Financial Reporting
One Unobservable Input Puts Your Entire ASC 820 Measurement in Level 3
ASC 820 grades a measurement by its lowest significant input: one unobservable number sends the entire mark to Level 3. What that costs you in disclosure.
By 409.AI Team - 2026-08-26
# One Unobservable Input Puts Your Entire ASC 820 Measurement in Level 3
Your analyst builds the mark exactly the way the textbook says to. Public comparables pulled from real trading data. A revenue multiple straight off that peer set. The portfolio company's own trailing revenue, tied to its audited financials. Three inputs, all observable, all defensible.
Then, because the company is private and a third the size of the smallest comp, the analyst applies a 25% company-specific adjustment.
The auditor looks at the file and puts the whole measurement in Level 3.
This catches finance teams off guard every reporting season. The instinct is that a number built mostly from market data should land somewhere in the middle of the fair value hierarchy, that "mostly observable" ought to buy an easier classification. It doesn't. ASC 820 doesn't average your inputs, and it doesn't grade on a curve.
The rule that does the work
The mechanic lives in ASC 820-10-35-37A. A fair value measurement is categorized in its entirety at the level of the *lowest* input that is significant to the measurement as a whole. One significant unobservable input, and the asset or liability is Level 3, however much observable data sits alongside it. PwC's [fair value measurement guide](https://viewpoint.pwc.com/dt/us/en/pwc/accounting_guides/fair_value_measureme/fair_value_measureme__9_US/chapter_4_concepts_u_US/45_inputs_to_fair_va_US.html) works through the mechanics of applying it.
There is no partial credit and no blended level. The hierarchy is decided by the weakest significant link, and everything else in the model comes along with it.
Worth saying clearly, because it gets read as an insult: the level is not a quality score. It describes how observable your inputs were, not how good your work was. A rigorous, well-documented Level 3 mark is better than a lazy Level 2 one. What the level does control is how much you have to show your work in the footnotes, and how hard your auditor looks.
"Significant" is a judgment, and it belongs to you
ASC 820 sets no bright line for significance. It doesn't require a quantitative threshold, and it doesn't tell you that 10% of the measurement is the cutoff. It asks for judgment based on factors specific to the asset.
PwC's guidance points at two questions worth asking about any input. How sensitive is the overall value to a change in it? And how likely is that input to move over the life of the position? The classic illustration is a ten-year interest rate swap with observable yield curve data for nine of those years. If the extrapolated tenth year barely moves the number, the measurement can stay in Level 2. If it moves the number materially, the whole swap is Level 3.
Run the same test on the mark above. Say the portfolio company has $12M of ARR and the peer set trades at 6x revenue, which puts enterprise value at $72M. The 25% company-specific adjustment takes it to $54M. If your fund holds 8% of the equity, that adjustment is the difference between a $5.8M position and a $4.3M one. An input worth $1.5M on a single holding is not a rounding item, and no auditor is going to accept that it wasn't significant.
One more wrinkle that trips people up on the second pass: an input can be insignificant when you first recognize the position and significant a year later, once markets or company assumptions have moved. Significance is retested at each measurement date, not decided once at entry. That is exactly why you want a written method for making the call and a habit of applying it the same way every quarter, rather than an argument constructed after the auditor asks.
The measurements you assumed were clean
Almost every direct private position was headed for Level 3 regardless. There is no ticker for a Series B stake, so the finding is rarely a surprise there. Our companion piece on [how funds mark startup positions under ASC 820](https://409.ai/articles/asc-820-level-3-fair-value-fund-portfolio-valuation) walks through that side of it.
The rule bites hardest on measurements you thought were mostly mechanical.
Contingent consideration is the common one. An earnout's structure is contractual and fully observable, but the revenue projections you probability-weight to value it are not, and they drive the answer. That is why [earnouts under ASC 805](https://409.ai/articles/earnout-contingent-consideration-valuation-asc-805) sit in Level 3 and stay there, remeasured every period until they settle.
Warrants and other instruments carried at fair value follow the same path. The risk-free rate and the contractual term are observable. Volatility for a private issuer is an estimate built from a peer group you selected. So is the expected time to a liquidity event, which is one reason [option pricing model and backsolve allocations](https://409.ai/articles/409a-allocation-methods-opm-pwerm-backsolve) are unobservable-input exercises no matter how much market data feeds the top of the model. A [discount for lack of marketability](https://409.ai/articles/discount-lack-marketability-dlom-409a-valuation) has the same character: supportable from studies and models, observable in no market anywhere.
