IFRS 9 for Non-Banks: Credit Impairment Modelling When You Hold Financial Assets

Why Non-Bank Fintechs Need IFRS 9 ECL Models

The misconception that IFRS 9 impairment accounting only applies to banks is one of the most consequential accounting errors a non-bank fintech CFO can make. IFRS 9 applies to any entity that holds financial assets at amortised cost or fair value through other comprehensive income. For fintechs, this includes a surprisingly wide range of assets:

• Trade receivables: Any B2B business with invoice-based billing holds trade receivables. Under IFRS 9, these must be assessed for expected credit loss at every balance sheet date, regardless of whether any specific customer has missed a payment.
• BNPL loan books: Consumer finance businesses offering buy-now-pay-later credit hold the deferred payment obligations as financial assets. These are squarely in scope for the full three-stage ECL model.
• SME lending portfolios: Any embedded lending or direct lending fintech holds loan assets at amortised cost. The ECL model is the primary accounting framework for these portfolios and drives both the P&L charge and the balance sheet allowance.
• Invoice finance receivables: Invoice finance businesses that hold the assigned receivables on balance sheet have financial assets requiring ECL assessment.
• Intercompany loans: Loans from a parent company to a subsidiary, or between sister entities, are financial assets in the lender's accounts. Where there is any doubt about the subsidiary's ability to repay (including where the subsidiary is loss-making), an ECL must be calculated.
• Bond and money market investments: Cash management instruments held in backing pools (for example, by e-money issuers or stablecoin issuers) are financial assets in scope for IFRS 9 classification, measurement, and impairment.

The Three-Stage ECL Model

IFRS 9's ECL model classifies financial assets into three stages based on the degree of credit deterioration since initial recognition. The stage determines both the measurement horizon for the ECL (12 months or lifetime) and the interest revenue recognition basis.

The most operationally challenging element of the three-stage model is the Significant Increase in Credit Risk (SICR) assessment for Stage 1 to Stage 2 transfers. IFRS 9 requires entities to compare the lifetime probability of default at the reporting date to the lifetime PD at origination, and transfer to Stage 2 where the increase is significant. For a non-bank fintech without a sophisticated credit risk model, this assessment must be based on observable indicators: days past due (30 days past due creates a rebuttable presumption of SICR), covenant breaches, changes in credit score or external rating, and qualitative factors such as industry deterioration.

The Simplified Approach for Trade Receivables

For trade receivables without a significant financing component (which is the case for most B2B SaaS or services businesses with standard 30 to 60 day payment terms), IFRS 9 permits the use of the simplified approach. Under this approach, the entity does not need to track SICR and stage transfers: it simply applies lifetime ECL to all trade receivables from the date of initial recognition.

The most practical implementation of the simplified approach is the provision matrix: a table that applies historical loss rates to receivables grouped by ageing bucket. A typical provision matrix for a B2B fintech might look like the following:

The ECL rates in the provision matrix must reflect both historical loss rates and forward-looking information. The historical rates are derived from the company's own collections data: what percentage of invoices in each ageing bucket have historically been written off. The forward-looking adjustment reflects macroeconomic conditions at the reporting date: if a recession is expected, loss rates may be adjusted upward from the historical average. For a small company with limited historical data, external benchmarks from trade credit insurers or sector-specific default databases may be used as proxies.

ECL for BNPL and SME Lending Books

For BNPL receivables, SME loans, and invoice finance assets, the simplified approach is not available, and the full three-stage model applies. This requires the entity to construct a credit model with the following components:

• Probability of Default (PD): The probability that a borrower will default over the 12-month (Stage 1) or lifetime (Stage 2 and 3) measurement period. For a non-bank with limited proprietary data, PD can be derived from credit score-to-default rate tables (where credit score data is available), from external benchmark data for the relevant borrower segment, or from a simplified scorecard model.
• Loss Given Default (LGD): The proportion of the exposure that will not be recovered if a default occurs. For unsecured BNPL or SME loans, LGD is typically 70 to 90 per cent. For invoice finance, LGD depends on the quality of the underlying invoice and the debtor.
• Exposure at Default (EAD): The expected outstanding balance at the point of default. For a term loan, EAD is the outstanding principal at the measurement date. For a revolving facility, EAD must account for the possibility of further drawdowns before default.

ECL = PD × LGD × EAD, discounted at the effective interest rate of the instrument.

P&L and Cash Flow Statement Treatment

The ECL charge and the allowance account interact with the financial statements in ways that are frequently misunderstood, particularly in the cash flow statement.

