Every credit provider in South Africa is required to conduct an affordability assessment before granting credit. That obligation sits in Section 81 of the National Credit Act and is given operational teeth by Regulation 23A of the National Credit Regulations. Most lenders meet it with a payslip, a declaration form, and a bureau pull. The regulation demands more than that.
If you are a compliance officer, risk manager, or credit operations lead, this is the reference piece. Everything else we write about NCA affordability links back here.
What Reg 23A actually says
Regulation 23A was introduced through Government Gazette 38557, published 13 March 2015, as part of the Affordability Assessment Regulations under the National Credit Regulator framework. It replaced the earlier, more loosely defined affordability requirements with a structured formula that every registered credit provider must follow.
The formula is straightforward in concept:
Discretionary income = Gross income − Statutory deductions − Existing financial obligations − Living expenses
Each component has specific requirements:
- Gross income (sub-reg 3): The consumer's total income from all sources, before deductions. The regulation requires the credit provider to take reasonable steps to validate declared income. For salaried consumers, this typically means a payslip or bank statement. For self-employed or informally employed consumers, three months of bank statements are the primary validation mechanism.
- Statutory deductions (sub-reg 4): PAYE, UIF, and any other deductions required by law. These are deterministic once gross income is known and can be calculated from payslip data or estimated from the net deposit on a bank statement.
- Existing financial obligations (sub-reg 5): All existing debt commitments, including credit agreements, maintenance orders, and other fixed obligations. The regulation explicitly requires the credit provider to obtain this information from the consumer and from a credit bureau. It does not say the bureau is the only source.
- Living expenses (sub-reg 6–10): The consumer's actual monthly living expenses, subject to minimum norms prescribed by the NCR. This is where the regulation becomes materially more demanding than most lenders realise.
The credit provider must then determine whether the consumer has sufficient discretionary income to service the proposed credit obligation without becoming over-indebted. If discretionary income is negative or insufficient, the credit should not be granted.
The minimum expense norms problem
Sub-regulation 10 of Reg 23A prescribes minimum living expense norms, published and updated by the NCR. These norms set a floor: the credit provider may not use a living expense figure lower than the prescribed minimum for the consumer's income band.
The norms were designed as a safety net. They exist to prevent lenders from accepting a consumer's declaration that they spend R500 per month on food when earning R25,000. The problem is that many credit providers have turned the floor into the ceiling. They use the NCR minimums as the expense figure rather than as the lowest permissible figure.
Consider a hypothetical consumer earning R28,000 gross. If the NCR minimum expense norms for that income band prescribe approximately R8,400 in total living expenses, that covers food, transport, housing costs (if applicable), education, and basic necessities at a subsistence level. Run the same consumer's bank statement through a categorisation engine and the picture changes: actual grocery spend R6,200, transport and fuel R3,800, insurance premiums R2,100, discretionary spending R2,400, giving total actual living expenses of R14,500, nearly double the prescribed minimum. (These figures are illustrative; actual norms vary by income band and are updated periodically by the NCR.)
The gap between norms and reality is not a marginal discrepancy. In practice, actual categorised expenses frequently exceed NCR minimums by a material margin, in some income bands by a factor of two or more, depending on geography and household composition. Urban consumers in Gauteng and the Western Cape are likely to show the widest gaps, given higher transport, housing, and food costs in those metros.
A credit provider that uses only the prescribed minimums will calculate a discretionary income figure that is materially higher than the consumer's actual position. That phantom headroom creates room for credit that should never have been extended. When the consumer defaults, the lender's affordability assessment is not just wrong; it is indefensible in the context of a reckless lending challenge under Section 80 of the NCA. The CDH legal analysis from November 2023 makes clear that an assessment conducted on inadequate inputs carries the same reckless lending risk as no assessment at all.
How bank statement affordability works
Bank statement affordability assessment replaces declared and assumed inputs with observed ones. Instead of asking the consumer what they earn and what they spend, you read it from the statement. Instead of relying solely on the bureau for obligations, you verify them against actual debit order activity.
