NPA classification and provisioning: a practical control guide

Distinguish delinquency, classification and impairment

Delinquency describes a payment position against contractual obligations. Prudential classification places an exposure into categories under the applicable regulatory framework. Accounting impairment measures expected or incurred loss according to the accounting framework that applies to the institution. Each answers a different question.

The Basel Committee's guidance on problem assets distinguishes prudential concepts and their relationship with accounting frameworks. ^[1] Indian institutions should use the relevant RBI directions and accounting requirements for the actual classification and measurement. Commercial-bank and NBFC directions must be assessed separately. ^{[4] [5]}

A useful data model retains these views independently and records how they relate. An accounting stage should not automatically overwrite a prudential category. Similarly, a collections label such as “promise to pay” should not change the underlying contractual overdue calculation.

Establish reliable contractual data

Classification begins with the loan agreement and approved changes to it. Maintain disbursement dates, repayment schedules, instalment components, due dates, interest terms and any authorised modifications. A schedule generated by a system should be traceable to the contract and its approved amendments.

Payment allocation is equally important. The system needs a controlled basis for applying receipts to outstanding components and periods. Reversals, failed payments, suspense balances and backdated postings can affect apparent arrears. Review how these items enter the classification process.

Keep a distinction between the date money was received, the accounting posting date and any value date used under the applicable treatment. Confusing them can shift delinquency across a reporting boundary. Exceptions should be investigated with evidence rather than resolved through an unexplained manual date change.

Control the classification population

Reconcile the exposure population to the institution's lending and accounting records. Include relevant facilities outside the main loan platform where required. A well-designed classification rule cannot identify risk in accounts excluded from its input.

Consider how borrower-level or connected exposure treatments apply under the relevant framework. The correct unit of assessment must be documented. Do not assume every rule operates independently at account level or that every facility shares the same treatment.

Test new products, acquired portfolios and migrated accounts separately. Legacy data may contain different date conventions, status codes or restructuring fields. A migration that preserves outstanding balances but loses repayment history can create serious classification uncertainty.

Use judgement without losing traceability

Some asset-quality decisions require information beyond payment ageing. Relevant evidence may include financial deterioration, insolvency proceedings, restructuring, disputed cash flows or other indicators specified by the applicable rules. Define who gathers this information and who approves its use.

Record the evidence and the reasoning behind a judgement. “Management assessment” is not enough when the decision materially affects classification or provisions. A reviewer should understand what information was considered, what alternatives were rejected and whether important facts remain uncertain.

The Basel credit-risk principles connect ongoing monitoring with credit administration and controls. ^[2] In operational terms, the classification process needs a reliable route for risk information from relationship teams, collections and legal functions, rather than relying exclusively on month-end data extracts.

A worked example: a receipt near period end

Consider a hypothetical borrower whose instalment remains unpaid in the loan system at period end. A receipt appears in a suspense account, but it has not been matched to the borrower. The collections team asks finance to treat the instalment as paid because the customer has supplied a transfer reference.

The reviewer needs more than the customer's statement. The team verifies the actual receipt, ownership, amount, timing and permitted allocation. It checks whether the payment was reversed or subject to another unresolved condition. The applicable rules and approved accounting treatment determine the reporting outcome.

The example illustrates why an automated overdue result may require investigation without becoming an invitation to arbitrary overrides. Once the facts are established, the institution should correct the underlying record, retain the evidence and assess whether other reports used the earlier position.

Keep prudential provisions and accounting allowances separate

A prudential provision and an accounting loss allowance may use different definitions, measurement methods and purposes. The relationship depends on the applicable regulatory and accounting framework. For an entity applying Ind AS, expected-credit-loss calculations require their own methodology and governance; they should not be assumed to equal a prudential provision.

Prepare a bridge between the relevant measures. Explain differences attributable to scope, classification, measurement, collateral treatment or other approved adjustments. Where a regulatory comparison, reserve or adjustment is required, document the applicable basis and the resulting treatment.

Avoid presenting the larger of two figures as automatically correct without examining the rules. Conservatism is not a substitute for a properly defined measurement. Both numbers need support, and the accounting entries and regulatory disclosures need a consistent explanation.

Validate models and parameters where they are used

If accounting impairment relies on models, identify the data, assumptions and parameters that drive the result. Depending on the method, these may include default estimates, recovery expectations, exposure profiles and forward-looking scenarios. Model output should be reviewed alongside its limitations.

Separate model changes from ordinary period-to-period movements. A higher allowance caused by a revised recovery assumption has a different explanation from one caused by newly deteriorated accounts. Maintain a movement analysis that allows users to distinguish these effects.

Management adjustments or overlays need a clear rationale, a defined population, avoidance of double counting and a review or release criterion. A long-standing overlay without a current explanation can hide a model weakness rather than address it.

Review upgrades, restructurings and write-offs carefully

Changes to favourable classifications require particular attention because they can reduce reported risk or provisions. Confirm that the relevant criteria have been met and retain supporting evidence. A short period of improved collections does not automatically establish that every required condition has been satisfied.

Restructuring and other modifications should retain their history. The institution needs to distinguish a commercial change from a concession associated with financial difficulty and apply the relevant rules. Deleting the previous schedule can make that assessment impossible.

Write-off, recovery action and legal enforceability are separate considerations. An accounting write-off should not be assumed to close collection or legal activity, and an active recovery process does not by itself justify retaining an unsupported carrying amount. Coordinate decisions across the responsible functions.

Reconcile the reporting story

The asset-quality narrative should align with the underlying data. Explain opening balances, additions, upgrades, recoveries, write-offs and other movements. Review whether the same population and definitions are used in management papers, regulatory returns and financial disclosures.

Analyse movements by meaningful segments. Product, vintage, geography, origination channel and borrower type can reveal emerging issues hidden by an overall ratio. Interpret these trends alongside changes in portfolio size and mix.

Pay attention to missing or low-quality data. A segment with incomplete repayment histories may appear unusually healthy because the system cannot identify deterioration. Report that limitation explicitly rather than treating the absence of a warning as evidence of good performance.

Examine collateral information critically

Collateral can affect recovery expectations or a prescribed treatment, but its existence does not make every exposure fully recoverable. Check ownership, enforceability, valuation date, realisation costs and the time needed to obtain proceeds where relevant to the measurement.

Keep these assumptions distinct from the contractual outstanding amount. A property valuation and an expected recovery are different quantities. If an allowance depends materially on collateral, the reviewer should understand the evidence supporting its value and the assumptions connecting that value to cash recovery.

Build an evidence-led review cycle

A practical review begins with population completeness, continues through classification and measurement and ends with approved entries and disclosures. Sample boundary cases, manual changes and unusual movements. Also test apparently routine cases to confirm that automated rules work as intended.

Track recurring defects to their source. If repayment schedules are repeatedly wrong, correcting classification each month leaves the root cause in place. Assign the system or process improvement to an owner and verify that it changes later results.

Reliable asset-quality reporting depends on preserving distinctions: contract versus cash receipt, arrears versus regulatory classification, and prudential provision versus accounting allowance. Keeping these distinctions visible allows banks and NBFCs to explain their results, challenge optimistic assumptions and respond earlier to deterioration.