Valuation diligence on illiquid and level-3 positions

A liquid book marks itself. Everything else is an assertion. Level 3 asset valuation — the fair value hierarchy’s term for positions priced on unobservable inputs — is where a manager’s judgment and compensation occupy the same cell of the spreadsheet. The marks set reported performance and, often, the fee base. Diligence here rarely means re-deriving a number; what an allocator can test is the process behind it, and the gap between paper and practice.

And it has scaled: roughly $18 trillion in private fund assets across SEC-registered advisers per late-2021 Form ADV data, much of it in positions that never print a quote between subscription and exit.

What the frameworks require

Both fair-value frameworks — ASC 820 in U.S. GAAP, IFRS 13 internationally — tier assets by the observability of their inputs, and level 3 is the floor: measurements built on unobservable inputs where there is little, if any, market activity at the measurement date. What must accompany them depends on the framework. IFRS 13 still requires the full set — a rollforward reconciling opening to closing level 3 balances, quantitative detail on significant unobservable inputs, and a description of the valuation process itself. U.S. GAAP no longer does: ASU 2018-13, effective for fiscal years beginning after December 15, 2019, removed the valuation-process description entirely and lets nonpublic entities — private funds among them — replace the full rollforward with transfers into and out of level 3 plus purchases and issues.

That divergence reshapes the paper trail. A manager reporting under IFRS who cannot produce a rollforward, an input table, or a written account of who marks what is failing the framework their own audited financial statements claim to follow. A U.S. GAAP manager may be violating no disclosure rule at all — which is precisely why allocators still ask. Those artifacts are the ordinary output of a functioning valuation process, and “the standard no longer requires it” answers a different question than “can you produce it.”

The finding is the gap, not the mark

When SEC examiners write up private fund valuation, they rarely argue with a discount rate. The June 2020 OCIE risk alert on private fund advisers reported advisers “that did not value client assets in accordance with their valuation processes or in accordance with disclosures to clients” — in some cases overcharging management fees and carried interest computed on overvalued holdings.

That is the policy-versus-practice test; it requires no modeling skill. Most weak processes do not fail because the model is wrong; they fail because the documents describe a process the firm does not perform.

The Division of Examinations’ January 2022 follow-up extended the same logic to write-downs. Examiners found advisers that failed to reduce cost basis in the management-fee calculation after selling, writing off, or writing down part of an investment — and others leaning on LPA terms like “impaired,” “permanently impaired,” or “written down” that were never defined or consistently applied. On an illiquid book, valuation diligence is fee diligence. Every undefined impairment term is a degree of freedom in the fee base.

A pricing vendor is not independence

The standard DDQ asks whether the manager uses a third-party pricing service. Infinity Q is the case study in why the checkbox is not the control. In February 2022, the SEC charged the founder and former chief investment officer of Infinity Q Capital Management with overvaluing the assets of a mutual fund and a private fund by more than $1 billion between at least 2017 and February 2021 — “by altering inputs and manipulating the code of a third-party pricing service” — while collecting more than $26 million in profit distributions. The vendor existed. Independence did not. The diligence question is not whether a pricing service appears on the service-provider page but who holds write access to the inputs, models, and overrides between the portfolio and the published NAV.

Regulators converge on the same point. Cayman’s CIMA rules on the calculation of asset values, finalized in July 2020, require a NAV policy ensuring the NAV is “fair, complete, neutral and free from material error and is verifiable” — the registered private funds rule adds “reliable” — direct that a mutual fund’s NAV be calculated by a competent service provider independent of the manager, failing which CIMA may require auditor or independent third-party verification, and leave ultimate responsibility with the operators, who must approve the policy and any pricing models at least annually. In the U.S., Rule 2a-5 — adopted December 2020, the SEC’s first comprehensive treatment of fund valuation in over fifty years — requires a registered fund’s board or valuation designee to assess valuation risks, test methodologies, and oversee pricing services. It binds registered funds, not private ones, but allocators increasingly expect private managers to approximate it. For private funds, the custody rule’s audit provision — annual audit by a PCAOB-registered and -inspected accountant, GAAP financial statements to investors within 120 days of fiscal year end — and the administrator’s NAV work supply the main recurring external contact with the marks.

Questions that expose a weak process

Five questions, each cheap to ask and expensive to fake.

Does practice match the paper? Read the valuation policy against the last four quarters of valuation memos and the PPM’s description; discrepancies are the deficiency examiners cite.

Who can change a mark? Map write access across the pipeline — vendor inputs, model code, manual overrides — and ask for the override log. If the people who can touch the number are paid on the number, ask what compensating control intervenes.

Can the model reproduce the entry price? The IPEV valuation guidelines (December 2022 edition; updated December 2025) center on calibration: at entry, assumptions are adjusted until the model reproduces the price paid, then rolled forward. Run it on a sample position. A model that cannot reproduce the manager’s own transaction is decorated with inputs, not disciplined by them.

How is “impaired” defined, and what happens to the fee base when it applies? Expect a definition, a decision-maker, and a worked example. Undefined terms are where write-down discipline goes to die.

Who outside the firm sees the marks, and when? The administrator’s role — calculating NAV or merely receiving it — the auditor’s level 3 procedures, and whether investors receive a level 3 rollforward — an IFRS 13 requirement, an allocator expectation everywhere.

Strong processes give boring answers

None of this requires forensic skill — that is the point. A sound valuation process produces dull artifacts on demand: a policy that matches the memos, a short and explained override log, a calibration that ties to entry, definitions that read the same in the LPA and the fee calculation. Weak ones fail on the first document request — not because anyone proves a mark wrong, but because the gap between disclosed and practiced process is itself the finding.

SetOne Labs pressure-tests valuation processes the way an examiner reads them — policy against practice, access against assertion — for managers facing institutional diligence and allocators conducting it. To discuss a valuation review in confidence, begin a conversation.

SetOne Labs provides advisory services and general information. Nothing here is legal, tax, or investment advice.