Lending & Credit

The moment a bank puts its own money at risk

A young couple sits at a kitchen table on a Sunday evening, staring at a mortgage offer on a laptop. They have found the house they want. The price is just above what they have saved. The bank's offer sits between them and the keys. To them, the loan is a single number: an amount, a rate, a monthly payment, a term. To the bank, that same loan is the moment the bank puts its own money at risk on the strength of a promise that the couple will repay.

Lending is the discipline of putting money at risk on the strength of a promise to repay. It is a major source of income and risk for many banks. The contribution differs by business model; transaction, custody, wealth and other services also have distinct economic purposes.

This chapter is the complete study of lending and credit in consumer and business banking. It covers what credit is, how a bank decides to lend, how a loan is structured and priced, how it is repaid, how it is monitored, how it is collected when it goes wrong, and how it is accounted for in the bank's books. It begins at first principles and ends at the architecture of a modern lending estate.

Learning objectives

By the end of this chapter, you will be able to:

• Define credit and lending, and distinguish them from deposits, payments, and investments.
• Describe the major lending products in consumer and business banking.
• Explain the five Cs of credit and how they shape the lending decision.
• Walk through the credit decisioning process: application, scoring, underwriting, decision, and offer.
• Describe loan structure: principal, interest, term, amortisation, security, and covenants.
• Explain interest calculation on loans: simple, amortised, and reducing balance.
• Distinguish product-level pricing from risk-based pricing.
• Outline the loan lifecycle: origination, disbursement, repayment, monitoring, and closure.
• Explain the stages of delinquency and default, and how the bank manages each.
• Describe credit risk management: exposure, probability of default, loss given default, expected loss, and provisioning.
• Outline the systems and architecture that support lending, including the data model and integrations.
• Apply the chapter content from the perspectives of business analyst, solution architect, developer, tester, operations, and production support.
• Identify common implementation mistakes, articulate best practices, and answer interview questions on lending and credit.

What credit and lending are

Credit is the trust that a borrower will repay. Lending is the act of providing money today against a promise to repay tomorrow. The two ideas are inseparable: lending is impossible without credit; credit is built through lending.

Three ideas sit at the heart of this chapter.

Credit is the bank's assessment of a borrower's willingness and ability to repay. It is expressed as a score, a rating, or a judgement, and it drives every lending decision.

A loan is a contract in which the bank gives money today and the borrower agrees to repay it, with interest, on a schedule. The loan defines the principal (the amount borrowed), the interest (the price of the money), the term (how long), the schedule (when payments are due), and the security (what the bank can take if the borrower does not repay).

Lending is the discipline of originating, structuring, pricing, disbursing, monitoring, and collecting loans. It is a cycle that begins when a borrower applies and ends when the loan is fully repaid or written off.

These three ideas — credit, loan, lending — together make up the subject of this chapter.

Why lending matters

Lending matters for four reasons that compound each other.

Revenue. Interest income is the largest single source of revenue for most banks. A bank that cannot lend is a bank that cannot earn.

Economic function. Lending lets households buy homes, students attend university, and businesses invest. Without credit, economic activity would be a fraction of what it is.

Risk concentration. Credit risk is the largest risk most banks carry. A bank that lends badly loses money through defaults; a bank that lends well earns a margin that funds everything else.

Regulation. Lending is regulated on both sides: who can be granted credit, what can be charged, how defaults are treated, and how much capital the bank must hold against the loan book. Credit regulation is extensive because credit harm is widespread.

The five Cs of credit

The five Cs are the timeless framework that underwrites every lending decision, from a $500 credit card to a $500 million syndicated loan. Understanding them is the foundation of credit literacy.

The five Cs are not a checklist; they are a lens. Every loan is assessed through all five, with the weight on each varying by product. A mortgage weighs collateral heavily. A credit card weighs character and capacity. A working capital loan to a business weighs capacity and conditions. The combination produces a judgement: is this borrower likely to repay, and if not, what can the bank recover?

Major lending products

Banks lend across many products, each engineered for a specific purpose, borrower, and risk profile.

Consumer lending products

Consumer lending products

The two markets share the same underlying machinery — credit assessment, structuring, disbursement, monitoring, collection — but the products and the depth of analysis differ. Consumer lending is high-volume, low-touch, rules-driven; business lending is lower-volume, higher-touch, analyst-driven.

The credit decisioning process

Every loan begins with an application and ends with a decision. The path between them is the credit decisioning process.

The whole flow may complete in seconds for a credit card application and weeks for a large commercial loan. The depth of underwriting scales with the loan size and complexity.

Credit scoring in depth

Scoring turns the five Cs into a number. The number predicts the probability that a borrower will default over a defined horizon, typically 12 to 24 months.

The score is typically a logistic regression or gradient boosting model trained on the bank's historical data, calibrated against bureau data. The output is a probability of default (PD), which is mapped to a score band.

Adverse action notices are a regulatory requirement in many jurisdictions: a declined applicant must be told the principal reasons for the decline and the bureau whose data was used.

Risk-based pricing

Risk-based pricing ties the interest rate to the borrower's risk. A lower-risk borrower pays a lower rate; a higher-risk borrower pays a higher rate. The relationship is not arbitrary; it is calibrated so that the higher rate compensates the bank for the higher expected loss.

Risk-based pricing lets the bank serve a wider range of borrowers while keeping its margin stable. It is the dominant model in consumer credit cards, personal loans, and auto loans, and it is increasingly common in mortgages and small business loans.

Loan structure

A loan is defined by its structure. The structure shapes the cash flows, the risk, and the accounting.

Loan structure is a design exercise. A long term reduces the monthly payment but increases total interest. An interest-only structure keeps payments low early but defers principal repayment. A bullet structure repays the entire principal at maturity, common in short-term business loans. Each structure serves a borrower need, but each carries risk that the bank must price for.