What Level 3 actually costs you in the footnotes
Here is where the common description of Level 3 goes wrong. The package people recite, a quantitative table of unobservable inputs, a sensitivity narrative, and a beginning-to-ending balance reconciliation, is the *public company* package. ASU 2018-13 rewrote the disclosure framework, and private funds and private operating companies got real relief.
Every entity still discloses which level a recurring measurement sits in, the quantitative information about the significant unobservable inputs used, and any transfers into or out of Level 3 along with the reasons for them.
Public business entities carry the rest: the range and weighted average of those unobservable inputs and how the weighted average was calculated (ASC 820-10-50-2(bbb)(2)(i)), the full roll-forward including sales, settlements, and realized and unrealized gains and losses, and the narrative description of measurement uncertainty in ASC 820-10-50-2(g), which explains how a reasonably different input would have produced a significantly different value.
Nonpublic entities are exempt from those three. Under ASC 820-10-50-2F and 820-10-50-2G, a private entity discloses an abbreviated version of Level 3 activity instead: purchases and issues, each shown separately, plus transfers in and out. No sales, no settlements, no gains and losses roll-forward, no sensitivity narrative. Deloitte's [fair value measurements roadmap](https://dart.deloitte.com/USDART/home/codification/broad-transactions/asc820-10/roadmap-fair-value-measurements-disclosures/chapter-11-disclosure/11-2-fair-value-disclosures-requirements) and [FASB's own text of ASU 2018-13](https://storage.fasb.org/ASU%202018-13.pdf) both lay the split out. ASU 2018-13 also dropped the old requirement to describe your policy for the timing of transfers between levels.
Relief is not the same as nothing, and this is the part worth internalizing. The quantitative unobservable input disclosure survives for everyone. You have to name the input and put a number next to it in the footnote. "We applied judgment based on company-specific factors" is not a disclosure. That 25% adjustment goes in print, in a table, where your LPs and your auditor read it. The classification decision made in a spreadsheet in February becomes a public statement about your assumptions in March.
Two things that aren't the escape hatch
Funds sometimes reach for the net asset value practical expedient. It applies to investments in entities that calculate NAV per share, a fund of funds holding interests in underlying funds, for example. Those investments aren't categorized in the hierarchy at all, though they still appear in the total column so the disclosure reconciles to the balance sheet. FASB narrowed this in [ASU 2015-07](https://storage.fasb.org/ASU%202015-07_2.pdf) to investments actually measured using the expedient. It does nothing for a direct stake in a portfolio company.
The second confusion runs the other way, and it hits portfolio-company finance leads more than fund CFOs. Share-based payment awards under Topic 718 sit outside ASC 820's disclosure requirements. Your option expense, and the 409A that feeds it, gets disclosed under [ASC 718's own rules](https://409.ai/articles/asc-718-stock-based-compensation-startup-guide), not as a Level 3 line item in a fair value footnote. Two different questions about the same company, answered in different parts of the financial statements. Reporting one under the other's framework creates an audit comment out of nothing.
What to fix before the audit, not during it
Write the significance method down before you need it, and use the same one across positions. Name every unobservable input in your models now, because you will be naming them in the footnote anyway. Keep a record of what changed when a position moves between levels, since the reason for the transfer is disclosable even for private funds. And anchor the model to your entry price and carry it forward rather than rebuilding the estimate from scratch each quarter, the calibration discipline covered in the companion piece above.
The hierarchy isn't asking how much work you did. It's asking a narrower question: which inputs could a reader verify independently, and which ones do they have to take from you? Answer that honestly in your own file, before an auditor answers it for you, and the classification stops being a surprise. It becomes the thing you already documented.
If you need Level 3 marks that hold up in an audit, that's what [409.ai's ASC 820 valuations](https://409.ai/products/asc-820) are built for. For the interpretive detail on measurement itself, EY's [Financial Reporting Developments guide on fair value measurement](https://www.ey.com/content/dam/ey-unified-site/ey-com/en-us/technical/accountinglink/documents/ey-frdbb1462-09-30-2025.pdf), refreshed in September 2025, is the reference practitioners actually keep open.