Practice chapter · in depth
By Niko Hatziiosifidis · last reviewed 23 Aug 2026
The return on a semi-liquid fund’s factsheet is almost always a time-weighted NAV return — closed-end private equity funds, by contrast, report a money-weighted IRR. The two figures measure different things and are not directly comparable: a 20% IRR at a closed-end fund produces the same terminal wealth over ten years as an evergreen at around 11% p.a. (Partners Group 2024, provider source). To examine a track record you need multiples (TVPI/DPI) and an understanding of the IRR pitfalls (subscription lines, timing) — and with evergreens additionally a critical look at seed portfolios, day-1 uplifts from secondaries and the cost base.
The IRR (internal rate of return) is the discount rate at which all the payments into and out of an investment come to a present value of zero. It is money-weighted: how much capital is at work in the fund, and when, helps determine the result. That is precisely what makes it the standard for closed-end funds — the manager decides on capital calls and distributions, so the measure is meant to reward that timing. But the same property leaves the IRR open to attack: an early, large book gain on a small amount of deployed capital produces a spectacular IRR that barely wears off later. And with subscription credit lines — borrowing with which the fund pre-finances purchases and delays capital calls — the IRR clock can be actively shifted: the effects measured in the academic literature range, depending on fund age, from a few tenths of a point to almost ten per centage points (details in the professional deep dive; Albertus/Denes 2019, working paper). Hence the famous title of a 2006 Oaktree memo by Howard Marks: “You Can’t Eat IRR” — you cannot live off a per centage figure, only off money that has actually come back. An IRR is not an account balance: it does not tell you how much wealth is there at the end, only how fast the capital tied up at any given time compounded for as long as it was tied up.
Multiples answer the question the IRR leaves open: how much money did how much money turn into? MOIC (multiple on invested capital) relates total value to invested capital — careful: often gross, that is before fees and carry, calculated at deal level. TVPI (total value to paid-in) is the investor’s view: (distributions + residual value) ÷ capital paid in, ideally net of all costs. With every track record, ask explicitly whether gross or net figures are being shown — the difference is substantial. TVPI breaks down into DPI (distributions to paid-in: what has actually been paid out) and RVPI (residual value to paid-in: what is still carried on the books as a valuation). DPI is the “hard” component; RVPI hangs on the manager’s NAV valuation. For context: according to Phalippou (2020), private equity funds have delivered ~1.5x net since 2006 ≈ ~11% p.a. — while paying managers ~US$230 billion in carry (paper). And Hamilton Lane does the arithmetic: 12% p.a. time-weighted produces 2.5x after 8 years — a level only ~6% of closed-end funds reach as a TVPI. An “unspectacular” evergreen can therefore build more terminal wealth than an IRR star.
The time-weighted return (TWR) cuts the investment period into sub-periods at every inflow and outflow, calculates the return of each and chains them geometrically. The effect: the timing of investor money drops out. That is the logic of the CFA Institute’s GIPS standards for open-ended vehicles — the manager of a fund you can subscribe to daily or monthly has no say over when investors come in and go out, so it should not make them look better or worse. Evergreen and ELTIF factsheets therefore show TWRs like any retail fund; in practice the period return is often approximated using Modified Dietz: (ending value − beginning value − net cash flows) ÷ (beginning value + time-weighted flows), calculated monthly and then chained.
A worked example — same fund, two truths. You pay in €100. Year 1 runs strong: +30%, NAV €130. The fund distributes €100; €30 stays invested. Year 2 runs weak: −10%, the €30 becomes €27, which is paid out. Time-weighted: 1.30 × 0.90 = 1.17 → 17% over two years, that is ≈ 8.2% p.a. Money-weighted (IRR): the cash flow series −100/+100/+27 gives ≈ 22.1% p.a. — almost three times as much, because most of the capital only lived through the good year. Your actual outcome: €127 out of €100, a TVPI of 1.27x (DPI 1.27x, RVPI 0). The 22% IRR applied only to the capital while it was working in the fund — the €100 distributed sat in your account afterwards. The rule of thumb: IRR measures the compounding of capital tied up, TWR the quality of the portfolio, the multiple your terminal wealth. Chapter 1 shows the consequence at scale (→ Chapter 1): a closed-end fund at 20% net IRR ≈ 2.8x over 10 years is the equivalent of an evergreen at ~11% p.a. — partly because up to half of the commitment is often never called (Partners Group 2024, provider source).
Four structural reasons. First, age and provenance: many semi-liquid vehicles are young; a track record of two or three years does not cover a full cycle. Some funds are also “incubated” with selected assets before distribution starts — that opening phase then feeds into the since-inception return even though it has little to do with today’s much larger portfolio. PitchBook analyst Hilary Wiek warns accordingly that a house’s historical drawdown IRRs are no reliable predictor of its evergreen returns (PitchBook, “Evergreen Funds: We Have Questions”, Q1 2026). Second, the day-1 uplift: evergreens frequently buy secondaries at a discount to NAV — buyout secondaries traded at 92% of NAV across the full year 2025 according to Jefferies. Hamilton Lane describes the mechanism openly: the buyer can acquire an asset below NAV “and then hold it at the reported NAV”. A purchase at 92, valued immediately at 100, shows an arithmetic instant gain of ~8–9% — a return from accounting rather than from value creation, and one that flatters young track records above all (Wiek makes the same point). Third, NAV smoothing: quarterly, model-based valuations understate volatility and make return series look more stable than the portfolio is (→ NAV chapter). Fourth, the fee basis: fees are charged on the NAV, performance fees partly on unrealised gains — with very different total costs depending on the share class. The same portfolio return can therefore arrive as a very different factsheet return.