Interest calculation on loans

Interest is the price the borrower pays for the use of the bank's money. The way it is calculated shapes the economics of the loan and the customer's experience.

Simple interest

Interest is calculated on the outstanding principal: interest = principal × rate × time. Simple interest is used for short-term loans and some instalment loans.

Reducing balance interest

Interest is calculated on the outstanding principal, which falls as the borrower repays. Each payment covers the interest due plus a portion of principal; the next period's interest is calculated on the lower balance. Reducing balance is the standard for mortgages, auto loans, and personal loans.

Amortised interest

The loan is structured so that a fixed monthly payment covers both interest and principal, with the interest share falling and the principal share rising over the term. Amortisation is the most common structure for long-term consumer loans.

This simplified schedule assumes principal of $100,000, 180 monthly payments and a fixed 0.4% monthly rate, with no fees or missed payments. The annuity payment is 100,000 × 0.004 / [1 − (1.004)^−180] = 780.414159…. The table calculates with unrounded intermediate values and displays cents. Actual cent-rounded instalments require a small final-payment adjustment. A production schedule uses the contract’s accrual basis, dates, calendars, rounding and payment allocation; daily accrual is not interchangeable with this simplified monthly model.

Day-count conventions

Day-count conventions determine how time is measured. The three common conventions are Actual/365, Actual/360, and 30/360. The convention is part of the loan contract and must be applied consistently for the life of the loan.

The loan lifecycle

A loan moves through a defined lifecycle, each stage with its own systems, controls, and accounting.

Each stage has its own systems and its own controls. Origination is rules-driven and fast; monitoring and collections are exception-driven and human.

Repayment and servicing

Once a loan is disbursed, the bank must collect what it is owed. Repayment is governed by the schedule, which is generated at origination.

Servicing also covers customer queries, payment holidays, modifications, and early settlement. Each is a controlled change to the loan, with audit and accounting impact.

Delinquency, default, and collections

Not every loan repays on time. Delinquency is the failure to pay on schedule; default is the more serious state where the bank concludes the borrower will not repay as agreed.

The following is an illustrative collections policy, not a universal default definition. Contractual, accounting and prudential definitions differ; unlikely-to-pay indicators can establish default before a 90-day counter. Apply the relevant product, jurisdiction and permitted forbearance rules.

Collections is a structured process, not a single action. Early delinquency is often accidental (the borrower forgot, or had a temporary cash flow issue) and is cured with a reminder. Serious delinquency often reflects genuine hardship and may be addressed through a modified payment plan. Default triggers assessment of recovery options, customer circumstances, legal rights and permitted forbearance. Enforcement is governed by local procedure and is not automatically authorised by a delinquency counter.

Detailed collections and recoveries practice is covered in its own chapter. The lending discipline's role is to originate well, structure well, and price well, so that the minimum volume of loans ever reaches collections.

Credit risk management

Credit risk is the risk that the borrower fails to repay. Managing it is the central discipline of lending.

Probability of default (PD)

PD is the likelihood that a borrower will default over a defined horizon. It is derived from the score, from behavioural data, and from macroeconomic factors. PD is expressed as a percentage: a 2 percent PD means the bank expects 2 out of 100 similar borrowers to default.

Loss given default (LGD)

LGD is the proportion of the exposure the bank expects to lose if the borrower defaults, after recovering what it can from collateral and guarantees. LGD is expressed as a percentage: a 40 percent LGD means the bank expects to recover 60 percent of the outstanding balance.

Exposure at default (EAD)

EAD is the amount the bank expects to be owed at the point of default. For a term loan, it is close to the outstanding balance. For a revolving facility, it is the expected drawdown at default, which may be higher than the current balance.

Expected loss (EL)

EL is the product of the three: EL = PD × LGD × EAD. Expected loss is what the bank prices for. The interest margin must cover the expected loss, the cost of funds, operating costs, and the capital charge.

Provisioning

Provisioning is the accounting recognition of expected loss. The bank holds a provision (a reserve) against the loan, which absorbs the loss when it materialises. Under the general IFRS 9 impairment approach, Stage 1 uses 12-month expected credit losses and Stages 2 and 3 use lifetime expected credit losses. Twelve-month ECL is the lifetime loss associated with default events possible within the next 12 months, not merely cash shortfalls in that period. Other approaches and standards differ; US CECL is not an identical three-stage model.

Provisioning is one of the most consequential accounting decisions a bank makes. Over-provisioning reduces current profits; under-provisioning stores up future losses. The discipline is to provision consistently, transparently, and with evidence.

Pricing a loan

Pricing a loan is the art of covering costs and earning a return, while remaining competitive.

Risk-based pricing adjusts the margin per borrower, so the bank earns its target return regardless of risk. A bank that prices too low loses money on defaults; a bank that prices too high loses customers to competitors.

Secured vs unsecured lending

Lending is broadly classified by whether it is backed by collateral.

Security reduces LGD: if the borrower defaults, the bank can seize and sell the collateral, recovering part or all of the exposure. The lower the LGD, the lower the expected loss, and the lower the rate the bank needs to charge.

Fixed vs floating rate loans

A loan's rate can be fixed for its life or can float with a reference rate.

The bank manages the interest-rate risk created by fixed-rate lending through its treasury function, often using interest-rate swaps to hedge. The hedging cost is factored into the loan's pricing.

Consumer vs business lending

Consumer and business lending share the same machinery but differ in depth and complexity.

The consumer side is an industrial process; the business side is a relationship and a judgement. Both must operate correctly for the bank to earn its lending margin.

Systems and architecture

A modern lending estate is a constellation of systems.

