Private credit valuation, explained.

Why a credit fund must price loans that never trade, every quarter

The life of one loan's quarterly mark, in five steps

1. The fund lendsVarrowmere Capital Corp. takes $186 million of a unitranche loan to Wrenhollow Dental Partners in March 2024, bought at 98.0.
2. The borrower reportsMonthly packages and a quarterly compliance certificate. Amendment No. 3 in August resets a covenant for extra interest paid in kind.
3. The quarter's mark is setFinancials re-read, covenants re-computed, the yield and coverage tested, the outside firm's range compared. This is where 98.50 becomes 95.40, with the reasons written down.
4. The committee approves, the board overseesSeven loans go to the October 22 committee as exceptions; the board's audit committee receives the quarterly report on November 4.
5. Investors trade at the resultThe 10-Q, the quarterly report to the SEC, is filed with every loan's fair value, investors redeem at net asset value, and the auditor tests a sample of 25 marks.

Step 3 is the subject of this guide. The fund, borrower, people and figures in this example are fictional.

Model-valued loans, and well-covered loans worth less than par

Fair value, Level 3, valuation designee and yield analysis

Fourteen terms are defined in the one-page summary.

Fair value under ASC 820: the price a buyer would pay today, worked out without a market

In shortUS accounting defines the fair value of a private loan as an exit price: what the fund would receive selling the loan to another market participant on the valuation date. When no price can be seen, the fund estimates one from its own assumptions about what such a buyer would assume, including a charge for risk. That makes the value Level 3, and Level 3 comes with heavy disclosure.

The rule is Accounting Standards Codification Topic 820, Fair Value Measurement. It defines fair value as "the price that would be received to sell an asset or paid to transfer a liability in an orderly transaction between market participants at the measurement date".7

Three phrases carry the weight. Received to sell means an exit price: what a buyer would pay today, whatever the fund paid for the loan or expects to collect by holding it to the end. Orderly transaction rules out a fire sale. Market participants means the buyer is a typical informed investor in such loans, with that investor's view of risk.

Three tiers of evidence, and why private loans sit in the last one

ASC 820 ranks the inputs to a valuation. The first tier is a quoted price for the identical asset in an active market, such as a listed share. The second, Level 2, uses other observable inputs: a quote for a similar asset, or a dealer price in a thin market.

Level 3 inputs are "unobservable inputs for the asset or liability", used where observable ones are not available.7 A unitranche loan to a dental group owned by a private equity firm has no quote, and the inputs that matter most, the return a buyer would demand on this loan and the multiple at which this business would sell, cannot be read off a screen. So the loan is Level 3.

Level 3 keeps the same bar. The standard says the objective "remains the same, that is, an exit price at the measurement date from the perspective of a market participant", and that unobservable inputs "shall reflect the reporting entity's own assumptions about the assumptions that market participants would use… (including assumptions about risk)". A model that leaves out a risk adjustment buyers would make "would not represent a fair value measurement".7 In plain words: the fund may use its own estimates, but they must be estimates of what a buyer would think.

Calibration: start from the price someone actually paid

When a fund buys a loan at a price it judges to be fair value, and will later value the loan with a model that uses unobservable inputs, ASC 820 says the model "shall be calibrated so that… the result of the valuation technique equals the transaction price".7 For a loan bought at 98.0, the fund finds the yield at which its model returns 98.0, then moves that yield each quarter as market spreads and the borrower's risk change. Calibration ties every later mark to a real trade.

The 2025 edition of the IPEV Valuation Guidelines (the International Private Equity and Venture Capital Valuation Guidelines), used widely outside the US, makes calibration at entry and at each valuation date a central step, and treats assumptions that never change from quarter to quarter as the exception.8

What a fund has to disclose about its Level 3 marks

Since 2011 a fund must publish, for its Level 3 assets, "quantitative information about the significant unobservable inputs", a description of its valuation processes, and a narrative of how sensitive the values are to those inputs.7

In a BDC's annual report that becomes a table: each class of investment, its fair value, the technique, the input and its range. One BDC's table for September 30, 2025 shows $2.04 billion of "one stop" loans valued by yield analysis at market interest rates from 3.8% to 20.5%, weighted average 8.6%, cross-checked against EBITDA multiples from 8.0x to 33.0x. Its auditor named the Level 3 valuations a critical audit matter, because the inputs "are highly judgmental, are sensitive to economic dislocations".3

Rule 2a-5 puts the mark on the board and its designee, and examiners turned to private credit in 2025 and 2026

In shortThe Investment Company Act puts fair value on the fund's board. SEC Rule 2a-5, required since September 8, 2022, says what "in good faith" requires and lets the board appoint the adviser as valuation designee. Rule 31a-4 keeps the evidence six years. In 2025 and 2026 examiners named private credit and BDC loans as priorities, the SEC brought a loan-pricing case, IPEV and IOSCO updated their guidance, and new proposals would open more retirement money to these funds.