Against closed-end funds: think in terminal wealth, not in IRR. Convert both sides into the same question: what becomes of €100,000 over ten years — including the capital that, in the closed-end fund, sits uncalled or lies idle after distributions? Neuberger Berman has modelled exactly that (assumption: an identical net asset return of 13.7%): a series of closed-end funds reaches ~2.7x over ten years, an evergreen with a 15% liquidity buffer ~3.2x, without a buffer ~3.6x — because the capital works immediately and continuously. That is a provider’s own calculation with favourable assumptions (no redemption effects, equal asset quality), but the direction is consistent with the Partners Group arithmetic from Chapter 1. Against ETFs: calibrate the yardstick first. The academic reference is the PME method (public market equivalent): all the fund’s cash flows are notionally invested in an equity index, and the comparison runs over identical payment streams — developed by Long/Nickels (1996), established by Kaplan/Schoar (2005). The result across >1,800 funds: buyout vintages before 2006 beat the S&P 500 by +3–4% p.a. (PME); from 2006 onwards they were ≈ in line with the market (Harris/Jenkinson/Kaplan 2016, paper). With evergreens the comparison is simpler — TWR against TWR — but costs decide it: according to Morningstar (2026) the ongoing costs of semi-liquid funds average ~3% p.a., and only a few beat their benchmarks after costs. A fair ETF comparison uses the same period, the same region and calculates after all costs — entry charge included.
The subscription-line effect, quantified. Albertus/Denes (2019, working paper) reconstruct from LP cash flows (Burgiss, US buyout, 2014–Q3 2018) what the IRR would have been without fund-level borrowing: funds with subscription lines report an IRR 6.1 per centage points higher (+25% relative to the sample mean); at young funds (vintage 2014 onwards) it is 9.7 points, at older ones only 0.7 — so the effect is largest exactly where track records are built for the next fundraising. Each additional unit of credit leverage adds ~50 basis points. Schillinger/Braun/Cornel (2019, working paper), using a different design, arrive at moderate effects under customary usage but “substantial” increases in time-sensitive measures under extensive usage. The consequence drawn by the ILPA guidance (2020): ask for the IRR with and without the credit-line effect. Important for the subject of this chapter: the line barely affects multiples (interest costs even push them down slightly) — where IRR and TVPI diverge sharply, that is a signal to look closer.
The gross-to-net bridge. Between the gross return of the assets and your factsheet return sit: the management fee (with evergreens on the NAV, not on commitments), the performance fee (with semi-liquid vehicles often on unrealised NAV gains — a conflict of interest criticised by PitchBook/Wiek), fund and vehicle costs (depositary, administration, and target-fund costs where there are fund-of-funds structures), plus share-class-specific distribution and service fees. Morningstar’s ~3% p.a. total cost ratio means: a portfolio has to run in double digits gross for beating an ETF net to be possible at all. The entry charge (with ELTIFs up to several per cent) does not show up in the fund return at all — it reduces your return, not the factsheet’s.
Track record questions for the manager. (1) Is the stated return net of all fees of the specific share class? (2) How long has the vehicle existed — and does “since inception” include an incubation phase with a seed portfolio? (3) What share of the return to date comes from write-ups, in the quarter of purchase, of secondaries bought below NAV? (4) IRR figures from the drawdown history: with or without the subscription-line effect, gross or net? (5) What is the DPI of the predecessor funds cited — not just TVPI/IRR? (6) Who values the NAV, how often, with what lag? (7) Does the same team manage the evergreen as the flagship funds — and how are deals allocated between the two? (8) Which benchmark does the fund measure itself against, and why that one?
The IRR is money-weighted: it depends on how much capital was in the fund and when, and it measures only the compounding of the capital tied up. The “return p.a.” on ETF and evergreen factsheets is time-weighted and screens cash flows out. A 20% IRR therefore does not amount to 20% growth in wealth per year.
It depends on the context: since 2006 private equity funds have delivered ~1.5x net on average (Phalippou 2020); a TVPI of 2.5x is reached by only ~6% of closed-end funds (Hamilton Lane). The DPI is what counts — money is only earned once it has been distributed; a high TVPI with a low DPI rests on valuations, not on cash.
Partly genuine performance, partly mechanics: early book gains on little capital, subscription credit lines (an effect of up to ~9.7 IRR points at young funds, Albertus/Denes) and the property of the IRR that it locks in early successes permanently. So never read an IRR without a multiple.
Methodologically yes — both report time-weighted returns. But it only becomes fair after all costs (avg. ~3% p.a. ongoing costs plus the entry charge), over identical periods and against a suitable benchmark; on top of that, the NAV visually smooths the ELTIF’s swings.
No. A distribution yield is a payment promise, not a performance measure: a fund can distribute while its NAV falls — part of the “return” is then a repayment of capital. What matters is the total return (NAV movement plus distributions).
At least one full market cycle, realistically 5+ years with a stable portfolio. Shorter histories are often distorted by seed portfolios, day-1 uplifts on purchased secondaries and the small base at the start — and a house’s drawdown successes are no reliable substitute (PitchBook 2026).
Last reviewed: 23 Aug 2026 · Methodology · Not investment advice.