What an Auditor Actually Tests in a Multiple-Based Valuation

There is a persistent misunderstanding about what happens in a valuation audit. Managers prepare to defend their judgement, why this multiple, why this peer, why this growth rate. Auditors mostly ask something narrower and harder: show me how you got here, and show me that you got here the same way last quarter.

Judgement is permitted. Unreconstructable judgement is not.

Consistency comes first

The first thing tested is whether the method is the same one used at the previous measurement date. Not because consistency is sacred, but because a change of method is the easiest way to move a valuation without admitting to moving it.

Changes are allowed and sometimes required. What is required with them is an explanation: what changed about the asset that made the previous method unsuitable. A switch from a revenue multiple to an earnings multiple in the quarter the company became profitable is straightforward. The same switch in the quarter the revenue multiple would have produced a write-down is not.

Auditors look at the pattern across the portfolio. One method change is a fact; five in the same direction in the same quarter is a finding.

The peer set, and whether it moved

Expect to be asked how the comparable set was constructed and why each constituent belongs. Sector label alone will not carry it, the question is whether these companies have comparable economics.

Then the harder question: has the set changed since last period, and if so, why? Removing the two lowest-multiple peers without a documented reason is the classic finding. If a peer was dropped because it was acquired or its business changed materially, say so in the file at the time, not in response to the question.

Dispersion within the set gets attention too. A set ranging from 2× to 14× invites a question about whether the median is meaningful, and the honest answer is usually to narrow the set.

Where every input came from

Each number in the calculation needs a source and a date. This sounds procedural until it is tested: a multiple with no traceable origin is, for audit purposes, not evidence.

The date matters as much as the source. A multiple has to be as at the measurement date, not as at whenever the file was last updated. Using a figure from two quarters ago during a period of rapid market movement is a real misstatement, not a technicality.

This is the single most common gap I see in practice, and it is also the easiest to close: use one documented, dated dataset consistently, and record which vintage was used each period.

Adjustments, and whether they double-count

Every adjustment between the peer multiple and the applied multiple gets examined separately: size, marketability, control, company-specific performance. Each needs a rationale and, where possible, a reference for its magnitude.

The specific thing being tested is double-counting. A discount for illiquidity applied on top of a peer set already composed of small, thinly traded companies is charging twice for the same characteristic. So is a size discount plus a small-company risk premium in the discount rate.

Laying the adjustments out as an explicit bridge, peer median, then each adjustment with its sign and size, then the applied multiple, makes double-counting visible to you before it is visible to the auditor.

Calibration back to what you paid

For portfolio holdings, expect to be asked how the current method reconciles to the entry price. If the technique would have implied a materially different value at entry than the price actually paid, that gap needs an explanation, and it needs to have been documented at entry, not reconstructed now.

Where calibration exists, subsequent valuations become a question of what changed: company performance against plan, and movement in the comparable set. Where it does not, every quarter is a fresh argument.

What a file that passes looks like

One page per holding, with: method and why it suits the asset; peer set and construction logic, with changes since last period flagged; every input with source and date; the adjustment bridge; calibration to entry; and a sensitivity showing what a reasonable variation in the key input does to the value.

The test to apply yourself, before anyone else does: hand the file to a colleague who was not involved and ask them to arrive at your number. If they can, it will hold. If they need you in the room to explain it, it will not.