The process maps directly to the Reg 23A formula:
Income: verified, not declared
Salary deposits, grant income, rental income, freelance receipts: the bank statement shows every credit that hits the account. For consumers with a single employer, the salary identification is straightforward: a recurring credit of consistent value from an identifiable source. For consumers with irregular or multiple income streams, three months of statement data establishes a pattern that declared income alone cannot. The statement also reveals income the consumer may not declare: a second job deposited periodically, a side business with irregular receipts, or a maintenance payment.
Expenses: categorised and norm-floored
Every debit transaction on the statement is categorised against standard expense categories that map to the Reg 23A living expense framework: groceries, transport, housing, utilities, education, medical, insurance, and discretionary spend. The categorisation is transaction-level. Each individual purchase or payment is assigned to a category based on the merchant name, transaction code, and payment pattern.
The critical mechanism is the norm floor. For each expense category, the system compares the consumer's actual spend against the NCR prescribed minimum for that category. The higher of the two figures is used. If a consumer's actual grocery spend is R3,200 but the norm prescribes R2,800, the actual figure stands. If their declared education spend is R0 but the norm prescribes R400, the norm figure is applied. This satisfies the regulation's floor requirement while capturing the reality that actual expenses commonly exceed it.
Obligations detected from debit orders
The bureau tells you what the consumer has disclosed to previous credit providers. It does not tell you about the R2,500 per month they send to a family member via EFT, the R800 stokvel contribution debited on the first of every month, or the informal loan repayment going to a colleague's bank account every payday.
Bank statement analysis identifies recurring debits (debit orders, stop orders, and regular EFT payments) and flags them as potential obligations. Known credit provider names are matched against the bureau data for cross-referencing. Unmatched recurring debits are flagged as additional obligations that the bureau does not capture. This is not a replacement for the bureau pull; it is a verification and augmentation layer that catches what bureau data structurally cannot.
The resulting discretionary income figure
With verified income, categorised expenses (norm-floored), and detected obligations all derived from the same data source, the discretionary income calculation produces a figure grounded in observed behaviour rather than declarations and assumptions. That figure is typically lower than the payslip-and-norms calculation, sometimes substantially so. It is also substantially more defensible.
The audit trail
Reg 23A does not only require the credit provider to calculate affordability correctly. It requires them to keep records that demonstrate the assessment was conducted and that the inputs were reasonable. The NCA's record-keeping obligations (Section 62 and Regulations 18 and 19) require retention of assessment records for the duration of the credit agreement. The NCR's compliance monitoring process evaluates not just whether an assessment was done but how it was done and whether the inputs can be traced.
Bank statement affordability creates a structural advantage here. Every number in the assessment traces back to a specific transaction on a specific statement.
Monthly income: R28,000. Source: salary deposit from XYZ Manufacturers (Pty) Ltd, reference SAL-2026-05, dated 25 May 2026, visible on page 3 of the statement. Grocery expenses: R6,200. Source: 47 individual transactions at Pick n Pay, Checkers, Woolworths, and Shoprite across the assessment period. Loan repayment: R3,200. Source: debit order to a registered credit provider, debited on the 1st of each month for three consecutive months.
When the NCR auditor reviews the file, they can follow the thread from the affordability output to the raw data. There is no gap between "we assessed the consumer's income as R28,000" and the evidence supporting that figure. Compare this to a payslip-and-declaration model where the auditor sees a declared income figure and a self-reported expense list with no independent verification. The first assessment is traceable. The second requires trust.
For a deeper look at how this audit trail specifically addresses the compliance gaps facing micro-lenders, see our article on micro-lenders and the Reg 23A audit trail.
What payslip and declaration miss
The payslip-and-declaration model captures a specific data point (employer-confirmed gross income) with reasonable accuracy. The problem is what it does not capture:
- Second and third income sources: A consumer earning R18,000 from their primary job and R6,000 from weekend freelance work will only declare the primary income if they think it improves their approval odds. Conversely, they may overstate income by including irregular receipts they cannot sustain.
