Most corporate ECL models fail at the same three points. Here is how to build one that auditors accept and finance teams can run every month.
Key takeaways
- Segment receivables by credit characteristics before you compute loss rates, not after.
- Historical loss rates must be built from completed collection cycles, not from the current ageing.
- The forward-looking adjustment has to be explicit, documented and refreshed, or auditors will challenge it.
Corporates are permitted to use the simplified approach under IFRS 9, which removes the need to track significant increases in credit risk and always recognises lifetime expected credit losses on trade receivables, contract assets and lease receivables. The practical expedient most entities adopt is the provision matrix. It sounds simple. In our experience across Gulf corporates, the same three weaknesses appear in matrix after matrix, and they are the reason audit adjustments arise late in the reporting cycle.
Weakness one: segmentation by convenience
A single loss rate per ageing bucket applied across an entire receivables book assumes that a government ministry, a listed contractor and a small distributor share the same credit behaviour. They do not. IFRS 9 requires receivables to be grouped on the basis of shared credit risk characteristics. In Gulf portfolios the segmentation that usually matters is customer type (government, government-related, large corporate, SME, retail), geography where the entity trades across borders, and product or contract type where collection terms differ materially.
The test of good segmentation is whether historical loss experience differs between segments. If it does not, merge them. If it does, keep them separate even if the segment is small. Auditors will ask for this analysis; produce it as part of the methodology paper rather than in response to a query.
Weakness two: loss rates built from a snapshot
The most common error is to derive loss rates from the current ageing report: for example, treating the proportion of the book currently over 360 days as the loss rate for that bucket. That is a point-in-time picture, not a loss experience. A defensible historical loss rate follows cohorts of receivables from origination to final outcome (collected, written off or still outstanding) and measures the proportion of each ageing bucket that ultimately went uncollected. This is a flow-rate or roll-rate analysis, and it requires at least two to three years of monthly ageing data with write-off and recovery detail.
Where the data does not exist, the answer is not to guess. Reconstruct what is possible from the ledger, document the limitation, use external benchmarks for the missing pieces, and put in place the data capture that makes next year's model better. Auditors are generally receptive to a documented, improving approach; they are not receptive to an undocumented number.
Weakness three: forward-looking information that is invisible
IFRS 9 requires ECL to reflect reasonable and supportable information about future economic conditions. Many corporate matrices either ignore this or bury an unexplained uplift in the loss rates. Neither survives scrutiny. The adjustment should be explicit: a documented view of the factors that affect collection (oil price and government spending for entities dependent on public sector receivables, real estate transaction volumes for developers, consumer indicators for retail), a statement of whether conditions are expected to be better or worse than the historical period, and a quantified adjustment with rationale. For most corporates a qualitative scalar of plus or minus 10 to 30 percent on historical rates, reviewed annually, is proportionate. For larger books with meaningful data, a simple regression of loss rates against one or two macroeconomic variables is achievable.
What a finished model looks like
A model that passes cleanly has five components: a methodology paper of eight to twelve pages, a data file reconciled to the ledger, a calculation workbook with locked formulas and integrity checks, a sensitivity table showing the effect of changing the forward-looking scalar and the segmentation, and a one-page control sheet that management signs each period. The finance team runs it monthly from the ageing report in under an hour. The forward-looking assumptions are refreshed annually or on a trigger event.
None of this requires a complex system. It requires disciplined design once, and a repeatable process thereafter.