In a modern architecture, these systems communicate through APIs and events. The LOS publishes LoanBooked; servicing subscribes; the risk engine consumes behavioural data; the GL receives accounting entries; reporting aggregates all.

Data model

A simplified data model for lending looks like this.

Each entity is normalised, audited, and effective-dated where appropriate. The loan record is the spine that connects every other entity.

Business rules and validation

Lending operates under hundreds of business rules. The most important are:

• A loan cannot be disbursed without a booked loan record and a valid funding source.
• A payoff quote includes the applicable principal, accrued interest, fees and other contractual amounts at its value date. Handle excess receipts through the governed overpayment credit, refund or suspense process rather than treating principal alone as the maximum receipt.
• An interest accrual must use the rate in force on the accrual date.
• An impairment allowance must follow the applicable accounting model and forward-looking evidence; IFRS 9 staging and US CECL are distinct models.
• A modification must be authorised and audited, with accounting impact recognised.
• A collateral valuation must be current; stale valuations produce inaccurate LGD.
• A covenant breach must trigger a defined escalation.
• A write-off must be authorised at the appropriate level.
• A repayment reversal must reverse the components actually posted by the original repayment, which may include principal, interest and fees; preserve the original allocation and reconcile the linked correction.
• A restructuring requires modification/derecognition assessment under the applicable accounting standard, required remeasurement and impairment treatment, and preserved original-loan history. A changed contract does not universally require creating a new loan record.

Validation rules are encoded in the core lending system, the decisioning engine, and the risk engine. They are tested exhaustively, because a defect in lending affects every loan of the product.

Risks and controls

Lending carries several categories of risk.

Each risk has a primary owner, a control set, and a portfolio-level reporting line. Credit risk and concentration risk are the two that dominate the lending conversation.

Exceptions and operational reality

Lending operations encounter exceptions daily.

Each exception has a defined process, an authorisation level, and an audit trail. Exceptions are not failures of the system; they are part of the operational reality the system must support.

Reconciliation and reporting

Lending operations close the books daily and reconcile at multiple levels.

A break at any level is investigated. Reconciliation is the daily proof that the lending books are correct.

Business Analyst perspective

A business analyst on lending focuses on translating commercial intent into rules.

• Requirement capture. Convert a lending proposition into product configuration, scorecard rules, underwriting policy, and integration requirements.
• Edge case definition. Specify behaviour at rate boundaries, term boundaries, LTV thresholds, and score thresholds.
• Traceability. Trace every lending rule to a configured attribute and a test case.
• Impact assessment. When a rate or policy changes, assess which products, segments, and loans are affected.
• Disclosure alignment. Ensure the disclosures match the configured rates and fees.

The BA's craft is precision, because a vaguely specified rule becomes a defect that affects every loan of the product.

Solution Architect perspective

An architect designing a lending estate thinks about scale, correctness, and evolution.

• Loan modelling. Decide how loans, schedules, drawdowns, and repayments are modelled. Each decision affects performance, extensibility, and integration.
• Decisioning as a service. Separate the decisioning engine so it can evolve independently and serve multiple products.
• Performance. Origination may be high-volume (credit cards); servicing must handle large back-books; collections must scale in a downturn.
• Consistency. Decide where strong consistency is required (disbursement, repayment) and where eventual consistency is acceptable (behavioural scoring updates).
• Event model. Define LoanBooked, PaymentReceived, DelinquencyTriggered, ProvisionUpdated events.
• Change management. Scorecards, rate plans, and policies change. The architecture must support versioned models and parallel running during migrations.

The architect's hardest job is keeping the model simple enough to operate and rich enough to support the bank's lending ambitions for decades.

Developer perspective

A developer implementing lending systems cares about correctness and precision.

• Money handling. Use decimal types or integer minor units; never floating-point. Round only at defined points.

• Idempotency. Disbursement and repayment APIs must be idempotent, so retries do not double-post.

• Schedule generation. The amortisation schedule is the contract; generate it carefully, test it exhaustively.

• Audit. Every state change (balance, status, rate, provision) writes to an append-only audit log.

• Tests. Unit tests cover interest calculation, schedule generation, and rule enforcement; integration tests cover end-to-end flows (origination, disbursement, repayment, default, closure); property tests cover edge cases (zero balance, negative rate, leap years).

Tester perspective

A tester validating lending attacks the boundaries and the rules.

• Interest calculation. Compare the system's calculation to an independent calculation across a range of principals, rates, and periods.
• Amortisation schedule. Verify the schedule sums to the principal plus total interest; verify the final balance is zero.
• Rule enforcement. Can a received repayment exceed the payoff amount? If so, allocate the amount due and handle surplus through the documented refund or suspense process; do not over-reduce principal. Can a loan be disbursed without a funding source? (No.) Test every rule.
• Provisioning. For each IFRS 9 stage, verify the provision reflects forward-looking expected loss.
• Reconciliation. After a synthetic day, verify the GL reconciles.
• Performance. At peak load, do disbursements and repayments complete within the SLA?

Operations perspective

Operations runs the loan book day to day.

• Daily close. Confirm all disbursements posted, all repayments collected, all accruals ran, all reconciliations passed.
• Exception handling. Work the queue of missed payments, covenant breaches, collateral impairments, and modifications.
• Collections. Work the delinquency queue; apply the right strategy within the SLA.
• Monitoring. Watch portfolio health: delinquency rates, default rates, provision coverage, concentration.
• Reporting. Produce portfolio reports for risk, finance, and regulators.

Production support scenarios

Production support handles incidents on the lending estate.