Section 2(a)(41) of the Investment Company Act of 1940 says a fund values securities at market value when quotations are "readily available" and otherwise at "fair value… as determined in good faith" by its board. The rule applies to registered funds and to BDCs.9 For decades the SEC's guidance on what that meant dated from 1969 and 1970. On December 3, 2020 the SEC adopted Rule 2a-5 and Rule 31a-4 and withdrew the old releases. The rules took effect on March 8, 2021, with an 18-month transition, so compliance has been required since September 8, 2022.10

What Rule 2a-5 requires

Determining fair value in good faith means four things, done for every investment without a readily available quotation:1,9

1. Assess and manage valuation risk, including conflicts of interest, periodically.
2. Select and apply methods consistently, review them, and watch for events that call for a new value.
3. Test the methods for "appropriateness and accuracy", with the tests and their minimum frequency written down.
4. Oversee pricing services, where used, including how prices are challenged.

The board may do this itself or designate a valuation designee: the fund's adviser, never a sub-adviser, or for an internally managed fund, its officers.1 The designee must report to the board at least quarterly on material valuation matters and changes, at least annually on whether the process is adequate and effective, with the results of its testing and the resources it has, and in writing within the time the board sets, "in no event later than five business days", when something materially affects fair value.9

The designee must also "reasonably segregate" fair value determinations from portfolio management, so the people whose performance rides on a loan do not decide its mark.1 A quotation counts as readily available only if it is an unadjusted price in an active market for the identical investment, which no private loan has.1

Rule 31a-4 then keeps the evidence: "appropriate documentation to support fair value determinations" for at least six years, the first two easily accessible; the reports to the board for six years after the end of the fiscal year in which they were given; and the list of investments the designee values, until six years after the fiscal year in which the designation ends.6

The SEC dropped a proposed requirement to keep the specific methodologies, assumptions and inputs behind every determination in all cases, so what counts as appropriate support is the fund's call, and the methodologies and inputs are what an examiner or auditor will usually ask for.10

The SEC's own count of what this costs

In August 2026 the SEC renewed both rules' paperwork approvals. Its staff estimate 10,047 funds subject to Rule 2a-5, each spending 34 hours a year on the reporting, a time cost of $33,422 per fund and $335,790,834 in all, plus $3,632 of outside cost per fund; Rule 31a-4 records add 36 hours and $15,984 per fund.9,11

These are averages across every registered fund and BDC that must determine fair value, and fewer than 200 of the 10,047 are BDCs (LSTA counts 179).12 A credit fund that values 142 loans by model every quarter spends far more, and no public figure measures how much.

What examiners said in 2025 and 2026

The SEC's Division of Examinations released its fiscal 2026 priorities on November 17, 2025.13 They list "alternative investments (e.g., private credit…)" among the products examiners will focus on, valuation among the core areas of every adviser's compliance program, and valuation among the topics for advisers new to private funds.14

The fiscal 2027 priorities, released October 23, 2026, go further. Examiners will review whether advisers "assessed the reasonableness of their valuation processes and methodologies", will prioritize advisers that recommend illiquid products "such as loans provided by private funds or business development companies (BDCs)", and will watch registered funds holding "private credit, and the associated issues regarding valuation, redemptions, and disclosures of conflicts of interest".15

Standards beyond the SEC

• AICPA Accounting and Valuation Guide (2019): the US practitioners' reference for portfolio companies, with guidance on valuing debt instruments, a chapter on calibration and one on back-testing, comparing past estimates with what was later realized.16
• IPEV Valuation Guidelines, 2025 edition: published December 2025, in effect for quarterly periods beginning on or after April 1, 2026. Stronger documentation, calibration at each date, more frequent valuation for evergreen funds (funds with no end date that take in and pay out money as they go), and a statement that automated models and AI do not replace professional judgment.8 In a 2025 UK review, 70% of the firms surveyed said they follow IPEV.17
• IOSCO: final report on valuing collective investment schemes, June 1, 2026, which updates and brings together its earlier, separate valuation principles for funds and for hedge funds, citing the rise of funds holding less liquid, harder-to-value assets, private assets among them, and of retail investors in them.18
• FCA: a multi-firm review of private market valuation, March 5, 2025, covering about £3 trillion of assets, on conflicts, committee independence, documentation, back-testing and triggers for unscheduled valuations.17

Retirement money is coming, and valuation is part of the test

Executive Order 14330 of August 2025 directed agencies to make room for alternative assets, private credit among them, in 401(k) plans.19 The Labor Department's proposal of March 31, 2026 names valuation as one of six factors a plan fiduciary weighs: the investment option must be "capable of being timely and accurately valued in accordance with the needs of the plan".20

The SEC's proposal of October 5, 2026 on interval funds (funds that offer to buy back part of their shares at set intervals) says the policies of funds holding private assets might include, as relevant, the valuation of illiquid and hard-to-value securities, including fair value methodologies and the use of third-party valuation providers; comments are due December 4, 2026.21 Neither proposal was final at the end of October 2026.

