To fee or not to fee…

The Rate Was Never the Problem. The Duration Is.

Five years ago this firm published a paper arguing that the management fee is the wrong thing to argue about. The data behind that paper has since been overtaken by the longest liquidity drought in the history of the asset class. This addendum re-tests every claim we made.

• Venture is the only major strategy whose fee did not move. Buyout's mean investment-period fee fell from 1.81% for 2015/16 vintages to 1.61% for 2025 vintages. Carta's median venture fee has been exactly 2.00% for eight consecutive vintages.
• "Fees are returned first" is a priority of payment, not a guarantee of payment. Fewer than 20% of 2017 and 2018 vintage funds have reached 1.0x DPI at ages eight and nine. Pooled DPI for the 2019 vintage is 0.12x.
• A fee designed for ten years is being drawn for eighteen. Total fee load rises from roughly 15% of committed capital to 21–24%. In extension years, 41% of funds keep charging the same rate.
• The alignment boundary moved without anyone renegotiating anything. Duration alone cut the present value of a partner's carry by 43%. The fee income did not move.
• Investors won transparency, not economics. 42% of LPs report better communication from their GPs over the past year. 3% report better fees.

A Fee in Winter: Written at the Exact Top

In 2021 my colleague Adam Day wrote that the management fee is a red herring. He was arguing against a real error — investors who reject an otherwise excellent fund over twenty-five basis points, while waving through a deal-by-deal waterfall and an uncapped expense agreement. That error is still being made. The paper was right to name it.

What the paper could not know is that it was written at the exact top. Between the fourth quarter of 2019, when its performance data was pulled, and today, United States venture capital has produced the longest stretch of negative net cash flow to investors in the history of the asset class — $196.9 billion more called than returned between the start of 2022 and the end of 2025. Paper value doubled. Cash did not follow.

That single fact changes the arithmetic of a fee. Not the theory — the theory in the original paper is sound, and most of it survives this review intact. But every argument in that paper rested on an assumption so ordinary in 2021 that nobody thought to state it: that a venture fund is a ten-year instrument that eventually returns more than it called. Remove that assumption and the conclusions do not all survive.

This addendum is written in the same spirit as the original: follow the data, and revise where the data says to. Golden Section charges management fees. We have an obvious interest in the conclusion that fees are unimportant. That is precisely why this paper leads with the places where the 2021 argument breaks.

The Scorecard: Five Claims, Marked

The Fee That Wouldn't Fall: Every Strategy Moved Except Venture

The 2021 paper opened with Preqin's 2016 fund-terms data: venture's mean investment-period fee at 2.07%, its median at 2.00%, higher than every asset class in the sample except distressed debt. It noted that fees were "in general trending down as competition increases for investors."

That trend was real. It simply did not include venture.

The 2024 venture mean almost certainly overstates the direction — see the caveats below. Preqin's median for the same vintages is 2.05%. Read the direction, not the decimals. Growth equity edged down; Preqin's published figures for it are behind a subscription.

Buyout's mean investment-period fee has fallen for three consecutive years to 1.61% for 2025 vintages — the lowest Preqin has recorded since it began tracking in 2005. Real estate reached a twenty-year low at 1.31%. Venture did not move.

There are two honest caveats. The first is that the 2024 venture mean of 2.24% is above the 2016 mean and almost certainly overstates the direction: 2024 was a thin venture fundraising year, and a mean computed across a small number of predominantly small funds drifts upward. The second is more interesting. Part of the reason the industry-wide average is falling is not that any general partner cut a fee. It is that capital concentrated into large funds, and large funds have always charged less. Roughly 46% of the capital raised in 2025 went to the ten largest funds, against 34.5% the year before. A falling average can be produced entirely by mix.

The Tail Thinned. The Middle Did Not Move.

Median: eight consecutive vintages (2018–2025), sample of roughly two thousand United States funds, median exactly 2.00% in every one. The per-vintage values behind the above-2.5% share between 2016 and 2023 are not published; only the endpoints are.