• Scenario 1: Disbursement failure. A batch of disbursements failed. Support identifies the failure point, restarts the batch with idempotency, and confirms reconciliation.
• Scenario 2: Rate mis-application. A rate change was applied to a broader set of loans than intended. Support identifies the affected loans, recalculates interest, and posts corrections under change control.
• Scenario 9: Provisioning model discrepancy. Provisions under a new model differ materially from the old. Support confirms the model is correct, the calibration is valid, and the difference is explained to finance and risk.
• Scenario 3: Collateral valuation lag. A collateral valuation feed failed; LGD figures are stale. Support re-runs the feed, refreshes LGD, and confirms provisioning impact.
• Scenario 4: Collections surge. Delinquency rates spike in a downturn. Support scales collections capacity, tunes the collection strategy, and reports to the risk committee.
• Scenario 2 (correction): Schedule generation error. A batch of schedules was generated with the wrong day-count convention. Support identifies the affected loans, regenerates schedules, recalculates interest, and communicates with affected borrowers.

Real banking implementation examples

A retail bank migrates from a rules-based decisioning engine to a machine-learning model. The legacy engine used static cut-offs on bureau score and income. The new model uses hundreds of features, including behavioural data from the bank's own systems. Shadow running confirms the new model approves more good risks and declines more bad risks, improving portfolio quality. The migration is run in parallel for months, with traffic gradually shifted, and reconciliation confirming equivalence. Post-migration, approval rates rise, default rates fall, and the bank's lending margin improves.

A bank introduces risk-based pricing for personal loans. Previously, all approved borrowers paid the same rate. The new system assigns each borrower a risk tier and prices the loan accordingly. Lower-risk borrowers pay less; higher-risk borrowers pay more. The change requires updates to the decisioning engine, the product configuration, and the disclosure library. After launch, approval rates rise (some higher-risk borrowers who would have been declined are now served at a higher rate), and the portfolio's risk-adjusted return improves.

A commercial bank implements covenant monitoring for its term loan portfolio. Historically, covenants were checked manually, annually, with delays. The new system ingests borrower financials quarterly, checks covenants automatically, and raises alerts on breaches. The commercial credit team receives real-time visibility into portfolio health; breaches are caught early, enabling earlier intervention. The implementation requires integration to borrower reporting systems, configuration of covenant rules, and a workflow for exception handling.

A bank overhauls its provisioning under IFRS 9. The legacy provisioning was incurred-loss. The new model is forward-looking, with three stages and lifetime expected loss for stages 2 and 3. The change requires a new credit risk engine, integration to the core lending system, and a re-baselining of provisions across the entire portfolio. Transition effects follow the applicable transition provisions; initial adoption can adjust opening equity rather than being a current-period earnings charge. Post-implementation, provisions move with the portfolio's risk, rather than lagging behind it.

Functional lending operating catalogue

A world-class lending estate is built from clear operating capabilities. Each capability below should have a business owner, source system, event contract, authority model, audit record, reconciliation path, and production support playbook. This catalogue is intentionally implementation-oriented: it shows what must be configured, what must happen at runtime, what evidence the bank should retain, and why the capability matters for both consumer and business lending.

Credit application capture

Credit application capture should define borrower identity, requested amount, purpose, channel, consent, affordability data, bureau references, documents, and product version. Runtime behaviour should ensure that consumer applications need clear disclosures and reason codes, while business applications need entity, owner, financial, and authority evidence. The capability should be effective-dated, visible to authorised servicing or operations users, and connected to downstream events for risk, finance, customer communication, and reporting. Tests should include happy path, ineligible request, missing data, manual override, backdated correction, duplicate event, downstream failure, audit retrieval, customer-facing explanation, portfolio-reporting impact, and finance reconciliation. Consumer lending versions should emphasise speed, fairness, affordability, and clear communication. Business lending versions should emphasise borrower group structure, authority, collateral, covenants, reviews, and negotiated terms.

Borrower group management

Borrower group management should define customer, co-borrower, guarantor, business, subsidiaries, directors, related parties, and connected exposure. Runtime behaviour should ensure that group structure drives approval limits, concentration reporting, cross-default review, collateral sharing, and collections strategy.

Eligibility and policy rules

Eligibility and policy rules should define minimum age, residency, account status, product availability, prohibited industries, credit appetite, fraud filters, and affordability rules. Runtime behaviour should ensure that policy rules should be effective-dated and explainable so declined applicants and reviewers can understand the decision basis.

Affordability assessment

Affordability assessment should define income, expenditure, existing debt, dependants, stress rates, disposable income, and verification source. Runtime behaviour should ensure that weak affordability implementation creates loans that book quickly but later fail through predictable repayment stress.

Business financial spreading

Business financial spreading should define balance sheet, profit and loss, cash flow, tax returns, account turnover, adjustments, and analyst commentary. Runtime behaviour should ensure that structured spreading lets risk compare companies consistently while preserving judgement and sector context.

Credit scoring

Credit scoring should define scorecard inputs, model version, score band, cut-off, adverse factors, champion or challenger status, and override reason. Runtime behaviour should ensure that models should support fair treatment, monitoring, drift analysis, and audit of each decision.

Manual underwriting

Manual underwriting should define analyst assessment, strengths, weaknesses, risks, mitigants, requested structure, conditions, and approval recommendation. Runtime behaviour should ensure that manual review should be structured enough for portfolio analytics rather than hidden in long unsearchable documents.

Approval authority

Approval authority should define delegated authority, committee approval, exception level, dual control, expiry date, and approval conditions. Runtime behaviour should ensure that the platform should prevent booking terms that exceed the approver's mandate or differ from approved conditions.

Offer generation

Offer generation should define approved amount, term, price, fees, repayment date, security, conditions, and disclosure pack. Runtime behaviour should ensure that the offer should be generated from the decision record so customers do not receive terms that operations cannot book.