How yield analysis prices a performing loan, worked through on one dental loan

In shortFind the return a buyer would demand on this loan today, then discount the loan's interest and principal at that return. Start from the yield that matched the last approved mark, add the change in market spreads and the change in the borrower's own risk, and let a floating base rate pass through the coupon. The Wrenhollow loan goes from 10.55% to 12.30%, and from 98.50 to 95.40.

For a loan that is paying as agreed, funds and their valuation firms mostly use a yield analysis. The Financial Stability Board, after engaging with market participants, describes it as "the primary approach for performing loans": projected cash flows discounted at a specified rate, with public loan and bond markets as benchmarks, calibrated at origination and adjusted for credit quality, performance and markets since.23 One BDC's annual report states the rule it applies: for debt investments "not determined to be credit impaired", a "market interest rate yield analysis".3

Step 1: the yield at the last mark

The fund bought into the Wrenhollow loan at 98.0 in March 2024 and calibrated its model to that price. At June 30, 2026 the committee approved 98.50, which on the model corresponds to a discount yield of 10.55%. That is the starting point. Starting from the last approved yield keeps every quarter's mark anchored to the original trade and to every decision the committee has made since.

Step 2: move the yield for the market and for the borrower

A floating-rate loan pays a base rate that resets with the market plus a fixed margin, the spread. Spreads are quoted in basis points: 100 basis points is one percentage point. Each driver below moves the required yield, and the table shows by how much.

The example's base rates (4.15% falling to 3.95%) are the fictional fund's sample inputs. In the real market, overnight SOFR (the Secured Overnight Financing Rate) rose over the quarter, from 3.68% on June 30 to 3.90% on September 30, 2026, after the Federal Reserve raised its target range to 3.75%–4.00%, effective September 17.24 With a floating coupon the direction matters little to the price: the base rate moves the coupon and the required yield together.

Step 3: from yield to price

The full model discounts each future payment, the cash interest, the interest paid in kind added to principal, and the principal at the expected repayment date, at 12.30%. A shortcut shows what drives it. For a small change in yield, price moves by the difference between the required yield and the coupon, times the loan's duration (its price sensitivity, in years):

The duration of 2.19 comes from an expected life of 2.5 years. The loan matures on March 15, 2030, but loans to companies owned by private equity firms are usually repaid early, when the company is sold or refinanced, so the fund assumes repayment in 2029. Expected life matters: the same 12.30% gives 96.24 on a two-year life and 94.60 on a three-year life. The shortcut is a check on the model; the model itself prices the actual cash flows.

Each yield change times the duration gives its share of the move: 0.35% × 2.19 is 0.77 points down for spreads; 1.60% × 2.19 is 3.50 points down for credit; the extra 0.50% of interest paid in kind raises the coupon and adds 1.10 points; a quarter less time to repayment adds 0.07. The base-rate change moves coupon and yield together and adds nothing. 98.50 − 0.77 − 3.50 + 1.10 + 0.07 = 95.40. On $186.16 million of principal, fair value is $177.60 million, against $183.21 million at June 30 on $186.00 million.

Two limits every yield analysis needs

A cap at the repayment price. A borrower can usually repay at par, or at a small premium during the first years. No buyer pays more than the price at which the borrower can take the loan away, so funds cap marks there. In the example, an aerospace parts loan whose model price is 101.25 is marked at its 101.00 call price.

A floor on the method. A yield analysis assumes the loan will be repaid. When the borrower's value no longer comfortably covers its debt, that assumption fails and the fund switches to a recovery analysis. The next section explains the test that decides.

The enterprise value coverage test chooses the method; it cannot set the price

In shortEstimate what the borrower's business is worth, usually a market multiple times EBITDA, and compare it with the debt that ranks ahead of and alongside the loan. Comfortable coverage means the loan is priced by yield analysis. Thin coverage means it is priced by what lenders would recover. Coverage tells the fund whether it is likely to be repaid; what a buyer would pay is a second question, and the SEC has sanctioned a BDC for treating coverage as the mark.

