IFRS 9 Expected Credit Losses: How to Audit the Provision Matrix

You're reviewing a trade receivables balance of €14 million and the client hands you a one-page provision matrix. Loss rates by ageing bucket, a small overlay for "macroeconomic conditions," and nothing else. The engagement partner wants to know if it's sufficient. You aren't sure where to start.

To audit the IFRS 9 expected credit loss provision matrix, test the completeness and accuracy of historical loss rates against actual write-off data, evaluate whether forward-looking adjustments reflect current conditions under IFRS 9.5.5.17, verify staging criteria against the standard's credit risk deterioration thresholds, and apply ISA 540 procedures to the estimate as a whole.

What the provision matrix is and why IFRS 9 allows it

IFRS 9.B5.5.35 permits entities to use a provision matrix as a practical expedient for measuring expected credit losses on trade receivables that do not contain a significant financing component. Instead of running a full discounted cash flow model, the entity groups receivables by shared credit risk characteristics (most commonly ageing buckets or geographic region) and applies a loss rate to each group.

The provision matrix exists because most non-financial entities lack the data infrastructure to run loan-by-loan probability-of-default models. For a company with thousands of trade receivable invoices, the matrix is the only realistic option. IFRS 9.5.5.15 requires lifetime expected credit losses for trade receivables when the entity applies the simplified approach (which most non-financial entities do). The matrix is the delivery mechanism for that requirement.

What makes this different from the old IAS 39 incurred loss model is the forward-looking element. Under IAS 39, you only recognised a loss when a trigger event had already occurred. Under IFRS 9.5.5.17, the entity must incorporate reasonable and supportable information about future economic conditions. A provision matrix built entirely on historical write-off percentages, with no adjustment for current or forecast conditions, does not comply with the standard. This is the single most common deficiency auditors find.

The three stages and why they matter for the matrix

Many auditors assume the staging model (IFRS 9.5.5.1-5.5.5) doesn't apply when the entity uses the simplified approach for trade receivables. That's partially correct. Under the simplified approach, the entity always recognises lifetime ECLs, so the distinction between Stage 1 (12-month ECL) and Stage 2 (lifetime ECL) collapses. But staging still matters for two reasons.

First, if the entity has contract assets, loan receivables, or lease receivables alongside its trade receivables, the general model applies to those. You'll need to assess whether credit risk has increased significantly since initial recognition (IFRS 9.5.5.9) for every instrument not covered by the simplified approach. Second, even within the simplified approach, IFRS 9.5.5.20 requires the entity to move receivables to Stage 3 (credit-impaired) when there is objective evidence of impairment. A receivable past due by 180 days from a customer in insolvency proceedings isn't sitting in an ageing bucket at the average loss rate. It's credit-impaired and should be assessed individually or within a group that reflects its actual expected recovery.

The staging assessment under IFRS 9.5.5.9-11 asks: has the credit risk of this instrument increased significantly since initial recognition? The standard provides indicators in IFRS 9.B5.5.17, including past-due status of more than 30 days as a rebuttable presumption (IFRS 9.5.5.11). When auditing the matrix, check whether the client has a process to identify receivables that should be individually assessed rather than pooled.

How to test historical loss rates

The loss rates in a provision matrix are derived from the entity's own write-off history. The audit procedure here is straightforward in concept but tedious in execution: you're testing whether the rates used in the matrix actually reflect what happened.

Start by obtaining the client's historical loss data for the past two to four years. For each ageing bucket in the matrix, compare the loss rate applied to the actual percentage of receivables in that bucket that were ultimately written off. Request the roll-forward data: receivables that sat in the 31-60 day bucket at each prior year-end, and what happened to them. Were they collected, written off, or still outstanding?

Recalculate the loss rate yourself. If the client says the loss rate for the 61-90 day bucket is 2.5%, you need to verify that number against actual outcomes. Pull the receivables that were 61-90 days past due at the prior two or three year-ends, trace them to collection records or write-off journals, and compute the percentage that became uncollectable. The IFRS 9 ECL calculator can help structure this recalculation by ageing bucket.