Both rows are true at once, and the tension between them is the whole story of venture fee pressure since 2021. The expensive tail genuinely thinned — investors won that fight. What they did not win was any movement in the middle. Across every vintage from 2018 to 2025, on a sample of roughly two thousand United States funds, the median venture management fee during the investment period is exactly 2.00%.

The same pattern shows up everywhere the data is granular enough to see it. Carried interest has not compressed either — 20% remains the standard, with a widening tail in both directions: a 15% floor appearing at the tenth percentile of the smallest funds, and a 25–30% premium tier at the top of the largest. Roughly a third of funds Cooley reviewed in 2026 carried some form of premium carry.

And the hurdle rate — the single mechanism the 2021 paper recommended most strongly — did not arrive. The paper closed one of its sections with the line: "A better way to protect downside is insist on a preferred return hurdle." Between 2021 and 2026 the risk-free rate went from roughly zero to roughly five percent and back down again. If ever there was a moment when hurdles should have entered venture, that was it. They did not. Fewer than one venture fund in ten under $10 million has a hurdle; fewer than one in eight above $100 million does.

What Investors Actually Won: Transparency, Not Economics

The ILPA asked 101 senior investor professionals what had improved in their general partners' behavior over the prior twelve months. The pattern is unambiguous, and it is not a fee story: 42% named communication. 3% named fees.

One respondent summarized it precisely: "slightly more leverage in transparency and governance topics, but no change in alignment of interest or economic topics." Investors won the right to see. They did not win the right to pay less.

They did win one thing, and it moved off the page entirely. Investors are now asking for a median of 33 cents of no-fee co-investment for every dollar of fee-bearing commitment — an effective reduction in manager revenue of roughly a quarter, achieved without touching a single headline rate. Two-thirds of managers charge no management fee at all on co-investment, and just over half charge no carry.

That is a real concession, and it is invisible in every fee table published anywhere, including the ones in this paper. It is also distributed extremely unevenly. Sixty-nine percent of investors managing more than $16 billion report more negotiating leverage than a year ago. Among those under $2 billion, only 33% do. The fee relief that exists in this market is going to the investors who least need it, in a form that never touches the document a smaller investor signs.

Return to Sender: A Priority of Payment Is Not a Payment

The load-bearing sentence of the 2021 paper is this one: "most funds must still return every dollar of called capital before any carried interest is earned. In other words, management fees are returned."

Structurally, this is still true, and more universally true than the paper claimed. Roughly 90% of the funds Cooley reviewed in 2026 use a European whole-fund waterfall. Every dollar of fee sits inside the capital that must come back before a general partner sees a dollar of profit. The paper called the fee "a loan that is paid back." The document still says so.

But a loan that is paid back and a loan that is prioritized for repayment are different instruments, and the difference is the entire content of the last five years. The sentence describes an ordering. It does not describe an outcome. For any fund that never crosses 1.0x, the fee is not returned. It is a permanent transfer, and it is the only compensation the manager receives.

In 2021 that distinction was academic, because most funds crossed 1.0x. It is no longer academic.

Three age points are shown because those are the ones the underlying source gives. Above age ten the comparison is flattering. At age nine and below, every cohort is behind. PitchBook Benchmarks, Q4 2019 and Q1 2025 editions.

This is the table that should change how the original paper is read. It holds fund age constant and asks what a venture fund of a given age had returned in 2019, when the original paper's benchmark was set, against what a fund of that same age has returned today. Above age ten the comparison is flattering — those funds harvested into the 2020–21 window and beat their predecessors handily. At age nine and below, every single cohort is behind.

Carta's earlier quarterly editions put numbers on the same pattern: as of 30 September 2025, 42% of 2020-vintage funds and 25% of 2021-vintage funds had made any distribution at all. As of 31 December 2025, the shares were above half for 2020, about a third for 2021, and just under a quarter for 2022 and 2023. Carta's universe is platform-selected and skews toward smaller and emerging managers, and is therefore a useful cross-check rather than a confirmation.