The test, on the example

Dental support organizations have recently sold for about 11.0 times EBITDA in the fund's set of nine 2024–2026 transactions (12.0x a year earlier). On reported EBITDA of $55.8 million that values Wrenhollow at $613.8 million. Net debt across all lenders is $337.8 million. Coverage is 613.8 ÷ 337.8 = 1.82x; put the other way, the debt is 55% of the business's value. The fund's policy moves a loan to a recovery analysis below 1.25x.

A downside case, EBITDA a further 10% lower at $50.2 million and a 10.0x multiple, gives $502.2 million of value and 1.49x coverage, a loan-to-value of 67%. The loan stays covered, so the yield analysis holds.

When coverage fails: a recovery analysis

Below the threshold, the fund values the loan as a claim on the business. It estimates what the business would sell or reorganize for, pays out the claims in order of rank (a "waterfall"), deducts the costs and time a restructuring takes, and discounts the loan's share. In the example, a freight company whose EBITDA fell 31% has enterprise value of only 0.94 times its debt; the fund moves it to a recovery analysis, proposes non-accrual (no longer booking the unpaid interest as income), and marks it at 79.50, down from 88.00.

One BDC describes the same split: an enterprise value analysis is run "to determine if debt investments are credit impaired"; if they are, it uses "the enterprise value analysis or a liquidation basis analysis"; if not, a yield analysis.3

KCAP, 2012: the case that separated coverage from the mark

In November 2012 the SEC settled charges with KCAP Financial, a BDC, and three of its executives. For its illiquid debt KCAP had used an enterprise value method: if the issuer's value exceeded the debt senior to and level with KCAP's, it "used the cost of the debt security as fair value". The SEC's order says the method "did not, however, calculate or inform KCAP – or the public – what the exit price was for that security".4

At December 31, 2008, KCAP carried a Ford Motor Credit debt instrument at 70% of par while it traded at 41% to 45%; it later restated it to about 41%. Its restatement found net asset value overstated by about 27% at that date. Two executives paid $50,000 penalties and one paid $25,000.4

The lesson stands in 2026. Direct lenders specialize in loans secured by the value of a going business: researchers at the NBER find their edge is in "enterprise-value based loans to private equity-backed firms", with claims built on "the continuation value of firms after default".25 That makes the coverage test the right question about repayment. The mark answers a second question: what return would a buyer demand to own this loan today? A loan can be well covered and still worth 95.

EBITDA add-backs, interest paid in kind and stale financials: the inputs that flatter a mark

In shortA mark is only as good as the earnings and dates beneath it. Covenant definitions of EBITDA allow add-backs that lenders and buyers would not pay for; interest paid in kind raises the coupon and often signals strain; and financials older than a quarter describe a business that may have changed. Each one pushes a mark up if nobody checks it.

Add-backs: the covenant's EBITDA and the buyer's EBITDA

Credit agreements define EBITDA for covenants, and the definitions often let the borrower add back costs and earnings that never appear in its accounts: savings expected from a merger, one-off costs, or "run-rate" earnings from locations not yet open.26

Researchers at the Federal Reserve Bank of St. Louis read the EBITDA definitions in syndicated loan agreements. The most common number of non-GAAP add-backs was two, and about 43% of definitions carried three or more. Each extra add-back raised the chance of a loan becoming 60 days delinquent within three years by 4.2 percentage points, against a base rate of 1.3%, and the chance of default by 1.6 points, against 1.1%. They also cite an S&P study in which projected adjusted EBITDA at the deal exceeded what borrowers actually earned over the next two years by about 30%.26

The study used the 2013 interagency leveraged lending guidance to explain the rise in add-backs. The OCC and FDIC withdrew that guidance on December 5, 2025, saying it was "overly restrictive" and pushed lending to nonbanks; the Federal Reserve's version is still posted.27

The example shows why a valuation policy should decide the add-back question once. Wrenhollow's certificate adds back $6.1 million for 14 offices, 9 not open. The credit agreement allows the add-back for the covenant, capped at 15% of EBITDA, so the covenant is met at 5.46x. A buyer of the loan would not pay for offices that do not exist yet, so the fund values on the reported $55.8 million and 6.05x. The cushion to the covenant falls from 19% as certified to 10% on reported earnings, and to 3% against the 6.25x covenant before the August amendment.

Interest paid in kind: a higher coupon that often signals strain

Paid-in-kind (PIK) interest is added to the loan's principal and paid when the loan is repaid. It raises the loan's stated return, which helps the mark, and it also often means the borrower is short of cash: Wrenhollow's lenders agreed to a looser covenant in exchange for 0.50% of extra interest paid in kind. LSTA reports PIK in 4.33% of new BDC investments in the first quarter of 2026, down from 9.47% a year earlier.12 When PIK appears by amendment, the mark should carry both effects: the higher coupon, and the credit risk that made the amendment necessary.

