How to Audit Estimates and Fair Values Under ISA 540

Why estimates are the hardest area to get right

Every other audit assertion has a natural anchor. Cash ties to the bank confirmation. Revenue ties to invoices and delivery documentation. Fixed assets tie to purchase contracts and physical existence. Estimates don’t tie to anything external in the same way. They’re forward-looking judgments built on assumptions, and the range of reasonable outcomes can be wide enough that two competent accountants could produce materially different numbers from the same data.

ISA 540.2 acknowledges this directly: accounting estimates vary widely in their nature, the degree of estimation uncertainty, and the complexity of the methods used to make them. That variation means your audit approach can’t be uniform. A straightforward inventory provision based on ageing data requires different procedures than a fair value (FV) measurement of an unquoted equity investment using a discounted cash flow model. Both are estimates. One needs a recalculation check. The other might need a management’s expert plus a sensitivity analysis.

The revised ISA 540 (effective for periods beginning on or after 15 December 2019) restructured the standard around four inherent risk factors. Estimation uncertainty and complexity are the first two. Subjectivity is the third. The fourth is how these factors interact with each other on a specific estimate. Every estimate gets assessed against these factors, and your response scales accordingly. A low-complexity estimate with low uncertainty (a depreciation charge using a standard straight-line method) sits at one end. A high-complexity estimate with high uncertainty and high subjectivity (a Level 3 fair value measurement with unobservable inputs) sits at the other. ISA 540.16 requires you to identify where each estimate falls on that spectrum before designing procedures.

What ISA 540 (Revised) changed and why it matters for your file

Under the previous version of ISA 540 , a more binary approach was acceptable: test management’s estimate, develop your own, or review subsequent events. The revised standard doesn’t remove those options (they appear in ISA 540.18 ), but it wraps them in a risk assessment framework that forces you to document why you chose the approach you chose.

What changed most in practice is the emphasis on “stand back” evaluation. ISA 540.34 requires the auditor to evaluate, based on the audit procedures performed and evidence obtained, whether the accounting estimates and related disclosures are reasonable in the context of the applicable framework. Paragraph 35 then requires a separate assessment of whether indicators of possible management bias exist. These aren’t review points. They’re required procedures with their own documentation.

Another change that affects mid-tier files: ISA 540.14 made the retrospective review a planning procedure, not an afterthought. You compare prior-year estimates to actual outcomes (or to subsequent re-estimates) to understand the entity’s estimation process, to identify inherent risk factors, and to detect indicators of management bias. The comparison must be specific enough to be useful. Noting that “prior year provisions were materially in line with actuals” doesn’t satisfy ISA 540.14 if the individual estimates moved in opposite directions that happened to net off.

For the ISA 540.14 retrospective review, the ciferi ISA 520 analytical review calculator can help you structure the prior-year-to-actual comparison across multiple estimate categories.

The three audit responses and how to choose between them

ISA 540.18 gives you the response options. Your choice depends on the nature of the estimate, the availability of data, the level of estimation uncertainty, and the assessed inherent risk factors.

Obtaining audit evidence from events ( ISA 540.18 (a))

This is the strongest form of evidence because it replaces the estimate with an actual outcome. If the client has a year-end provision for a legal claim and the case settles before you sign the audit report, the settlement amount is better evidence than any estimate. Use this response where subsequent events resolve the estimate entirely or narrow the range substantially. The limitation is timing: many estimates don’t resolve within the audit window. An expected credit loss provision won’t crystallise into actual defaults within two months of year-end.

Testing how management made the estimate ( ISA 540.18 (b))

We’ve seen this on about half the engagements we’ve worked: testing management’s process is the default response on mid-tier files. You test the method, the data, and the assumptions. ISA 540.18 (b) breaks this into components: whether the method is appropriate under the framework, whether the data is complete and accurate, whether the significant assumptions are reasonable, and whether the calculations are mathematically correct. For an FV estimate, “testing the method” means evaluating whether the valuation technique (market approach, income approach, cost approach under IFRS 13.61 -66) is appropriate given the nature of the asset or liability and the availability of observable inputs.

The risk with this response is that testing each component in isolation can miss the overall picture. The method might be appropriate. Data inputs might be accurate. Each assumption might be individually defensible. But the combination of individually optimistic assumptions produces an estimate biased in aggregate. That’s why ISA 540.34 requires the stand-back evaluation even when you’ve tested the components. Nobody enjoys challenging a CFO’s valuation model, but skipping that conversation is how the file gets flagged in inspection.

Developing an auditor’s point estimate or range ( ISA 540.18 (c))

You build your own estimate using different data, a different method, or different assumptions, then compare your result to management’s. This works well for estimates where the inputs are available but management’s judgment is questionable. For an expected credit loss provision, you might run the client’s debtor book through your own probability-of-default model (or a published model for the client’s industry) and compare the output to management’s provision. If your range is €180,000 to €240,000 and management’s provision is €195,000, the estimate sits within a reasonable range. If management’s provision is €85,000, you have a presumptive misstatement.

Developing your own estimate is more work than testing management’s process. But it produces the strongest evidence for estimates with high subjectivity because you’re not anchored to management’s assumptions. For fair value measurements, this may involve engaging an auditor’s expert under ISA 620 to produce an independent valuation.