Watch for these problems:

• Loss rates that haven't changed in years despite changes in the customer base or payment behaviour
• Write-offs that were delayed (the client waits 18 months to write off a balance that was clearly uncollectable at 6 months, which depresses the apparent loss rate in the earlier buckets)
• Exclusion of specific large balances from the historical data ("we took that one out because it was unusual"), which skews the rate downward
• No distinction between customer segments that have different credit risk profiles

If the client's revenue has shifted toward a riskier customer segment, loss rates from three years ago may understate current expected losses even before you get to the forward-looking adjustment.

How to evaluate forward-looking adjustments

IFRS 9.5.5.17(c) requires the entity to consider reasonable and supportable information about past events, current conditions, and forecasts of future economic conditions. For a provision matrix, this means the historical loss rates must be adjusted to reflect where the economy is heading.

In practice, most forward-looking adjustments on trade receivable matrices are small overlays. The client might add 0.5% across all buckets and call it a macroeconomic adjustment. Your job is to test whether that adjustment is reasonable, not whether it's precise. ECL measurement is an estimate, and ISA 540.A9 acknowledges that estimates inherently involve uncertainty.

Ask for the basis of the adjustment. What data did the client use? GDP forecasts, unemployment projections, industry-specific indicators, customer payment trend data? If the answer is "professional judgement" with nothing behind it, that's a problem. IFRS 9.B5.5.52 requires the entity to consider information that is reasonably available without undue cost or effort. For a trade receivable matrix at a mid-sized entity, you'd expect at least a reference to published economic forecasts from the national statistics office or central bank, and some documented reasoning about how those forecasts translate into higher or lower loss rates.

Test the directional consistency. If the client's sector is experiencing rising insolvencies and lengthening payment terms, a forward-looking adjustment that reduces the provision below the historical average requires strong justification. Conversely, if the macroeconomic outlook has improved and the customer base is stable, a zero adjustment might be reasonable.

Check the granularity. A single percentage added to all ageing buckets assumes the economic outlook affects all customers equally. For some entities that's defensible (a retailer with thousands of homogeneous customers). For others (a construction company with 30 large contract debtors in different countries), it is not.

Document your evaluation against the specific forecast data the client used, the reasonableness of the link between that data and the loss rate adjustment, and whether the adjustment was applied at an appropriate level of granularity.

How to assess staging criteria

Even under the simplified approach, you need to confirm the client correctly identifies credit-impaired receivables (Stage 3) and assesses them separately. IFRS 9.5.5.20 and the definition in Appendix A require objective evidence of a loss event affecting estimated future cash flows. Indicators include significant financial difficulty of the debtor, breach of contract, probability of bankruptcy, and disappearance of an active market.

The audit procedure has two parts. First, test whether the client's process for identifying credit-impaired receivables is designed to catch them. Is there a trigger mechanism? Does someone review the aged listing and investigate balances past a certain threshold? Does the credit control team flag customers that have entered administration?

Second, test whether the process operated. Select a sample of receivables that were past due by more than 90 days at the balance sheet date (or another threshold appropriate for the entity's payment terms). For each, determine whether the client assessed it for individual impairment or left it in the general matrix. If a customer that filed for insolvency in November is still sitting in the 91-120 day bucket at the December year-end, the matrix loss rate for that bucket is almost certainly too low for that balance.

IFRS 9.5.5.11 provides the 30-day past-due rebuttable presumption for determining significant increase in credit risk. Under the general model, this moves the receivable from Stage 1 to Stage 2. Under the simplified approach, lifetime ECLs already apply, so the 30-day presumption is less critical. But the 90-day presumption for default (IFRS 9.B5.5.37) matters for identifying Stage 3 balances regardless of which approach the entity uses. If the entity rebuts this presumption for specific receivables, document the basis.

ISA 540 overlay: auditing the ECL as an accounting estimate

The ECL provision is an accounting estimate, and ISA 540.13 requires you to understand how management made the estimate, including the method, assumptions, data sources, and the basis for selecting them. For the provision matrix, this translates to understanding the methodology for deriving loss rates, the data behind forward-looking adjustments, and the criteria for moving receivables between stages.