When financials and marks go stale

A mark built on June figures in October describes a business four months out of date. Policies set a limit (90 days is common) after which inputs count as stale and the loan gets a closer look. The same applies to the mark itself: a loan held at the same price for three quarters while comparable spreads widened 0.45% needs a written reason.

Researchers find positive serial correlation in BDC fair value changes, "evidence of delayed updating": a mark that moves this quarter tends to keep moving next quarter.28 Section 10 shows why that matters most in funds investors can leave each quarter.

What independent valuation firms check, and why the committee sits apart from the deal team

In shortMost BDCs hire an independent valuation firm to value or review part of the book each quarter, so every loan is seen at least once a year. The firm gives a range or a positive assurance that the fund's mark is reasonable. The fund still decides; the outside view tests it. Research finds the reviews work in normal times and weaken in stress, and regulators look hard at who sits on the committee.

Third-party valuation and positive assurance reviews

A third-party valuation firm either values a loan itself or reviews the fund's preliminary value and says whether it is reasonable, usually with a range. One BDC's annual report describes the second model. When no outside firm values the whole portfolio, one is engaged "to provide positive assurance of the portfolio each quarter (such that each investment is reviewed by a third-party valuation firm at least once on a rolling 12-month basis and each watch-list investment will be reviewed each quarter)".29 The watch list is the fund's list of loans under closer monitoring.

Another BDC has outside firms review about 50% of its Level 3 value each quarter, each investment at least every other quarter.3 In both, the adviser's valuation committee determines the fair value.

Do outside reviews make marks better?

Young Soo Jang and Ginha Kim used data from a third-party valuation firm to compare BDCs' reported marks with the firm's recommended ranges. At least 76% of BDCs used third-party appraisal in 2022, and 58% were required to by their own lenders. In normal times deviations were "infrequent and small": 4.0% of reported fair values sat above the recommended range by more than one point of par.28

During COVID-19, nearly 5% of loans were marked more than 5 points above the range, mostly by funds that were not the lead lender, and lead lenders received higher appraisals even on the same loan, which the authors trace to information from renegotiations.28 The FSB heard the same from market participants: they pointed to independent valuation firms and oversight committees as the way to address subjectivity, while some FSB members cited comparable cases valued differently across managers.23

The committee, and who should not be on it

Rule 2a-5 requires the designee to keep fair value decisions reasonably apart from portfolio management.1 The UK regulator's 2025 review shows what goes wrong in practice: in some firms the valuation function was mainly administrative while investment professionals shaped inputs and assumptions; some committees had senior investment professionals in all or most voting seats; minutes did not record how decisions were reached; and reasons for changing key assumptions such as discount rates were vague.17

Good practice included independent voting members, inviting auditors to committee meetings, comparing two providers' values for the same asset each year, and a "value bridge" showing how much of each change came from earnings, multiples or markets.17

Marks on stressed loans, and how close BDC marks come to what lenders recover

In shortIn aggregate BDC marks sit close to cost, and most of the gap on discounted loans comes from the price paid at entry. Stress shows at the edges: loans near maturity, loans just below par, and the minority that move from near par to a large loss in a quarter or two. No public study follows the same loans from their last performing mark to final recovery across the market, so back-testing your own book is the only way to know how good your marks are.

BDC fair value near cost while PIK rose

Researchers at the Federal Reserve Bank of Boston studied BDCs' quarterly schedules of investments. The ratio of fair value to cost "remained relatively stable near 1.0" even as PIK and other risk signs rose, and because most middle-market loans have no liquid market, "these valuations rely substantially on the BDCs own models and the judgment of their management". The stock market reads the marks: a one-standard-deviation fall in a BDC's fair value ratio went with about 0.5% lower abnormal returns on its shares the next quarter.30

Moody's analysts looked at the same filings from two angles. Across 6,387 positions in 34 listed BDCs, they found that for loans bought below 98, about 130 of the 172 basis points of discount to par came from the entry price and only about 42 from movement since, so a below-par mark is often what the fund paid.31

Across 6,570 loans in 32 listed BDCs at March 31, 2026, loans due in 2026 carried an average mark of 88.8% of cost, against 99.7% for loans due in 2030 or later; 11% of all loans were below 90% of cost, and 52% were above par.32 PIMCO compared the 25th percentile of BDC marks, 97.51, with 94.95 for broadly syndicated loans at the same date.33

When marks met reality: a 19% NAV cut traced to six companies

The cases where marks met reality are documented one at a time. KCAP's 2010 restatement, in section 05, is one.4

KBRA followed 79 middle-market borrowers it had flagged as at risk in 2024: 56 deteriorated further in 2025, and 25 restructured or missed a payment.34

What lenders recover sets the floor beneath a stressed mark. Moody's historical data, as cited by Dechert, show senior secured loans "consistently recovered 60% or more" of face value in default, with first-lien bank loans recovering 53.6% in the worst conditions of 2008. The same article cites an estimate that realized losses on all BDCs' loans were 0.70% in 2025.35 Those are averages over many defaults; a single loan can recover far less.

