Fees, Lock-Ups, and Why the Reported Returns Are Not Comparable

Reading-order note. The v8 architecture places this article last, as #13. It is drafted second, because the pillar directive requires it to be consistently fee- and risk-forward and an article about fees placed last cannot inform the eleven before it. The figures below are this pillar's single arithmetic source, and every subsequent article references rather than re-derives them — the discipline Pillar 20 #6 established after the reverse ordering produced a correction there.

This article establishes all three, because the opening article showed the category has no shared properties — and these three features are the closest thing to one.

The fee stack, computed

The standard structure is a management fee plus a performance fee, and its arithmetic is worth doing rather than describing. Management fee, conventionally around 2% annually — and critically, often charged on committed capital rather than invested capital, so a fee accrues on money the manager has not yet deployed. Performance fee or carried interest, conventionally 20% of profits above a hurdle, often 8%. And frequently a third layer — a feeder fund, platform, or fund-of-funds charge of around 1%, which is the layer most retail-accessible structures add. The reference table below is computed and is this pillar's single source. It assumes a ten-year fund at 2% and 20% over an 8% hurdle, with the management fee applied to the annual return and carry taken on the terminal value in excess of an 8% compounded hurdle.

Three observations, of which the second is the one to carry. The drag rises with performance, because the performance fee takes a share of a larger profit — which is the structure working as designed rather than a flaw. And the right way to express the cost is as a share of the profit, not as a percentage of assets. At a 15% gross return over ten years, gross profit is 305% and the investor keeps 215% — so the fee structure took 29% of the profit. That figure is the honest one, and it appears in no fee table anywhere. And a third layer changes it materially: adding a 1% feeder charge to the 15% case reduces the net return from 12.1% to 11.3%, and at a 10% gross return the net falls from 8.0% to 7.0% — the retail-access layer costs roughly a tenth of the net return at ordinary performance levels.

Lock-ups, and the three reasons the returns are not comparable

Committed capital is not invested capital, and the distinction has consequences. A fund typically calls capital over its first several years and returns it over the following several, with a total life around ten years. Four practical effects. You cannot choose when the money leaves. Capital calls arrive on the manager's schedule, so a committed investor must hold liquid reserves against them — which means the effective allocation is larger than the amount invested at any moment. You cannot exit. Secondary markets for fund interests exist and transact at discounts, so an early exit is available at a price. The J-curve: fees are charged from the start while gains arrive later, so reported performance is typically negative in early years — which is normal, and is also why a short track record says almost nothing. And the commitment outlasts most people's plans. A ten-year lock-up is longer than many readers' job tenure, housing situation, or stated investment horizon. Now the measurement problem, which is the most important content in this pillar and the least widely understood. Private-market returns are usually quoted as an internal rate of return, and an IRR is not comparable to a public-market total return. Three separate reasons, and they stack. IRR is sensitive to timing in ways a time-weighted return is not. Because the manager controls when capital is called and returned, an early distribution can lift a reported IRR substantially without improving the total money returned — which is why the same fund can show an impressive IRR and an unimpressive multiple, and why both figures should always be read together. Unrealised holdings are valued by appraisal rather than by trading. Pillar 20 established what follows: appraisal-based valuation produces smooth series, so reported volatility and reported drawdowns understate what a traded market would have shown. Smoothness in a chart is not stability in an asset, and in private markets a meaningful share of a reported return may be unrealised marks rather than money received. And the aggregate statistics are survivorship-affected. Funds that failed report less completely or stop reporting, and databases are voluntary — so an industry average describes the funds that survived to be counted. The consequence is a rule worth applying to every figure in this pillar: ask whether it is realised or marked, whether it is an IRR or a multiple, and whether it is net of every layer. A gross IRR on partly unrealised holdings from a voluntary database is three distortions in one number.