Facility setup

Facility setup should define limit, currency, product, maturity, availability period, borrower group, collateral, guarantees, covenants, and status. Runtime behaviour should ensure that facility setup distinguishes approved commitment from drawn loan account and enables revolving or staged lending.

Loan account booking

Loan account booking should define account number, principal, schedule, interest basis, fee rules, repayment mandate, statement settings, and GL mapping. Runtime behaviour should ensure that booking should not require rekeying approved terms because rekeying creates silent contract and accounting errors.

Drawdown control

Drawdown control should define availability, conditions precedent, facility status, covenants, sanctions, arrears, collateral coverage, and funding route. Runtime behaviour should ensure that drawdown is where approved risk becomes cash, so all unresolved blockers should be visible before funds move.

Disbursement payment

Disbursement payment should define beneficiary, account verification, value date, payment rail, settlement status, and reversal handling. Runtime behaviour should ensure that consumer disbursement may be instant while mortgage or business disbursement may require legal and treasury coordination.

Collateral record

Collateral record should define asset type, owner, value, valuation date, forced-sale value, insurance, lien, perfection, priority, and release terms. Runtime behaviour should ensure that collateral quality changes expected loss, pricing, drawdown availability, recovery strategy, and regulatory capital treatment.

Guarantee record

Guarantee record should define guarantor, guarantee type, amount, expiry, conditions, legal document, and release rules. Runtime behaviour should ensure that guarantees should not be remembered only by relationship managers because enforcement depends on precise documentation.

Covenant management

Covenant management should define definition, threshold, frequency, source documents, due date, tested value, breach status, waiver, and cure period. Runtime behaviour should ensure that business lending needs covenant workflows that block or escalate drawdowns when borrower performance deteriorates.

Rate setup

Rate setup should define fixed rate, floating index, margin, floor, cap, reset frequency, observation date, and customer notice rule. Runtime behaviour should ensure that rate setup must flow into accrual, schedule, statements, payoff quotes, and regulatory disclosures.

Fee setup

Fee setup should define origination, commitment, utilisation, renewal, late, prepayment, waiver, covenant, and servicing fees. Runtime behaviour should ensure that fee rules need authority, caps, tax treatment, GL mapping, and customer or borrower disclosure evidence.

Amortisation schedule

Amortisation schedule should define principal, interest, due date, frequency, balloon amount, grace period, holiday, and maturity. Runtime behaviour should ensure that the schedule is the customer's repayment promise and the bank's expected cash-flow record.

Payment mandate

Payment mandate should define repayment account, direct debit authority, card payment route, payroll deduction, or manual payment instruction. Runtime behaviour should ensure that mandate failure should create a servicing and collections event before delinquency becomes avoidable harm.

Payment allocation

Payment allocation should define fees, interest, principal, arrears, current due, suspense, escrow, disputed amounts, and future instalments. Runtime behaviour should ensure that allocation rules should respect contract and regulation and be visible on statements and servicing screens.

Prepayment handling

Prepayment handling should define extra payment, partial prepayment, full payoff, prepayment fee, interest-to-date, quote expiry, and closure logic. Runtime behaviour should ensure that borrowers expect payoff amounts to be precise, time-bounded, and explainable down to daily interest.

Schedule recalculation

Schedule recalculation should define rate reset, term extension, hardship plan, overpayment, underpayment, capitalisation, and restructure. Runtime behaviour should ensure that recalculation should preserve audit history and explain how old and new schedules differ.

Servicing authority

Servicing authority should define date change, fee waiver, address change, collateral update, payoff quote, hardship, refinance, and restructuring permissions. Runtime behaviour should ensure that servicing users should not change credit exposure, interest, or security without the right approval path.

Customer communication

Customer communication should define approval, decline, offer, disbursement, rate change, payment due, missed payment, hardship, default, and closure notices. Runtime behaviour should ensure that communications should be event-driven and consistent with the exact product terms and account state.

Delinquency classification

Delinquency classification should define days past due, amount past due, missed instalments, expired facility, overdraft excess, covenant default, and cure amount. Runtime behaviour should ensure that classification should drive collections treatment, risk staging, reporting, and customer support.

Collections strategy

Collections strategy should define early reminder, promise-to-pay, hardship review, formal demand, restructuring, legal action, write-off, and recovery. Runtime behaviour should ensure that strategy should vary by product, borrower vulnerability, collateral, balance, risk, and likelihood of cure.

Hardship arrangement

Hardship arrangement should define reduced payment, payment holiday, term extension, rate concession, capitalisation, and exit plan. Runtime behaviour should ensure that hardship must alter billing, collections, risk classification, provisioning, and customer communications consistently.

Default event

Default event should define contractual default, unlikely-to-pay, bankruptcy, fraud, death, insolvency, covenant breach, and collateral impairment. Runtime behaviour should ensure that default should trigger non-accrual review, provision update, recovery strategy, and required reporting.

Write-off and recovery

Write-off and recovery should define charge-off balance, legal balance, recovery account, tax treatment, customer reporting, and post-write-off collections. Runtime behaviour should ensure that write-off is an accounting recognition of loss expectation, not proof that all recovery activity has ended.

Provisioning data

Provisioning data should define PD, LGD, EAD, stage, collateral value, delinquency, macro overlay, model run, and finance posting. Runtime behaviour should ensure that expected-loss calculations depend on clean loan data and timely risk events from servicing and collections.

Risk grading

Risk grading should define borrower grade, facility grade, behavioural grade, override, watchlist status, and review date. Runtime behaviour should ensure that risk grades should update through monitoring rather than remain frozen from origination.