What SEC cases and exam priorities show about how loan valuations fail

In shortFour public matters cover the main ways a loan valuation fails: a method that never asks what a buyer would pay (2012), falsified inputs (2019), a pricing formula left running when markets moved (2026), and the fee conflict built into valuing your own book (alleged in a 2026 lawsuit). Each points to a control that would have caught it.

Of the four, the 2026 order is the one to read in full, because nothing in it involves fraud. Madison Capital Funding originated senior loans and sold portions to funds it advised after 30 to 60 days, at "fair value" or "fair market value" according to the funds' documents. In practice it sold at par less the unamortized loan fee. Between March and May 2020, as credit spreads "widened substantially" and loan prices fell, it executed 143 such sales "without any market adjustments" and certified to the funds' independent review party each time that the price was fair market value.5

It checked only that each loan's internal rating was still B or better. All but one loan later performed or was repaid. The SEC found violations of Sections 206(2) and 206(4) and Rule 206(4)-8 without needing intent, noted that the adviser had already repaid the funds $5,010,854.90 plus $203,819.69 of interest after a 2021 examination, and imposed a $900,000 penalty.5

BDCs, semi-liquid funds and redemptions: why the mark carries more weight in 2026

In shortPrivate credit is about $1.4 trillion of US company debt. More of it now sits in funds that let individual investors leave every quarter at net asset value, and in early 2026, for the first time, accepted redemptions from perpetual BDCs exceeded new money. Marks fell at most BDCs, non-accruals rose and defaults reached a record on Fitch's measure. Each mark now prices a real exchange between investors.

The Federal Reserve's May 2026 Financial Stability Report puts private credit at about $1.4 trillion, 10% of the debt of US nonfinancial companies and about a third of their below-investment-grade debt outside banks. Perpetual BDCs held $306 billion of gross assets ($161 billion net) and credit interval funds $119 billion ($80 billion); together these semi-liquid funds were about 20% of net assets in private credit vehicles.2

Most perpetual BDCs intend to cap redemptions at 5% of net asset value a quarter. In the first quarter of 2026, after investor sentiment turned on defaults and worries about AI and software borrowers, "accepted redemption requests somewhat exceeded new inflows… the first time this has occurred since these vehicles were created".2

LSTA's quarterly review of BDCs counts 179 funds with $575 billion of gross assets in the first quarter of 2026, up 21% in a year. Nine large perpetual BDCs received $13.4 billion of redemption requests in the quarter and paid $7.8 billion, while taking in $5.6 billion of new subscriptions. 77% of BDCs reported a lower weighted average mark on their term loans, up from 67% in the fourth quarter of 2025; on average 7.6% of a BDC's portfolio was marked below 85; and the weighted average non-accrual rate rose to 1.99%.12

Fitch's private credit default rate reached 6.3% for the twelve months to August 2026, a record on its measure, with health care and industrial borrowers at 9.9% each.36 When a fund's investors can leave each quarter, a mark that is too high pays those who leave at the expense of those who stay, and a mark that is too low does the reverse. The FSB warns that "perceived or actual stale valuations may create a first-mover incentive during stress events".23

What software and AI can take on in a valuation, and what stays with the committee

In shortMost of a quarter's hours go to reading, re-keying, re-computing and writing up, which software can do and cite. Choosing methods and inputs, approving marks and overriding with reasons stay with people, because the rule puts the determination on the board or its designee. Whatever does the work must be tested, and its output kept six years.

Broken into steps, a quarter's work falls into three kinds. Reading monthly packages and compliance certificates, re-computing leverage and covenants by each agreement's definitions, collecting comparable spreads and other funds' public marks with their dates, running the yield model and the coverage test, comparing the result with the outside range, drafting the memo and the bridge, and filing the evidence are reading, arithmetic and writing. Deciding whether an add-back belongs in valuation EBITDA, whether a loan has moved from performing to impaired, which comparables are comparable, and what to propose are judgments. Approving the mark is a decision.

