Private Markets, Retail Edition

Executive summary · free edition · September 2026

Complete in itself. The finding is here, not held back to make you buy the report.

The finding, and it has a date on it

The gate is not a tail risk. It is the product working as designed, and it has already been tested.

Across twenty-three months, 36 repurchase offers were oversubscribed across 19 named vehicles — 29 of them, across 17 vehicles, in the last twelve — with fill rates from 7.38% to 95.15%. The largest fund in the sample filled 48.5% of what investors asked to redeem — 191.8m shares requested, 93.1m accepted. Another filled 13.1%, in two consecutive quarters.

And the timing is the story.

Three of the largest interval funds printed exactly the 7% regulatory ceiling in the March 2026 quarter, and one states outright that it prorated in all four.

Why nobody has published this

A repurchase offer's result is filed but never collected.

The form that announces every offer never reports its outcome. The annual form records only a yes or a no. The portfolio form has nothing at all. The regulator does not gather it, and no aggregated source carries it.

We built it by reading 1,439 filings from 328 filers. The evidence was public, complete, and unassembled — which is the shape of finding this house looks for.

And the table checks against the funds' own arithmetic. Where a fund states its fill rate and files the share counts, our recomputation reproduces its number — eleven times, twice to three decimal places.

Three promises, examined separately

"You can get out quarterly"

Periodic and conditional — and the condition has now bound. One fund repurchased 0.47% of its shares in early 2025 and, a year later, filled 43.1% of a capped offer. One vehicle ran six consecutive oversubscribed offers with the fill rate falling from 92% to 7.38%.

An interval fund's stated repurchase percentage is a ceiling, not a promise.

"The fee is 1.25%"

Interest expense is the largest single line in every levered vehicle, at 55% to 65% of total expenses. A reader comparing "the fee" is comparing the smallest component.

And some cost sits outside every expense ratio. One fund's acquired-fund line of 0.78% explicitly excludes the underlying funds taking "approximately 15%–20% of net profits." One exchange-traded fund's 0.75% unified fee excludes both management and performance fees charged inside its holding structures.

The comparison, without a verdict: one private-credit share class costs 9.96%; a listed high-yield bond fund costs 0.49%. That is 9.47 points. We state both and stop — what they buy is not the same thing, and the comparison is the reader's to make.

"It returned 10% with low volatility"

It reports steady marks. Level 3 shares run 84.9% to 97.1% across the large non-traded vehicles. One reports 0.02% of assets at quoted market prices. Four report no quoted-price investments at all.

A risk-adjusted statistic computed on manager-set marks inherits their smoothness and says nothing. So we did not compute one. Instead, for companies whose same assets have both a reported value and a traded share price, we measured the difference in variability over the same ten quarters:

Same assets, same period, same company. The smoothness is in the measurement, not in the business.

What happens when a mark meets a market

One listed vehicle went from a +2.95% premium to a 47.37% discount in four quarters — a 50.32-point swing. Four listed vehicles spanned +1.10% to −37.91% against the same valuation date.

And where the distribution comes from

Return of capital as a share of distributions runs from 0% at three funds to 26.99% at one and 39.70% at another — the latter being 57.93% of the per-share distribution in a single year. One company's figure was first filed at $10.4m and finally determined at $69.9m, 6.7 times higher.

Two of the five large non-traded vehicles do not publish the split at all.

What is actually inside these things, in both directions

One product marketed on private markets barely holds any. It is 6.21% level 3, 37.07% Treasuries and agency paper, and 0.00% restricted securities across 325 holdings — its ten largest positions are government bonds and a money-market fund.

At the other extreme, one interval fund is 97.13% level 3 or marked at another fund's net asset value — and holds $1.29bn of four vehicles that were themselves prorating in the same quarters. A fund whose own liquidity depends on funds that were gating.

"Can I hold it in my retirement account?"

And the object is narrower than the headline. The proposal concerns an asset-allocation fund that may include private assets — and its own worked example is 90% liquid.

What we cannot tell you

• Twelve months is a short window and the proration jump is recent. Two quarters back below 5% would change the reading — which is why it is the report's first marker.
• "Determinable" is our word. Not every offer's outcome can be established from filings.
• Level 3 is a disclosure category, not a judgement. 90% level 3 says the assets do not trade. It does not say the marks are wrong.
• The smoothing measurement compares a reported value with a share price, and a share price carries sentiment the underlying assets do not. It shows the two series behave differently, not that one is correct.
• A comparability trap we found in replication: one fund reports its expense ratio excluding interest expense, so its 3.30% is not the same measurement as another's 8.50%.
• The rule is a proposal. Nothing here forecasts what a final rule will say.

Verification. 3,145 automated checks across two research passes, none failing. 227 planted mutations across nine rounds, all caught. Twenty-six errors were found and corrected inside the packs — including a dropped duplicate holding worth $49.2m and a near-miss in which one fund's figures were almost attributed to another. Twenty-two holes were found in the packs' own checkers. The most useful became rule 10 of the house standard: finding a number in a source is not verifying it.

Access. A government full-text search was CAPTCHA-gated and a market-data site used a proof-of-work check. Neither was bypassed. The printing office's own service answered the first, and the second was routed around onto the issuers' own filed tables — better sourcing than the blocked route would have given.

Disclosures. The author holds no position in any fund or security named here. No issuer payment, no advertising. Impersonal research, identical for every purchaser, tailored to no one. No statement about what any reader should hold, no forecast. Finnacl Research is not an investment adviser.

The full report adds the eleven-vehicle snapshot series ordered by fill rate, the complete gate table, the fee-stack chapter, the adviser chapter and the evidence appendix.