while peers’ results move. Asking for the distance from the threshold, alongside the label, helps avoid assigning too much significance to small differences.
5. Persistence exists, but use the information available at the time
Harris, Jenkinson, Kaplan and Stucke find persistence in VC performance even when using information available during fundraising. Their 2020 working paper, with data through June 2019, reaches different conclusions for buyouts. The finding is neither that track records are useless nor that they identify subsequent winners with certainty. Source: study, sections 3 and 5.
For an LP, the more interesting test is to reconstruct the decision. What numbers were available when the GP sought commitments to its next fund? Which companies supported the valuation? Which parts of the thesis were subsequently confirmed, and which developments surprised the manager too?
Assessing a past investment today using a quartile that became established years later introduces information the investor did not have. This is hindsight bias: selection looks straightforward because the outcome is already known.
Historical evidence must then connect to the current vehicle. If the decision-making partners, fund size, entry stage or geography change, the manager needs to explain why the capabilities behind previous results should transfer. Continuity of brand does not establish continuity of process.
6. The 20VC question: access to what, and at what weight?
In his August 4, 2025 conversation with Harry Stebbings, Miles Dieffenbach asks new allocators a difficult question: can they actually access managers capable of justifying the risk? Around 13:21–16:23, the discussion also addresses the ability to select early-stage opportunities outside the consensus. These are an investor’s judgments, not a universal statistical rule about the segment. 20VC episode and context; transcript consulted.
The study cited above also finds an average PME above 1 for second-quartile VC funds against the S&P 500. Claims that only the top decile suffices therefore need to be read alongside their chosen benchmark and sample.
Our interpretation is that “access” must translate into an amount and concrete terms. Receiving a pitch deck, being admitted to a vehicle and obtaining a meaningful allocation are three different situations.
An LP might secure a small allocation to the desired fund and deploy most of its budget elsewhere. The relevant return is that of the actual combination, including any additional layers of costs. Access to a well-known firm should also be distinguished from access to its particular investment strategy.
For an emerging manager, the evidence is different: it cannot offer a final ranking for its first fund. It can make the origin of opportunities, decisions taken before consensus formed, prices paid and reasons for declining investments verifiable. An incomplete track record is uncertainty to assess; it establishes neither superiority nor inferiority.
7. Beating other funds is not enough to justify illiquidity
A fund can outperform its peers and still be unattractive to a particular investor. The GP’s competitive reference and the alternative available to the LP are not necessarily the same thing.
A public market equivalent, or PME, compares private cash flows with a public-market investment while accounting for their timing. Different methodologies exist, so identify the one being used. Cambridge, for example, describes its mPME variant explicitly in its benchmark commentary. mPME methodology.
In our assessment, a public index should be selected for an explicit economic reason, not because it creates the most flattering comparison. A broad equity index and a technology-heavy index represent different alternatives. For an LP whose reference currency is the euro, cash-flow currency and any hedging also need consistent treatment.
PME does not automatically eliminate every difference in risk, concentration and liquidity. Nor does it capture every cost of the LP’s program on its own: capital kept available for calls, manager-selection work or intermediary structures. A sound comparison is a starting point for deciding whether the expected benefit justifies those constraints.
8. Read distributions alongside capital still committed
Liquidity has at least two levels: what an individual fund generates and what is needed to sustain the investor’s overall program. A mature fund can distribute while younger funds call capital. The LP needs to understand the balance, and how predictable that balance is.
A practical test asks what would happen if expected distributions arrived two years late while capital calls continued. This is a working scenario, not a forecast. It reveals whether the allocation would force sales of other assets, interrupt new commitments or prevent participation in subsequent opportunities.
Even a rule of considering only liquidated funds is not always a solution. At an interim observation date, it may disproportionately select funds that finished earlier, excluding still-active funds with different outcomes. A better approach keeps the entire cohort visible, distinguishing distributions, residual value, age and funds that have actually closed.
Realized performance provides stronger evidence than estimated value. But asking a manager to maximize immediate distributions can conflict with value creation. Exit decisions should reflect achievable prices and alternatives, while the LP sizes its commitment to withstand uncertainty about timing.
9. What to request before the next commitment
An investment memo should allow an outside reader to reconstruct the comparison. I would request six things:
Benchmark definition: source, inclusion rules, sample size, vintage, stage, geography, currency and date, with an explanation for changes.
Reconcilable results: net performance, paid-in capital, distributions and residual value, with enough cash-flow detail to recalculate the metrics.
Valuation history: results presented during earlier fundraises, not just the latest available pitch deck.
An explanation of returns: the contribution of major holdings, the timing of investments and the role of the people who will manage the new fund.
An obtainable allocation: the amount available in the desired vehicle, economic terms and implications for the rest of the LP’s portfolio.
A liquidity scenario: outstanding calls, assumed distributions and the ability to accommodate delays without forced sales.
The final question is more demanding than “are you top quartile?”: what result does the current process make plausible, for the capital I can actually allocate, with what risks and against which alternatives? A credible manager should help the LP build that answer, even when the most flattering benchmark tells a simpler story.
For more on Atlas’s approach, read the fund manifesto.
To discuss the assessment of early-stage and AI venture capital funds, contact Vitantonio Santoro on LinkedIn or email vitantonio.santoro@atlassgr.com.
Comments