Portfolio monitoring

Portfolio monitoring should define vintage, roll rate, loss rate, prepayment, approval cohort, product, channel, sector, LTV, and score band. Runtime behaviour should ensure that portfolio analytics reveal whether growth is profitable or merely accumulating future losses.

Concentration limits

Concentration limits should define borrower group, sector, geography, product, collateral, currency, risk grade, and maturity concentration. Runtime behaviour should ensure that new approvals and drawdowns should check concentration appetite, not only individual borrower affordability.

Regulatory reporting

Regulatory reporting should define credit bureau, arrears, forbearance, large exposure, capital, expected loss, conduct, and complaints reporting. Runtime behaviour should ensure that reports should use governed definitions and reconcile to finance, risk, and operations totals.

Subledger to GL reconciliation

Subledger to GL reconciliation should define principal, interest, fees, disbursements, repayments, write-offs, recoveries, suspense, and provisions. Runtime behaviour should ensure that daily tie-out prevents customer balances and financial statements from drifting apart.

Interest accrual control

Interest accrual control should define balance, value date, rate, day count, non-accrual status, compounding, rounding, and adjustment reason. Runtime behaviour should ensure that accrual errors are high-risk because they affect customer statements, income, tax, and remediation.

Suspense management

Suspense management should define unknown payment, wrong reference, overpayment, closed-loan payment, partial payment, and currency mismatch. Runtime behaviour should ensure that aged suspense should have owner and resolution because it may represent customer money or misapplied cash.

Collateral release

Collateral release should define payoff confirmation, cross-collateral check, legal hold, unpaid fee, pending transaction, and release instruction. Runtime behaviour should ensure that release should be prompt when valid and blocked when the bank still relies on the security.

Annual review

Annual review should define fresh financials, covenant results, risk grade, collateral value, pricing, limits, and renewal decision. Runtime behaviour should ensure that business facilities should not roll forward silently on stale borrower information.

Renewal workflow

Renewal workflow should define expiry alert, borrower refresh, risk assessment, terms update, documentation, approval, and booking. Runtime behaviour should ensure that renewal is a credit decision, not only an administrative extension of maturity.

Restructuring workflow

Restructuring workflow should define problem diagnosis, viable terms, borrower capacity, concession type, accounting treatment, and monitoring plan. Runtime behaviour should ensure that restructuring should avoid hiding default while still supporting borrowers who can recover.

Fraud and application abuse

Fraud and application abuse should define synthetic identity, income manipulation, document fraud, mule accounts, broker fraud, and repeat applications. Runtime behaviour should ensure that lending fraud controls should connect onboarding, account behaviour, disbursement route, and early arrears.

Broker and partner channels

Broker and partner channels should define lead source, commission, disclosure, document quality, conversion, early performance, and complaint tracking. Runtime behaviour should ensure that partner growth should be measured against loan quality and customer outcomes, not just volume.

Open banking income verification

Open banking income verification should define customer consent, account data, income classification, expense detection, refresh date, and fallback rules. Runtime behaviour should ensure that automated verification improves speed only if classifications are transparent and challengeable.

Treasury funding forecast

Treasury funding forecast should define expected disbursement, repayment, prepayment, maturity, currency, rate reset, and large borrower movement. Runtime behaviour should ensure that lending changes the bank's liquidity profile and should feed funding and interest-rate risk management.

Profitability analysis

Profitability analysis should define net interest margin, expected loss, fees, capital cost, funding cost, servicing cost, and collection cost. Runtime behaviour should ensure that loan profitability should be measured after risk and operations, not only on headline interest rate.

Conduct monitoring

Conduct monitoring should define decline reasons, affordability overrides, complaints, hardship outcomes, fees, collections contacts, and vulnerable customers. Runtime behaviour should ensure that good lending platforms measure whether borrowers are treated fairly after the loan is booked.

Data lineage

Data lineage should define source of application data, decision data, loan balance, GL balance, risk attributes, and report transformations. Runtime behaviour should ensure that lineage lets the bank defend numbers in audits, exams, model validation, and customer remediation.

Regression testing

Regression testing should define product rules, interest, fees, schedule, allocation, delinquency, provisioning, reports, and lifecycle changes. Runtime behaviour should ensure that lending changes should be tested on realistic accounts because small formula changes compound over years.

Maturity management

Maturity management should define facility expiry, balloon payment, interest-only period end, promotional-rate end, covenant review, and refinancing window. Runtime behaviour should ensure that maturity alerts prevent customers and relationship managers from discovering renewal needs after credit availability has already expired.

Closure workflow

Closure workflow should define zero balance, pending payments, unpaid fees, legal holds, linked collateral, cross-collateral exposure, and final statement. Runtime behaviour should ensure that closure should stop billing and reporting correctly while preventing premature security release.

Production support evidence

Production support evidence should define incident ID, affected accounts, failed batch, correction method, customer impact, finance impact, and preventive fix. Runtime behaviour should ensure that support teams need enough data to repair bad postings or schedules without inventing manual workarounds.

Model monitoring

Model monitoring should define score stability, approval drift, override performance, discrimination testing, loss by band, and reject inference where used. Runtime behaviour should ensure that decision models should be governed after launch because economic conditions and applicant populations change.

Document management

Document management should define required document type, version, source, validation result, expiry, retention, legal hold, and customer visibility. Runtime behaviour should ensure that missing or stale documents should block the right actions and feed operations queues.

Collections contact governance

Collections contact governance should define channel, time, frequency, consent, vulnerability, dispute status, promise history, and regulatory contact limits. Runtime behaviour should ensure that collections intensity should be controlled so recovery work does not become customer harm.

Remediation workflow

Remediation workflow should define affected population, error cause, recalculation, refund, interest correction, customer notice, and regulator reporting. Runtime behaviour should ensure that when lending defects occur, the platform should support precise remediation rather than broad manual estimates.