What the rules and guidance expect of automation

• Rule 2a-5 still applies: methods must be tested for "appropriateness and accuracy", and any pricing source overseen. A model or a tool that drafts marks is part of the method.1
• IPEV 2025: automated models and AI are no substitute for professional judgment.8
• The SEC's FY2027 priorities say registrants increasingly use generative tools (the document calls them "super intelligence") for research, drafting and portfolio management, and that examiners want to understand "how the technologies influence outcomes and whether firms continue to comply with existing, technology-neutral obligations".15
• Records: the documentation supporting each mark is a Rule 31a-4 record, whoever or whatever prepared it.6

The limits are practical. Software can read a certificate and flag an add-back; it cannot decide the fund's policy on add-backs. It can find the same loan in three other funds' schedules; it cannot know what the lead lender learned in a renegotiation. It can draft a reason; a person has to agree with it and sign.

How valuation teams run a quarter today: people, hand-offs and the weeks between close and committee

In shortA valuation team of a few people re-reads every borrower's figures, rebuilds a model per loan in linked workbooks, collects market evidence by hand, reconciles the outside firm's ranges, and writes memos and a committee pack, all in the three weeks between quarter end and the committee. Most of the time goes to collecting and re-keying.

The valuation head, analysts, outside firm and committee behind a mark

Collecting and re-keying: about half of a quarter's valuation days

No regulator or study publishes the hours a quarter's valuation takes. A vendor reports one private credit manager's preparation for its quarterly portfolio review falling "from two to three weeks to two to three days" after adopting the vendor's software, which gives a sense of how much of the work is collection.39 The split below is an estimate for a book of about 140 loans, to replace with your own.

Workbooks, an outside firm, positive-assurance rotation or valuation software

The Wrenhollow dental loan marked from 98.50 to 95.40 on new inputs, then proposed as one of seven committee exceptions

In shortThis is the example from the top of the page, inside the product. Over the weekend the agents proposed marks for 141 of the 142 loans. On Monday Elena Marchetti, head of portfolio valuation, runs the last one on its new inputs: six agents read the package and the market, the loan moves from 98.50 to 95.40, and every number in the evidence links to its document. She adds a downside case and proposes the mark, which goes to the committee as one of seven exceptions.

1

The whole book against its peers

Elena opens the home screen on Monday, October 12: every loan in the fund drawn at its net leverage (across) and the spread over SOFR a buyer would demand (up), sized by principal. The rings are June's approved marks; the dots are where this quarter's evidence puts them. Loans priced on recovery sit in the red band at the top.

One ring waits with "new inputs · run": Wrenhollow Dental, still at 98.50. Its September package and compliance certificate arrived on Friday at 6:40 pm, after the weekend run.

2

Six agents on the new inputs

She presses "Run valuation on the new inputs". The rail lists each agent as it works. The borrower financials reader finds EBITDA for the last twelve months of $55.8 million, down 13.1%, net leverage 6.05x, and the certificate's $61.9 million that adds back $6.1 million for 14 offices not yet open. The credit event watcher finds Amendment No. 3 of August 14: the leverage covenant reset from 6.25x to 6.75x for four quarters, for 0.50% of extra interest paid in kind and a 0.25% fee.

On the map, the two quoted dental loans widen as the market analyst prices them.

3

The dot leaves its ring, with the evidence beside it

The dot moves to 6.05x and 835 basis points, and the evidence mark settles at 95.40: a 12.30% yield, 3.10 points down. The rail lists why, in four groups, 16 sources each linked: the same loan in three other funds' June schedules (98.75, 97.90, 98.25), two quoted dental loans that widened 52 and 44 basis points, a new oral surgery deal priced on September 18 at SOFR + 6.25%, the dealer survey's +0.35% for single-B-rated healthcare loans, the six comparable loans in the book (median 95.93), and the fund's leverage grid adding 1.60% for 6.0x to 6.5x.

4

A proposed mark, and three reasons it goes to the committee

The position opens with the run that produced it, finished in 1 minute 11 seconds across 9 documents and 3 data feeds, and the reasons it is an exception: a move of more than the 2.00 points the policy allows without the committee, an amendment with interest paid in kind, and leverage up 0.88x in one quarter. A fourth line records the finding: the add-back was left out of valuation EBITDA, with the certificate's schedule one click away.

Owen Tran, the valuation analyst, reviews the inputs against the package: "Inputs tie to the package".

5

A downside case, then a signature with its meaning

Before she signs, Elena adds a downside case to the memo: EBITDA a further 10% lower at a 10.0x multiple still covers net debt 1.49 times, so the yield method holds; with another 0.50% of credit premium the price would be 94.31, below the outside firm's range. Then she proposes 95.40. The dialog records the meaning of her signature, "Reviewed — propose to the committee", her name and her password, and says what is written to the audit trail and kept for six years.

She could have overridden the mark instead; an override shows in the memo and to the committee with its reason.