- Undeclared debt obligations: Family loans, stokvel commitments, informal lending circle repayments, and private vehicle purchase agreements do not appear on the bureau. They appear on the bank statement as recurring debits.
- Lifestyle inflation: A consumer who was approved for a vehicle loan six months ago at norms-based expenses may now be spending significantly more on fuel, insurance, and maintenance than the norms assumed. The statement reflects the current expense profile, not the historical one.
- Gambling and high-risk discretionary spend: Regular transfers to betting platforms or casino withdrawals indicate a spending pattern that materially affects affordability but will never appear on a declaration form.
- Cash withdrawals: Persistent large cash withdrawals indicate undisclosed expenses or cash-based debt repayments invisible to both the bureau and the payslip.
Consumers routinely underestimate their own expenses, not by rounding error, but structurally. The bank statement corrects for this without asking. For more on how behavioural signals contribute to credit scoring, see our overview of behavioural credit scoring and bank statement scoring.
Draft amendments: August 2025 and beyond
In August 2025, the Department of Trade, Industry and Competition published draft amendments to the National Credit Regulations for public comment. Several of the proposed changes directly affect how credit providers must conduct affordability assessments.
The key amendments relevant to Reg 23A include:
- Tightened expense norms: Upward revision of minimum expense norms across multiple income bands, reflecting updated household expenditure data. Credit providers using norms-only calculations will see calculated discretionary income fall, potentially below zero for borderline approvals that currently pass.
- Expanded record-keeping under Reg 18 and 19: More granular requirements for what records must be retained and for how long. The NCR wants to see the workings, not just the answer.
- Stricter validation of income: Self-declaration of income, without independent verification, may no longer be sufficient for certain credit categories. Bank statement verification is explicitly referenced as an acceptable method.
The comment period closed 12 September 2025. Promulgation is expected during 2026. Credit providers already conducting bank statement-based affordability assessments are well-positioned for these changes without material process overhaul. Those relying on payslip-and-declaration models face a more significant adjustment.
Practical implementation with AffyScore
AffyScore's affordability output is structured to satisfy Reg 23A evidence requirements. When a bank statement is processed, the decision pack returns:
- Income (verified): Salary and other income sources identified from statement deposits, with source attribution and consistency scoring across the assessment period.
- Expenses (categorised, norm-floored): Transaction-level categorisation mapped to Reg 23A expense categories. Each category shows both the actual spend and the applicable NCR minimum, with the higher figure used in the affordability calculation. The thresholds applied are AffyScore defaults based on publicly available NCR norms; credit providers can configure their own floors where their product policies require stricter standards.
- Existing obligations (detected): Recurring debit orders and instalment payments identified from transaction patterns, cross-referenced against known credit provider names. Bureau-matched and unmatched obligations are reported separately.
- Discretionary income: The resulting figure after all verified inputs are applied to the Reg 23A formula.
- Affordability ratio: Proposed instalment as a percentage of discretionary income, giving the credit provider a single metric for decisioning thresholds.
Every figure in the output links to the underlying transactions. The decision pack is designed to be stored as the affordability assessment record, supporting the record-keeping requirements without requiring manual compilation. An NCR auditor reviewing the file sees the formula, the inputs, the source transactions, and the result, all in one document.
The extraction and categorisation runs via API. The credit provider submits the bank statement PDF and receives the structured JSON decision pack. For lenders processing high application volumes, the assessment becomes part of the automated pipeline rather than a bottleneck.
The affordability output is one component of the broader AffyScore extraction. The same bank statement also produces a behavioural credit score based on spending patterns, income stability, and financial behaviour signals. The affordability assessment answers the regulatory question: can this consumer afford the proposed credit? The behavioural score answers the commercial question: is this consumer likely to repay? Together, they give the credit provider both compliance coverage and risk intelligence from a single document.