Management control pack

Management control pack should define new lending, approvals, declines, drawdowns, arrears, provisions, exceptions, breaks, complaints, losses, and profitability. Runtime behaviour should ensure that executives should see a reconciled view of growth, risk, customer outcomes, and financial performance.

Lending acceptance scenarios

The borrower's journey: lending from the customer's side

Lending chapters usually describe the bank's machinery; the borrower's experience of that machinery decides whether the lending franchise grows, and the two views must be designed together.

The journey begins with a need and a fear: the customer needs money for something that matters, a home, a van, a season's stock, and fears the process, the jargon, the judgment and the fine print. The application journey either respects this or exploits it: clear questions in human language, documents requested once with reasons, honest indication of likelihood and price before the hard credit check, and a decision communicated when promised, form the journey customers recommend; opaque processes, repeated document requests and decisions that arrive late form the journeys customers leave. The offer stage is where understanding is won or lost: the rate, the total cost, the monthly obligation and the conditions must be comprehensible to the actual customer, not to a credit officer, and regulation increasingly tests comprehension rather than signature.

In-life, the borrower experiences the loan through statements, rate changes, and the moments that matter: the payment difficulty faced early with options, the life event that changes circumstances, the request to borrow more or pay down faster. A lending operation that treats the performing borrower as a relationship, proactive communication, fair restructuring, easy prepayment, builds the book's quality and the franchise's reputation together; one that treats the borrower as a receivable learns the difference at collections time. The final stage, the loan repaid and closed cleanly, the mortgage deed released, the relationship graduated, is the lending journey's referral engine, and it is shocking how many banks fumble the ending with slow releases and silent closures after years of perfect payments.

Credit policy: the constitution every loan answers to

Behind every credit decision stands a document most customers never see and every credit professional lives inside: the credit policy, the bank's constitution for lending.

The policy's architecture follows the credit lifecycle: appetite statements define what the bank wants to lend, to whom, secured how, at what concentrations; origination standards define minimum evidence, assessment methods and approval authorities; product standards define each lending product's terms, covenants and structures; in-life standards define review, renewal and restructuring treatment; and impairment standards define classification, provisioning and recovery approaches. Each element carries an owner, an approval level and a review date, and the hierarchy of the policy, what may be varied, by whom, with what escalation, is the constitution's amendment process, applied every time a deal strains at a boundary.

Exceptions are the policy's stress test: every exception granted is a loan outside the constitution, and disciplined banks record, limit, report and review them, because a portfolio of unrecorded exceptions is a policy that exists only on paper. The policy's life in practice is decided by its readability and its enforcement: a policy that credit officers can navigate approves consistently; one that requires legal interpretation produces divergence; and enforcement through automated policy rules in the origination workflow, checking each deal against the constitution at booking, turns the policy from literature into machinery. Supervisors read credit policy before they read portfolios, because the policy tells them what the bank intended, and the portfolio tells them what actually happened, and the distance between the two is the credit function's true report card.

Lending operations: the machinery after approval

Approval is the midpoint of lending, not the end, and the operations that follow decide whether good decisions become good loans.

Documentation and completion translate the decision into contract: facility documents generated from approved terms, signed with proper authority, conditions precedent verified before money moves, and security perfected, registered, deposited, insured, before drawdown, because an unperfected security discovered at enforcement is the most expensive paperwork failure in banking. Disbursement is a controlled financial event: funds released per the contract's mechanics, to verified destinations, with the drawdown recorded against the facility and the ledger's first entries reconciled, and business lending adds utilisation management as the facility lives, drawdowns and repayments tracked against limits, availability computed honestly, and commitment fees accruing on the undrawn part.

In-life administration is the long middle: repayments collected and applied per the waterfall, interest and fees computed and communicated, covenants monitored with evidence, insurance renewed, security revalued on cycle, and the borrower's periodic information received and reviewed. Each is a queue, a calendar and a control, and their aggregate quality is the portfolio's operational risk: covenant breaches found late, security lapsed unremarked, repayments misapplied, are the defects that turn performing books into problem books. The loan that finishes well, fully repaid, security released promptly, records complete, is the operation's product, and borrowers remember the release letter's speed long after they forget the rate.

Portfolio management: lending as a managed whole

Individual loans are underwritten one at a time; the book they form behaves as a portfolio, and portfolio management is the discipline of seeing the forest the credit officers' trees compose.

Concentration is the first lens: exposure by sector, geography, product, collateral type and counterparty group, measured against appetite, because a portfolio of individually sound loans can be collectively fragile when they share a hidden correlation, the property developer book that is one regional market, the merchant cash advances that are one retail season. Vintage analysis is the second: cohorts of loans tracked by origination period, their delinquency curves compared across time, because vintage curves detect underwriting drift years before the aggregate numbers move, and every credit crisis in history was visible in the vintages before it was visible in the book.

Portfolio actions follow from the lenses: appetite adjusted as sectors and cycles move, origination standards tightened or loosened deliberately, pricing re-tiered against demonstrated risk, growth steered toward segments the portfolio needs and away from concentrations it must not deepen, and occasionally books sold or hedged when no origination change can fix the shape. The reporting that drives these actions is the credit function's instrument panel: the portfolio's risk composition, migration between risk grades, early-warning indicators, provisioning trajectory and stress-test results, reviewed on a governance calendar with authority to act. Lending done well at the individual level and ignored at the portfolio level ends the same way as lending done badly, only later; the portfolio lens is what makes the difference visible in time.