ISA 540.18 requires the auditor to assess the risks of material misstatement related to the estimate. Ask yourself: where could this estimate be wrong enough to matter? The three most common sources of material misstatement in a provision matrix are understated loss rates (historical data is stale or cherry-picked), inadequate forward-looking adjustments, and failure to individually assess credit-impaired balances.

ISA 540.21-22 set out the auditor's response. You can test how management made the estimate (reperform the calculation, test the inputs), develop an independent estimate for comparison, or review subsequent events that confirm or contradict the estimate. For the provision matrix, testing management's estimate is usually the most efficient approach. Developing an independent estimate requires access to the same granular receivable and write-off data the client used, which you'll already be testing anyway.

Watch for management bias indicators under ISA 540.32. A provision matrix that has been consistently too low (actual write-offs exceed the provision each year) and the client hasn't adjusted the rates upward is a bias indicator. A forward-looking adjustment that always rounds down is another. Document your assessment of whether the estimate falls within a reasonable range and whether any directional bias exists.

ISA 540.A130 reminds auditors that the point estimate selected by management may differ from the auditor's point estimate. If the difference is individually immaterial, record it on the ISA 450 summary of misstatements. If it is material, discuss with management and consider the effect on the audit opinion.

Worked example: auditing the provision matrix at Dekker Wholesale B.V.

Scenario: Dekker Wholesale B.V. distributes industrial cleaning supplies across the Benelux region. Revenue for the year ended 31 December 2025 was €28 million. Trade receivables at year-end totalled €4.2 million. Dekker uses the simplified approach under IFRS 9 and measures ECLs through a provision matrix with five ageing buckets. Overall materiality is €140,000. The client's total ECL provision is €189,000.

1. Obtain and understand the matrix.

Dekker's matrix applies these loss rates: Current (0.3%), 1-30 days (1.0%), 31-60 days (3.0%), 61-90 days (8.0%), 91+ days (25.0%). A flat 0.4% overlay is added across all buckets as a forward-looking adjustment.

Documentation note: Record the matrix structure, loss rates, and overlay in the working paper. Reference the client's ECL policy document and confirm it states the simplified approach is applied under IFRS 9.5.5.15.

2. Test historical loss rates against actual write-off data.

Request Dekker's receivable ageing at 31 December 2023 and 31 December 2024, plus all write-offs in the subsequent 12 months for each cohort. For the 61-90 day bucket, the 2023 cohort had €310,000 in receivables. Of those, €28,500 was written off (9.2%). The 2024 cohort had €275,000, with €19,250 written off (7.0%). The two-year average is 8.1%, against the client's applied rate of 8.0%.

Documentation note: Show the recalculation for each bucket. State which years were used and why. Flag any bucket where the applied rate differs from the recalculated average by more than 20% relative.

3. Evaluate the forward-looking adjustment.

Dekker's CFO points to a Netherlands Bureau for Economic Policy Analysis (CPB) projection showing GDP growth of 1.2% for 2026, down from 1.8% in 2025. The CFO added 0.4% to all buckets. Challenge: Dekker's customer base is concentrated in the Belgian hospitality sector (40% of revenue), where payment delays increased 12% in H2 2025. A uniform overlay doesn't reflect this concentration risk.

Documentation note: Record the data source for the adjustment (CPB March 2026 forecast). Document the auditor's assessment that the adjustment is directionally appropriate but may be understated for the Belgian hospitality segment. Quantify the potential understatement: if the 31-60 and 61-90 day buckets for Belgian customers carry a 1.0% higher loss rate, the additional provision would be approximately €11,000 (below materiality but recorded as an unadjusted difference).

4. Test staging (credit-impaired identification).

Select all receivables past due by more than 90 days. Total: €126,000 across 8 customers. One customer (Maison Bruges SPRL, balance €47,000, 140 days past due) filed for judicial reorganisation in November 2025. Dekker has this balance in the 91+ bucket at 25%. Expected recovery from judicial reorganisation proceedings in Belgium averages 15-30% for unsecured trade creditors. The matrix rate of 25% implies 75% recovery, which is too high.