Set that against the 2021 paper's own benchmark. It described "something below a 1.3 DPI" as underperformance relative to the pooled average of mostly-realized funds, and modeled its downside case at 1.2x. That figure was accurate: the 2004–2008 vintages averaged 1.31x pooled DPI as of the end of 2019. Today, the equivalent mature cohort — 2010 through 2014, now aged eleven to fifteen — averages 2.11x. Funds that lived long enough to harvest did better than the paper assumed, not worse.

The distinction matters because a fee is charged in the present against an outcome in the future. Every fund in the 2016–2022 range is paying management fees today on the strength of a return that has not happened and, on current evidence, will happen later than anyone underwrote. The cohort now paying fees — the 2018 vintage, at age seven — sits at 0.39x pooled DPI.

Marked, Not Made: What Happens When Marks Are Tested

A reasonable objection at this point is that DPI is the wrong lens on a young fund. Venture takes time; unrealized value is not lost value; the money is in the portfolio. That objection is correct as far as it goes. Here is how far it goes.

For every vintage from 2016 onward, between 63% and 98% of reported value has never been converted to cash. That is not, by itself, an indictment — young funds are supposed to look like that. The question is what happens to those marks when they are tested. The 2016 through 2019 vintages are the test case, because they are old enough to have been marked up in the boom and old enough to have been marked back down.

PitchBook Benchmarks, Q1 2022 and Q1 2025 editions. The Q1 2022 edition is no longer publicly retrievable; see the method note. The per-vintage unrealized-share series is stated as a range only in the underlying material.

Between the first quarter of 2022 and the first quarter of 2025, the 2019 vintage's pooled paper value fell by 37%. Its realized cash moved from 0.10x to 0.12x. Its pooled internal rate of return went from 57.29% to 5.72%. Management fees were charged, in full, throughout.

Across all strategies the average was 87%, two hundred basis points lower than the year before. Jefferies 2025 Global Secondary Market Review; PitchBook Q3 2025 Quantitative Perspectives.

The secondary market prices the remainder honestly. Whatever the exact figure, the market's considered view is that a dollar of unrealized venture net asset value is worth roughly four-fifths of a dollar — and that the discount widens with every year the fund stays open. The management fee, meanwhile, is charged at par.

The Running Ledger: Fees Drawn Against Cash Returned

Here is the arithmetic the 2021 paper did not need to do, because in 2021 it would have been unremarkable. For a fund of any size, at each vintage's current age, compare two numbers: the management fees drawn to date, and the cash returned to date.

The 2019 and 2020 figures are given in the analysis as "very nearly as much in fees as the investor has received in cash" and the 2021 figure as "almost four dollars for every one returned"; both are shown as approximations. Fees are modeled on Carta's median step-down path applied to committed capital.

Three things must be said about this table, or it will be misused.

1. It is a snapshot of an unfinished process, not a verdict. Every one of these vintages holds substantial unrealized value. As that value converts, the ratio falls, and for the good funds it will fall a long way. A 2021-vintage fund at 0.02x DPI after four years is not a failed fund; it is a young fund in a closed exit market.
2. It is modeled, not measured. Fees are computed on Carta's median step-down path applied to committed capital. Funds that shift the fee base to invested capital net of write-offs after the investment period — 79% of managers do — will run below this line.
3. It does not net the two. DPI is already reported net of fees. The table puts the two flows side by side as an investor's running account, which is how an investment committee actually experiences them.

The Long Goodbye: The Variable We Underweighted

If there is one variable the 2021 paper underweighted, it is not the fee rate, the carry percentage, the hurdle, or the waterfall. It is time.

The paper's model, like every fund model of its era, assumed a ten-year life: five years of full fee, five years at half. That assumption is now wrong, and it is wrong in a direction that compounds.

In April 2026 Robert Bartlett and Paolo Ramella published the first systematic study of the question, using PitchBook data on funds formed between 1995 and 2014. Their title is the finding: The Disappearance of the Ten-Year Fund. Unrealized value late in fund life has risen sharply across vintages, they conclude, and especially in venture capital. Many funds continue distributing capital past year twenty. Practitioners describe early-stage fund life as fifteen to eighteen years; SVB's work puts full return of capital for top-quartile funds at sixteen to twenty.