Common implementation mistakes

• Floating-point money. Rounding errors compound across millions of transactions into material amounts.
• Hardcoded rates and rules. Embedding rates or limits in code makes every change a release.
• Missing schedule versioning. Changing a loan's schedule without preserving history breaks historical interest.
• Weak underwriting. A weak scorecard approves bad risks; the losses accumulate slowly and then suddenly.
• Ignoring concentration. Concentration in one sector or geography can overwhelm portfolio-level metrics.
• Stale collateral valuations. Outdated valuations understate LGD and provisions.
• Inconsistent day-count conventions. Mixing conventions between products produces reconciliation breaks.
• Poor idempotency. Disbursement and repayment APIs that double-post on retry create financial loss.
• Underestimating collections. A bank that under-resources collections in a downturn suffers higher losses.
• Model drift. Scoring and provisioning models that are not refreshed become blind to new patterns.
• Disclosure drift. Letting disclosures fall out of sync with configured rates creates regulatory exposure.
• Manual exception handling. Unsystematised exceptions create audit gaps and inconsistent treatment.

Best practices

• Decimal money, never floats. Use decimal types or integer minor units; round only at defined points.
• Configuration over code. Rates, policies, and scorecard thresholds are data, read at runtime.
• Effective-date everything. Every rate, fee, and threshold carries effective dates.
• Versioned models. Scorecards and provisioning models are versioned; shadow-run new models before cutover.
• Concentration limits. Enforce limits at borrower, sector, and geography level.
• Forward-looking provisioning. Provisions reflect expected loss, not just incurred loss.
• Fresh collateral valuations. Valuations refreshed per policy; stale valuations flagged.
• Daily reconciliation. Disbursement, repayment, interest, provision to GL.
• Observability. Portfolio health, delinquency curves, provisioning coverage visible on dashboards.
• Tested downturn plans. Collections capacity and strategies tested against stress scenarios.
• Model monitoring. Scorecard and provisioning model performance monitored and refreshed.
• Disclosure at configuration time. When a rate or fee is configured, the disclosure is updated in the same change.
• Strong idempotency. Disbursement and repayment APIs are idempotent, identified by client correlation ID.

Summary

Lending is the discipline of putting money at risk on the strength of a promise to repay. It is the largest source of revenue for most banks and the largest source of risk. Every loan is an assessment of the five Cs (character, capacity, capital, collateral, conditions), structured as a contract (principal, interest, term, schedule, security), priced to cover expected loss and earn a margin, disbursed, serviced, monitored, and eventually either fully repaid or written off.

We have walked through the credit decisioning process, scoring, risk-based pricing, loan structure, interest calculation, the loan lifecycle, delinquency and default, credit risk management, provisioning under IFRS 9, consumer vs business lending, systems and architecture, data model, business rules, risks and controls, exceptions, and reconciliation. We have viewed the discipline through the eyes of business analyst, solution architect, developer, operates, and production support, and grounded it in real examples, common mistakes, best practices, and interview-ready reasoning.

The bank that masters lending earns a margin on money at risk; the bank that does not, loses it. The choice, made in every loan, every score, every schedule, every provision, is consequential and enduring.

Key takeaways

• Lending is the discipline of putting money at risk on the strength of a promise to repay; it is the largest source of revenue and the largest source of risk for most banks.
• The five Cs of credit — character, capacity, capital, collateral, conditions — are the timeless framework that underwrites every lending decision.
• Credit scoring turns the five Cs into a number that predicts the probability of default; risk-based pricing ties the rate to the risk.
• A loan is structured through principal, interest rate, term, amortisation, schedule, security, covenants, and fees; each choice shapes the cash flows, the risk, and the accounting.
• The loan lifecycle runs origination, disbursement, repayment, monitoring, delinquency, default, and closure; each stage has its own systems, controls, and accounting.
• Credit risk is managed through probability of default (PD), loss given default (LGD), exposure at default (EAD), and expected loss (EL = PD × LGD × EAD); the bank prices for expected loss and provisions against it.
• The general IFRS 9 approach uses forward-looking 12-month or lifetime expected losses according to credit deterioration and impairment; the horizon is not lifetime for every performing loan.
• Consumer lending is high-volume, rules-driven; business lending is lower-volume, analyst-led; both must operate correctly for the bank to earn its lending margin.
• The lending estate is a constellation of systems — origination, core lending, decisioning, disbursement, servicing, collections, collateral, risk engine, GL, reporting — increasingly cloud-native and event-driven.
• Decimal money, configuration over code, effective-dating, versioned models, concentration limits, forward-looking provisioning, daily reconciliation, and tested downturn plans are what separate a lending estate that scales from one that does not.

Working-capital lending: approve the source of repayment

A fictional wholesaler turns inventory in 45 days, collects invoices in 30 days and pays suppliers in 25 days. Its cash conversion cycle is 45 + 30 - 25 = 50 days. A seasonal revolving line may fit this timing better than financing permanent assets through a short overdraft. Analyse customer concentration, disputed receivables, ageing, inventory obsolescence and whether a proposed draw is eligible under the facility.

If eligible receivables are 200,000 and the agreed advance rate is 70 percent, the gross borrowing base is 140,000. Deduct reserves and prior utilisation before comparing it with the legal facility limit. A 10,000 disputed invoice excluded from the base reduces capacity by 7,000 under this simple assumption; a 50,000 facility headroom calculation alone misses that restriction. Collateral value is not unrestricted repayment cash and a guarantee is not a deposit.

Document borrower and guarantor authority, conditions precedent, perfected security where required, monitoring covenants and expiry. Approval, signed contract, available facility and actual disbursement are different states. The delinquency table is an illustrative collection model; default, non-performing classification, impairment and charge-off use different applicable definitions. Apply Regulation B adverse-action and notification requirements, 12 CFR 1002.9, to covered US credit decisions. Business-credit notice and record rules have specified distinctions; do not assume every SME receives identical consumer procedures.