The expected credit loss model in IFRS 9 asks a question older accounting never did: not "which of our receivables has gone bad?" but "how much do we expect to lose on everything we are owed, starting today?" Provisions are booked from day one of every loan and receivable, sized on probability-weighted, forward-looking estimates rather than waiting for a default to prove itself. Introduced after the financial crisis exposed the too-little-too-late flaw in the old incurred-loss rules, ECL is now the daily reality for every IFRS reporter — and a source of persistent confusion for UK companies on FRS 102, who keep being told they need it when they do not.
The three-stage general model
For loans, debt investments and similar assets, IFRS 9 tracks credit quality through three stages:
- Stage 1 — performing. On day one, recognise 12-month ECL: the losses expected from defaults possible within the next year. Interest income runs on the gross carrying amount;
- Stage 2 — significant increase in credit risk. When credit risk has risen significantly since origination (not merely become bad in absolute terms), the allowance steps up to lifetime ECL — all losses expected over the asset's remaining life. Interest still on gross. This cliff between stages 1 and 2 is where most of the model's judgement, and most of its earnings volatility, lives;
- Stage 3 — credit-impaired. Default has effectively arrived; lifetime ECL continues, and interest income switches to the net amount — the carrying value after the allowance.
Two rebuttable presumptions anchor the judgement: 30 days past due suggests a significant increase in credit risk; 90 days past due suggests default. The arithmetic underneath is the familiar credit triad — probability of default × loss given default × exposure at default — computed across multiple forward-looking macroeconomic scenarios and probability-weighted. That scenario requirement is the model's post-crisis signature: a base case alone is non-compliant, and the downside scenarios are exactly what drove the large, early provisions banks booked in past shocks.
The simplified approach: what most companies actually use
For trade receivables and contract assets, IFRS 9 skips the staging entirely: lifetime ECL, always. The accepted practical tool is the provision matrix — receivables bucketed by age, historical loss rates applied per bucket, then adjusted for what forward-looking information says about the year ahead. A wholesaler might run 0.5% on current balances, stepping to 3%, 8% and 25% through the ageing — with the rates flexed when its customer sector visibly deteriorates.
Done honestly, this is an afternoon's modelling refreshed annually, and it produces something managerially useful: a live, quantified view of collection risk by customer segment. Done as a copied-forward spreadsheet with 2018's loss rates, it fails both the auditor and the point. The common defects are consistent — no forward-looking adjustment at all, related-party and intercompany balances excluded from scope (they are in scope, and intercompany ECL is a recurring audit finding in groups), and no documentation of why the rates are what they are.
Who this applies to in the UK — and who it doesn't
ECL binds entities reporting under UK-adopted IFRS: listed groups, banks and building societies above all, plus IFRS-reporting subsidiaries. FRS 102's own model is not ECL. The FRC considered importing it during the recent periodic review and deliberately declined — UK GAAP's default remains incurred loss even in the amended standard taking effect from 2026. One nuance keeps this from being absolute: FRS 102 paragraph 11.2 offers an accounting policy choice to apply IFRS 9's recognition and measurement instead, so an FRS 102 company can elect into ECL — typically a subsidiary aligning with an IFRS group. Absent that election, the FRS 102 question remains objective evidence of impairment, and "expected credit loss provision" language in default-policy FRS 102 accounts is importing a framework the company has not actually adopted. Where it does bite private companies is group reporting: subsidiaries of IFRS parents supply ECL numbers for consolidation packs regardless of their local GAAP, which is how a provision matrix lands on a Midlands finance manager's desk each December.
The standard itself has kept moving at the edges — 2024 amendments effective from 2026 tidy classification questions (ESG-linked loan features, electronic payment timing) — but the post-implementation review left the impairment model intact: ECL, in its current shape, is settled law.
Making it defensible
The audit-season checklist we run with clients: loss-rate history refreshed against actual write-off experience; a documented forward-looking overlay tied to something observable (sector insolvency data, customer concentration changes); intercompany and related-party balances assessed, not assumed benign; stage-transfer criteria written down for anything under the general model; and disclosures that reconcile the allowance movement rather than asserting it. None of this is heavy for a trading business — the whole point of the simplified approach is proportionality — but it has to be visibly done.
Acumon prepares and audits ECL positions across the size range — IFRS accounts and group reporting packs, financial services audit where the general model runs at full complexity, and financial reporting support for FRS 102 groups feeding IFRS parents. If your provision matrix predates the last economic cycle, it is due a rebuild before the auditors say so.