Nobody has published the obvious next number: what those extra years cost. So we computed it.

Modeled on the step-down path actually observed in the market. The 2021 paper's own assumption produced 15% over a ten-year life; the observed path produces 15.5%. Golden Section estimate.

Under the 2021 paper's own assumption, a fund draws 15% of committed capital in fees over its life. Under the step-down path actually observed in the market, a ten-year fund draws 15.5% — the paper had it essentially right. Extend that same fund to fifteen years and the load rises to 21.0%. Extend it to eighteen and it reaches 24.3%. On a $50 million fund, that is the difference between $7.5 million and $12.2 million, drawn from the same investors under the same document at the same headline rate.

Nobody Wrote Down What Happens in Year Eleven

The step-down is real and it is more common than the 2021 paper suggested. Carta finds that 81.9% of venture funds step the fee down at least once, and 51.5% do so twice or more. But the step-downs are gentler than the mental model: the median first step-down is ten basis points, and it takes six of them to travel from 2.0% to 1.1%.

More importantly, the schedule usually stops stepping down before the fund stops running. In the largest recent survey of fund terms — 110 United States alternatives managers — 60% of funds permit two extensions and 30% permit three, almost always in one-year increments. On what the fee does during those years, the survey is blunt: 41% of funds continue charging the same management fee rate through extension years, and 91% grant those extensions in one-year increments, decided one year at a time.

Two funds in five wrote a fee schedule for a fund life that no longer exists, and then carried the same rate straight through the years that replaced it. This is not predatory. It is an artifact — nobody drafting a venture partnership agreement in 2018 was pricing an eighteen-year fund. But it means the negotiation that matters most now is one that mostly is not happening, and where it does happen it happens at the worst possible moment: after the money is committed, mid-extension, one year at a time, with the manager holding the portfolio.

Investors themselves have reached the same place by a different road. Asked in early 2026 what should happen to funds that outlive their usefulness, 54% of limited partners named management fee step-downs as the preferred remedy — the most-selected answer, well ahead of resetting incentives at 18%. The same survey found 54% expect the number of such funds to rise over the next two years.

The Model, Re-Run: The Framework Was Right. The Prior Was Wrong.

The heart of the 2021 paper was a sensitivity analysis: a $50 million fund, ten-year life, run at 2/20, at 1.5/25, and at 0/30, first at a 2.5x multiple and then at a 1.2x. Its conclusion, in the paper's own italics, was "lower management fees, worse results" — because at 2.5x, the fund that gave up fee for carry returned less to investors than the fund that kept its fee.

That arithmetic is correct and we reproduce it exactly. Here is the original table, verified:

Now the same three structures, at the same fund size, with two changes: a fifteen-year life instead of ten, and the step-down path actually observed in the market rather than the assumed halving. Nothing else moves.

Golden Section model, on the convention described in the method note. The stated multiple is a gross value multiple on committed capital.

The 2021 conclusion survives at the top: at 2.5x the three structures still land within four percent of each other, and the fee-light structure is still no bargain. What changed is everything below 1.5x. At a 1.0x gross multiple — not a disaster, simply a fund that returns what it called — the spread between the highest- and lowest-fee structure is $10.5 million on a $50 million fund. That is 21% of committed capital, decided entirely by a term the paper told investors to stop worrying about.

The paper anticipated this. It wrote: "giving up carried interest to reduce management fees seems to create some downside protection if a fund performs poorly." Then it asked the reader the right question — "How well do you expect the fund to perform?" — and, reasonably for 2021, treated the strong answer as the base case.

There is a second, subtler revision. The paper argued that an investor expecting 3x or better from venture should prefer the lowest carried interest, because at high multiples carry costs more than fees. That remains arithmetically true. But it now sits alongside a fact the paper could not see: the distribution of outcomes has widened at the bottom and the timeline has lengthened at both ends. An investor underwriting 3x should still weight carry over fee. An investor underwriting the actual current distribution of venture outcomes should weight both, and should weight duration above either.