How to run the retrospective review properly

ISA 540.14 requires a comparison of prior-period estimates to the actual outcomes (or the subsequent re-estimates made in the current period). Most files include this as a table: prior year estimate in one column, actual or current year re-estimate in the other, difference in the third. That format is correct. The analysis that accompanies it is where files fail.

The purpose of the retrospective review isn’t to confirm that the client’s estimates were “approximately right.” It serves four distinct functions under ISA 540.14 : understanding the entity’s estimation process, identifying estimation uncertainty indicators, detecting possible management bias, and informing the current-year risk assessment. If your retrospective review note says “prior year estimates were within an acceptable range of actual outcomes” and nothing else, you’ve addressed none of these functions specifically.

A proper retrospective review for the expected credit loss provision on a €30M debtor book might look like this. Prior year provision: €210,000. Actual write-offs during the current year relating to prior-year receivables: €285,000. Difference: €75,000 under-provision (36%). This comparison tells you the client’s estimation process underestimated defaults. It identifies a high estimation uncertainty indicator (the actual outcome was materially different from the estimate). And it raises a potential management bias flag if the under-provision is in the same direction for two or more consecutive years.

The key: do this for each material estimate individually. Don’t net provisions against warranty claims against restructuring costs. An accurate total can hide directional bias in the components.

Identifying management bias in estimates

ISA 540.32 requires you to evaluate whether indicators of possible management bias exist. ISA 540 .A129 lists the types of indicators you’re looking for. The critical ones for mid-tier engagements are directional consistency, anchoring behaviour, selective use of assumptions, and optimistic weighting of scenarios.

Directional consistency means the estimates all move in the same direction. If every judgmental estimate on the file sits at the optimistic end of the range (low provisions, high asset valuations, long useful lives, aggressive revenue recognition on estimates), that’s a bias indicator. No single estimate may be unreasonable on its own. The pattern is the problem.

Anchoring happens when management sets the current-year estimate equal to the prior year without updating for changed circumstances. If the client’s warranty provision has been exactly €120,000 for four consecutive years despite revenue growth of 30%, the provision hasn’t been re-estimated. It’s been copied. ISA 540.14 should catch this through the retrospective review, but only if you compare the provision to the underlying risk drivers (warranty claim rates, revenue volume) rather than just to the actual claims.

Selective use of assumptions is the subtlest indicator. The client uses conservative assumptions for some inputs and aggressive ones for others, always in the direction that produces a favourable result. A property developer might use conservative cost estimates (reducing profit recognition on percentage-of-completion contracts) while simultaneously using aggressive selling price assumptions (increasing the net realisable value of inventory). Both assumptions are individually within a defensible range. Together, they produce an NRV calculation that’s systematically overstated.

Document your bias assessment explicitly. ISA 540.35 requires a conclusion. “No indicators of management bias identified” is a valid conclusion, but only after you’ve documented the analysis that supports it. A blank working paper section doesn’t satisfy the requirement.

The practical difficulty with bias assessment on mid-tier engagements is that you’re often dealing with a small number of estimates. If the client has two material provisions and one fair value measurement, a directional pattern across all of them is suggestive but not definitive. Still, document what you observe. If all estimates move in the same direction, note it. If the direction is consistently favourable to management’s compensation targets or loan covenants, note that too. ISA 540 .A130 identifies these as specific bias indicators worth evaluating. You don’t need to prove bias exists. You need to demonstrate that you looked for it and recorded what you found. A well-documented bias assessment that concludes “no bias identified after analysis of directional patterns across five estimates” is far stronger than a pro-forma conclusion generated at the end of the audit without analysis behind it.

Worked example: Janssen Bouwgroep B.V.

Client scenario

Janssen Bouwgroep B.V. is a mid-sized Dutch construction company. Revenue for 2024: €85M. The balance sheet includes a warranty provision of €640,000, an expected credit loss provision of €195,000, and a percentage-of-completion revenue accrual on four active projects totalling €12.4M. Fair value measurement: one investment property carried at €3.8M under IAS 40 , valued internally by management using the income capitalisation approach.

Conclusion. The file shows a quantified misstatement on the investment property (€292,000), a directional bias pattern across four estimates, and a specific retrospective review that identified two consecutive years of warranty under-provision. The reviewer sees the individual estimate assessments, the independent valuation with market data, and a documented bias evaluation that connects the findings across estimates.

Your engagement checklist

Common mistakes regulators flag

• The FRC’s 2022-23 Audit Quality Inspection Report found that audit teams routinely failed to assess estimation uncertainty, complexity, and subjectivity as separate inherent risk factors. Many files combined them into a single assessment, which prevented the team from designing responses targeted at the specific risk driver. ISA 540.16 requires separate consideration.

Related content

• Accounting estimates. The ciferi glossary entry explains estimation uncertainty, the inherent risk factor framework, and the key ISA 540 paragraph references.
• IFRS 9 expected credit loss calculator. Calculates ECL provisions using the simplified approach and produces working paper documentation for the probability-of-default and loss-given-default assumptions.
• How to audit revenue recognition under IFRS 15 . Variable consideration estimates under IFRS 15.56 follow the same ISA 540 framework for testing assumptions and identifying bias.

Word count target: 3,000-4,500 words.

Related ciferi content

Related guides:

Put audit concepts into practice with these free tools:

Frequently asked questions