IPEV in Practice: What Fund Reporting Actually Requires

Most fund managers can name the IPEV guidelines. Fewer can say, without checking, what their own valuation file would look like to someone reading it cold six months later. That gap is where most audit friction comes from.

IPEV sits on top of fair value as defined in the accounting standards. It does not invent a separate definition, it explains how to apply fair value to assets that do not trade. Understanding that relationship makes the rest of it much easier to follow.

Fair value is the anchor, not a technique

The objective is the price that would be received to sell the asset in an orderly transaction between market participants at the measurement date. That definition comes from the accounting standards, IFRS 13 and, in the US, ASC 820, and IPEV is guidance on applying it.

Two things follow. It is a market participant view, not the holder’s view: what a buyer would pay, not what the fund hopes to realise. And it is measured at the date, which means conditions after the reporting date are generally not reflected, however tempting.

Cost is not fair value. Recent transaction price is evidence of fair value at the transaction date, and its weight decays as time and events accumulate.

Choosing a technique, and sticking to it

Market approaches, multiples of earnings or revenue drawn from comparable companies or transactions, dominate in practice, because they are the ones a market participant would actually use for an operating business.

The requirement is not that you pick the theoretically best method. It is that the method suits the asset, that it is applied consistently period to period, and that a change in method is explained when it happens. Switching technique in the quarter a valuation would otherwise fall is the pattern auditors look for hardest.

Income approaches have their place, particularly where cash flows are predictable. For early-stage assets with no earnings and no reliable forecast, calibration to the last round, adjusted for what has changed since, is frequently the most defensible route.

Calibration is the step most files are missing

Calibration means: at entry, work out what your chosen technique would have produced, and reconcile it to the price you actually paid. If your comparable set implies 6× revenue and you paid 9×, that gap is information. Document it.

At subsequent measurement dates, you then move the valuation by what has changed, the company’s performance against plan, and the movement in the comparable set, rather than re-deriving from scratch each quarter. This is what produces valuations that move for reasons you can name.

Files without calibration tend to show a flat carrying value for several quarters followed by a sudden step. The step is usually correct; the flat period was not.

Inputs, and why their source matters

The fair value hierarchy ranks inputs by observability: quoted prices first, other observable inputs second, unobservable inputs last. Private company valuations sit largely in the third level, which raises rather than lowers the documentation burden.

What that means concretely: every input needs a source and a date. A sector multiple recalled from memory is not an input. A multiple taken from a dated, documented dataset is. The distinction is invisible until someone challenges the number, at which point it is the whole argument.

Where a multiple is adjusted, for size, for marketability, for the company’s position against its peer set, the adjustment and its rationale belong in the file, separately from the base figure.

What a defensible file contains

For each holding: the technique used and why it suits the asset; the comparable set and how it was built; the inputs with sources and dates; the calibration to entry price and what has changed since; each adjustment with its rationale; and the resulting value with a sensitivity around it.

None of that is exotic. What makes it hard is doing it every quarter for every holding, which is why the funds that do it well treat it as a standing process with a fixed data source rather than a quarterly scramble.

The test is simple: could a reader who was not in the room reconstruct your number? If yes, the file will hold. If no, it will not, regardless of how good the judgement behind it was.