Moving the Boundary: Duration Cut the Carry by 43%

The most original section of the 2021 paper borrowed Tversky and Kahneman's cumulative prospect theory to define what it called the alignment boundary: "the point at which a general partner no longer thinks the carry is worth working for." The paper argued that boundary is crossed when fee income grows so large, and carry so uncertain, that the manager stops weighting the carry at all.

Its test was a thought experiment. A $200 million fund at 2% supports a $250,000 partner salary; that partner's share of carry in a moderately performing fund is around $2.5 million; ten times the annual salary, delivered at the end. The paper concluded, reasonably, that this passes the sniff test.

Re-run the thought experiment with one change — not to the fee, not to the carry, only to when the carry arrives.

Discounting alone — before any adjustment for the probability that the fund clears its preferred return, before any haircut for wider dispersion, before the ambiguity the paper itself invoked — removes 43% of the incentive between year ten and year eighteen. Meanwhile the salary is unchanged, certain, and now paid for eight additional years. Golden Section model.

The 2021 paper's own logic delivers the conclusion. It argued that carried interest is weak compensation because it is remote from the work: "the further out this compensation is from the work effort in terms of time, the more disconnected the reward is from the agent." It used the analogy of a salesperson paid a commission eight years after closing the contract, and concluded — correctly — that such a person needs a real base salary.

The analogy has aged. It is now a commission paid eighteen years after closing. Every argument the paper made for paying managers a proper fee got stronger. And every argument for believing that carry aligns the manager got weaker, by exactly the same mechanism.

Two Venture Markets: One Name, Two Economies

The 2021 paper made a point about fund size that has since been overtaken by events. It observed that despite a doubling of venture assets under management, the median fund was $50 million smaller than its 2007 counterpart and the average $24 million smaller — and drew the reasonable inference that "the typical venture fund is not structured to maximize fees."

The median has kept falling. The average has gone the other way, violently.

A ratio of almost twelve to one. PitchBook-NVCA Venture Monitor, Q2 2026; PitchBook revises fundraising figures substantially between editions.

The median is falling and the average is climbing, at the same time, in the same market. Two-thirds of funds closed were under $50 million and took 4% of the capital; funds over $1 billion took 68.3%. Three firms — Andreessen Horowitz, Founders Fund and Thrive — raised 48% of all United States venture dollars in that half-year.

The 2021 paper's conclusion that overcapitalized funds are not necessarily misaligned remains logically correct. But the sentence was written about an exception. It now describes where most of the industry's fee-generating capital lives. Two-thirds of the fee base sits in vehicles large enough that the fee is, unambiguously, a business rather than a cost-recovery mechanism — and 52.3% of undeployed capital sits in funds of $500 million or more, which are 6.7% of the funds closed in the last four years.

Both halves of the paper's fund-size argument still apply. They simply apply to two different markets that now share a name.

What Still Holds: Four Arguments Strengthened

Four of the 2021 paper's arguments have not merely survived. They have been strengthened by data that did not exist when it was written.

1. Fees pay for a business, and the smallest funds are not making money on them. The paper argued that a micro fund charging 2% would exhaust its operating budget before partner salaries. Carta has since measured it: funds between $1 million and $10 million spend 3.4% of committed capital on operating costs over five years, against 1.0% for funds above $100 million. Legal is 8.6% of a small fund's expenses and 22.7% of a large one's; tax is 12% against 6.1%. Fixed costs do not scale, and the small manager's higher rate is an arithmetic consequence, not a premium.
2. The small manager is not the misaligned one. Smaller venture funds commit more of their own capital, not less: the median general partner commitment is 2.0% of fund size for funds between $1 and $10 million, with a 75th percentile of 6.15%, against 1.5% and 2.1% respectively for funds above $250 million. If skin in the game is the alignment test, the smallest funds pass it most convincingly.
3. Structure still beats rate, in the specific places the paper named. Every term the paper told investors to read is still where the money is. The waterfall is nearly universal in its investor-friendly form. Expense agreements have quietly drifted the wrong way — median organizational expense caps rose from around 20 basis points in 2019–21 to 25 by 2024–25, which is why ILPA in May 2026 proposed capping them at the lower of 5 basis points of target assets or $10 million, a fivefold tightening. Recycling remains the one place venture is the most disciplined asset class in private markets: only 15% of venture funds leave recycling uncapped, against 40–45% in private equity and 78% in credit.
4. The hurdle recommendation was right and remains unheeded. The paper's single sharpest recommendation — that the better way to protect downside is a preferred return, not a fee cut — is more correct now than it was then, and less adopted. In a world where the modal outcome for a live vintage is somewhere near 1.0x, a hurdle is the term that does the work a fee negotiation cannot. Venture has spent five years not adopting it.