Documentation note: Propose an adjustment to individually assess Maison Bruges. At an estimated recovery of 20%, the specific provision should be €37,600 versus the matrix provision of €11,750. Difference: €25,850. Record on the summary of unadjusted differences alongside the €11,000 from step 3. Combined: €36,850 (below materiality of €140,000, but both directional and therefore a bias indicator under ISA 540.32).

5. Conclude.

The matrix methodology is appropriate under IFRS 9.B5.5.35. Historical loss rates are supported by write-off data. The forward-looking adjustment exists but is understated for the Belgian hospitality concentration. One credit-impaired receivable requires individual assessment. Total unadjusted differences of €36,850 are below materiality but both adjustments reduce the provision upward, indicating potential management optimism. Report to those charged with governance.

Documentation note: Cross-reference to ISA 450 schedule. Note the directional consistency of unadjusted differences. Conclude that the ECL provision, after considering identified misstatements, is not materially misstated.

Sensitivity testing and range of outcomes

ISA 540.A98-A101 expect the auditor to consider the range of possible outcomes for an accounting estimate. For the provision matrix, this means testing what happens if assumptions change.

Run two scenarios. In the first, increase loss rates by 25% across all buckets (reflecting a deterioration in the economic environment). In the second, increase them by 50%. Calculate the resulting provisions and compare them to overall materiality. If a 25% increase in loss rates produces a provision change that exceeds materiality, the estimate has high estimation uncertainty, and you need to document this in the completion memo.

Also test the forward-looking adjustment sensitivity. What if the overlay doubled? What if it were removed entirely? If removal of the overlay produces a provision still within a reasonable range of the recorded amount, the overlay is not the critical assumption. If removal produces a material change, the overlay is the critical assumption, and your evidence about its reasonableness needs to be correspondingly strong.

This sensitivity analysis takes 20 minutes with a spreadsheet. It gives you two things: evidence that you considered the range of outcomes (ISA 540 requirement), and a documented basis for concluding whether the recorded provision falls within a reasonable range.

Audit procedures checklist

1. Confirm the entity applies the simplified approach under IFRS 9.5.5.15 for trade receivables and that the receivables qualify (no significant financing component).
2. Recalculate historical loss rates for each ageing bucket using at least two years of write-off data, traced to the general ledger and supporting write-off approvals.
3. Test the forward-looking adjustment against identified data sources (published economic forecasts, sector-specific indicators, customer payment trends) and evaluate whether the granularity of the adjustment matches the risk profile of the receivable book.
4. Select all receivables past due by more than 90 days (or the entity's default threshold) and verify that credit-impaired balances are individually assessed rather than pooled in the matrix at the standard loss rate.
5. Perform sensitivity analysis on the key assumptions (loss rates and forward-looking overlay) to establish a range of reasonable outcomes and assess whether the recorded provision falls within that range (ISA 540.A98-A101).
6. Assess management bias under ISA 540.32 by comparing the provision to actual write-off experience over the past two to three years and documenting whether unadjusted differences are directionally consistent.

Common mistakes

• Accepting the provision matrix without recalculating the loss rates against actual write-off data. The FRC's thematic review on IFRS 9 implementation found that auditors frequently tested the mathematical accuracy of the matrix but not the underlying loss rate derivation.
• Treating the forward-looking adjustment as immaterial and therefore not testing it. Even a small overlay becomes significant when applied across a large receivable balance, and the absence of any overlay is a non-compliance issue under IFRS 9.5.5.17.
• Failing to identify credit-impaired receivables that should be removed from the matrix and assessed individually, particularly when the entity does not have a formal process for flagging Stage 3 balances.

Related content

• Expected credit loss (ECL). Glossary entry covering the IFRS 9 ECL model, the difference between 12-month and lifetime ECLs, and how the simplified approach works for trade receivables.
• IFRS 9 ECL calculator. Calculator that structures the provision matrix by ageing bucket, applies historical loss rates, and layers in forward-looking adjustments for a documented ECL output.
• ISA 540 accounting estimates: auditing assumptions, data, and methods. [FUTURE POST] Application guide covering the ISA 540 audit approach for all accounting estimates, including the procedures for testing management's method and developing auditor's independent estimates.