The solution's design targets are a committee-ready pack within three days of the valuation date and documented support for 100% of marks before the committee. They are targets, to be measured on your own book.

Dealer quotes, the yield build, the coverage test, the fund bridge and the auditor's sample: the rest of the quarter's screens

In shortOne loan is the story. These screens show the rest: the quarter replayed, a dealer quote opened at its source, five quarters of performance with the add-back finding, the yield build and the coverage test, the mark beside its peers and its bridge, the memo and sign-off, the committee's seven exceptions, the fund's mark movement, the auditor's sample and the valuation policy as settings.

Select any screen to open it larger. The fund, borrowers, people, rates and figures are fictional.

118 + 17 + 7 = 142: each loan lands in exactly one group, so the committee's time goes to the seven that need a decision.

Where the valuation agents stop: no agent sets a mark, changes a method or decides a non-accrual

In shortEach agent has one job and limits it does not cross. None of them sets or approves a mark, changes a method, decides a non-accrual or edits the outside firm's range. The head of valuation proposes every mark with a signature; the valuation committee, chaired by the CFO, approves; the board receives the quarterly report.

Six valuation agents, from borrower financials to the draft memo

The valuation head proposes; the committee approves; the board oversees

The outside firm's range is shown, never edited. The auditor reads the evidence file read-only. Every agent step, signature, mark and memo version is written to the audit trail and kept six years, the Rule 31a-4 period.

Elena and Martin before the October 22 valuation committee

What the valuation file holds at quarter end: a cited memo per loan, an exceptions list, a board bridge and the auditor's evidence

In shortIt reads the documents and data a credit fund already receives, applies the fund's own valuation policy, and leaves behind a proposed mark for every loan with its memo, bridge and evidence, an exceptions list for the committee, a fund bridge for the board, and an evidence file for the auditor. The fund's systems of record stay where they are.

Administrator, market data and outside firm feeds for each mark

For each, check whether it has an API, a file feed, or only screens, and who owns it.

Sources, a versioned policy and named people behind each mark

What an unsupported or late mark costs a fund in hours, exam findings and investors' money

In shortStart with your own costs, because nobody publishes how long a quarter's marks take or how often they change at committee. Three published figures frame them: what an analyst hour costs, what the SEC itself counts for the Rule 2a-5 paperwork alone, and what one 2026 loan-pricing case cost an adviser.

Last quarter's marking time for ten loans, step by step

Value moved between investors, exam findings and restatements

Hours per loan per quarter and the NAV a small mark error moves: a calculator for your fund

In shortPut in your own numbers. The calculator starts from a published analyst wage and asks you for the rest: how many loans you mark, the hours each takes a quarter now and after, and what the tool and your testing cost. It also shows how much net asset value a small error across the book moves. Every default marked illustrative is an assumption to replace.

Linked inputs, a faster pack and fewer full outside revaluations

Your quarterly marking hours and mark-error exposure

Each figure is $80,000 ÷ (hours saved × 4 quarters × $71), rounded up. A fund marking 140 loans that saves two hours on each sits at the break-even line; one marking 40 loans should expect the case to rest mostly on evidence and speed.

Your valuation policy, inputs, outside firm and hours per loan, scored for readiness

In shortPick the answer closest to your team today. Eight questions score 0 to 2; the last is for reflection. The meter adds up your score as you go, and each question says why it matters and what a strong answer sounds like.

A valuation team's next step: write the policy or run a parallel quarter

Eighteen tests of how valuation software builds a yield, runs the coverage test and keeps Rule 2a-5 evidence

In shortAsk every vendor the same questions, us included. The answers show how a mark is built: what the tool reads, how it prices, what evidence you keep, whose policy applies, who decides, and how you would measure it on your own loans.

Marking one quarter in parallel with your own team, judged by your valuation committee

In shortWrite the policy down, measure a quarter, then run the next quarter in parallel on your own loans and compare. Decide with four numbers. None of these steps commits you to a vendor.

1. Write the policy down: methods by credit status, the recovery threshold, add-backs, the call-price cap, committee thresholds and the stale-input rule.
2. Measure last quarter for ten loans: hours by step, days to the pack, marks changed at committee.
3. Pick 30 to 50 loans, including last year's hardest: an amendment with PIK, a move to recovery, a mark outside the outside firm's range.
4. Run one quarter in parallel, with your people proposing and approving every mark from both processes.
5. Decide with four numbers: agreement with the approved marks, issues found that your process missed, false alarms, and team hours.

Proposed against approved marks, in points, loan by loan

Answers on Level 3 marks, Rule 2a-5 designees, yield analysis and stale private credit marks

In shortThe questions valuation teams, fund boards and investors search for most, each answered on its own.