A New Set of Questions: In Order of What Each Is Worth

The 2021 paper gave investors a rate screen: worry if the fee is above 2.75% or below 1.2%. That screen cannot see the thing that now matters. Replace it with five questions, in order of how much money each one is worth.

1. What is the total fee load, in dollars, to year eighteen? Not the rate — the schedule, multiplied out, on committed capital, with the extension years included at whatever the document actually says. Under 18% of committed capital is unremarkable. Above 24% deserves an explanation. Any manager who cannot produce this table in an afternoon has not thought about it, which is itself the answer.
2. What does the agreement say about years eleven through eighteen? If the answer is "nothing," that is the negotiation. Ask for an automatic, written schedule to a cost-recovery amount or to zero, on a date certain. This is the highest-value term available in a 2026 venture partnership agreement and it is nearly free to the manager.
3. Is there a preferred return, and does it compound? Still the cleanest downside protection in the instrument, still rare in venture, still worth more than any concession on rate.
4. What is the general partner's salary, and what is the profit multiple on it over a realistic fund life? The paper's own test, run to eighteen years rather than ten. A multiple above 15x on realistic duration is a manager whose incentive still points the right way.
5. What is the running ledger? Fees drawn to date against cash returned to date, disclosed each quarter, for every prior fund. No manager currently reports this. Every manager can compute it in a spreadsheet. It is the single most informative number an investor could be given, and the fact that nobody publishes it is telling.

Conclusions, Revised: Amended Rather Than Replaced

Four Open Questions This Research Could Not Answer

• What share of venture funds are currently operating past their stated term? No one publishes it.
• What fees are actually charged inside continuation vehicles and secondary special-purpose vehicles — the fastest-growing fee-bearing structures in the market? No public dataset exists.
• Do smaller funds still outperform on realized distributions, or only on marks? Every published cut of that finding uses TVPI or IRR.
• What is the aggregate incremental fee the industry has drawn from the extension of fund life since 2020? The estimate in this paper is the only one we are aware of, and it is a model, not a measurement.

Method, Caveats and Notes

The modeling in this paper follows the 2021 original's convention exactly, so that the tables are directly comparable. In that convention, the stated multiple is a gross value multiple on committed capital; the general partner receives fees plus carried interest on value above committed capital; the limited partners receive the remainder. General partner figures are inclusive of management fees. This is a simplification of fund economics — it ignores capital call timing, recycling, expense agreements outside the fee, and tax — and it was a simplification in 2021. It is retained here for comparability, not because it is complete.

Three Data Points Readers Should Treat With Care

• The frequently quoted figure that private markets distributed only 6% of assets under management in the year to June 2025, against a 16% historical average, comes from McKinsey's Global Private Markets Report 2026 and describes all private markets, dominated by buyout. McKinsey does not break out venture. It is widely miscited as a venture statistic. It is not used as one here.
• The finding that smaller venture funds outperform larger ones in recent vintages is real but is measured in TVPI and IRR — both marks — in every published source we could find. Given what the performance sections show about the reliability of marks in this period, that finding should be held loosely until someone publishes it on realized distributions.
• Preqin's 2024-vintage venture mean of 2.24% conflicts with chart readings from Preqin's own 2025 Fund Terms Advisor webinar materials, which appear to show materially lower venture means on a different weighting basis. We have used Preqin's published figures and flagged the conflict rather than choosing between them.