P&L treatment: The movement in the ECL allowance (the increase or decrease in total expected credit loss from period to period) is recognised as a credit impairment charge (or reversal) in the P&L. A deteriorating portfolio increases the allowance, generating a charge. An improving portfolio reduces the allowance, generating a credit. The gross carrying amount of the financial asset remains unchanged on the balance sheet; it is the allowance account that changes.

Balance sheet treatment: The financial asset is shown at its gross carrying amount less the ECL allowance, giving a net carrying amount that represents the CFO's best estimate of the recoverable value. For a BNPL loan book with a gross value of £5m and an ECL allowance of £300,000, the net carrying amount is £4.7m.

Cash flow statement treatment: The ECL charge is a non-cash item: no cash has left the business when the provision is raised. It must therefore be added back in the operating activities section of an indirect method cash flow statement, alongside depreciation and amortisation. This is one of the most commonly omitted line items in growth-stage cash flow statements, resulting in understated operating cash flow.

Building an ECL Model Without Banking-Grade Data

The most significant practical challenge for non-bank fintechs is that ECL modelling is designed with banks in mind, and banks have decades of historical credit data across economic cycles. A two-year-old BNPL business has, at most, two years of performance data across a single economic environment. Several approaches can address this data limitation:

• External benchmark data: Trade credit risk data providers (Dun and Bradstreet, Experian Business, Creditsafe) publish default rate data by industry sector and company size. These can provide calibration points for the provision matrix where internal data is insufficient.
• Proxy PD and LGD data: Published academic research and central bank publications provide industry-level PD estimates. The Bank of England's Financial Stability Report includes sector-level default rate estimates that can calibrate a lending book's PD model.
• Conservative adjustment for data immaturity: Where internal data covers fewer than 3 to 4 economic cycles, auditors and regulators expect an upward adjustment to the historically-derived rates to reflect the possibility that the data was collected during a benign economic period. This conservatism adjustment is judgment-based and should be documented.
• Vintage analysis: Even with limited data, breaking the portfolio into origination cohorts (vintages) and tracking the payment behaviour of each cohort over time provides valuable insight into the maturation pattern of default risk in the portfolio.

Disclosure Requirements

IFRS 7, which accompanies IFRS 9, requires extensive quantitative and qualitative disclosures about credit risk. For a non-bank fintech, the minimum disclosures that must appear in the annual accounts are:

• A description of the inputs, assumptions, and estimation techniques used in the ECL model
• A reconciliation of the opening and closing ECL allowance for each class of financial asset, showing the movements in the period
• A credit risk exposure table showing the gross carrying amount at each stage (Stage 1, 2, and 3) by class of financial asset
• Sensitivity analysis: the effect of a reasonable change in one or more assumptions on the ECL amount
• Qualitative description of any forward-looking economic assumptions applied and how they have been reflected in the model

These disclosures are frequently inadequate in growth-stage company accounts, partly because auditors at the early stages may not focus intensively on credit risk disclosure, and partly because the ECL model is often maintained in a spreadsheet that makes disclosure preparation laborious. As the loan book or trade receivables base grows, investing in a proper ECL modelling tool that can generate audit-ready disclosures automatically becomes worthwhile.

Key Takeaways

• IFRS 9 ECL modelling applies to any fintech holding financial assets: trade receivables, BNPL books, SME loans, invoice finance assets, intercompany loans, and investment-grade debt securities.
• The three-stage model classifies assets as Stage 1 (12-month ECL), Stage 2 (lifetime ECL, significant credit deterioration), or Stage 3 (credit-impaired, lifetime ECL, credit-adjusted EIR for interest). Stage transfers are triggered by the SICR assessment.
• For trade receivables, the simplified approach (provision matrix) is permitted. The matrix applies historically-derived default rates, adjusted for forward-looking information, to receivables grouped by ageing bucket.
• For BNPL and lending books, the full three-stage model applies. ECL = PD × LGD × EAD, discounted at the effective interest rate. PD and LGD can be estimated from external benchmarks where internal data is limited.
• The ECL charge is a non-cash item: it must be added back in the operating activities section of the cash flow statement under the indirect method. Omitting this adjustment is a common error.
• Sensitivity analysis and a reconciliation of the allowance account are required IFRS 7 disclosures. These must appear in the annual accounts and must be updated every year to reflect current conditions.
• A documented, consistently applied methodology is more important in the audit conversation than the specific numerical outcome. The CFO must be able to explain the model clearly and demonstrate that it has been applied consistently.