Underwriting

Underwriting is the process of evaluating a credit application to decide whether to lend, how much to lend, and on what terms. The underwriter assesses the applicant's ability and willingness to repay, the quality of any security offered, and the fit with the lender's credit policy, then converts that assessment into a decision: approve, decline, or approve with conditions.

Underwriting is where a lender prices risk. Everything downstream — collections effort, provisioning, portfolio quality — is largely determined by decisions made at this stage.

Underwriting in other contexts

The term appears in three distinct industries and the meanings diverge:

• Lending — assessing a borrower's creditworthiness before advancing funds. This is the sense used throughout this page.
• Insurance — assessing an applicant's risk profile to decide whether to issue a policy and at what premium.
• Securities — an investment bank agreeing to purchase and distribute a new issue of shares or bonds, assuming the risk that it does not sell.

The common thread is the historical origin: a party writing their name under a statement of risk to signal they accept it.

Underwriting vs. related terms

These stages are frequently collapsed into one another, which causes confusion in process design and in job titles.

• Loan origination — The whole front-end journey from application intake through to disbursement.
• Credit analysis — The analytical work of interpreting financials, cash flow and repayment capacity.
• Underwriting — Analysis plus the risk judgement and recommendation against credit policy.
• Credit approval — The formal authorisation decision, made by an officer or committee holding the required limit.
• Loan processing — The administrative work of collecting documents, verifying data and preparing the file.

In practice, credit analysis feeds underwriting, and approval follows it. A small lender may have one person doing all four; a larger institution separates them deliberately, because the person who recommends a loan should not be the person who authorises it.

The five Cs of credit

The traditional framework underwriters work through, still the clearest organising structure for a credit assessment:

Character — the applicant's track record and willingness to repay. Evidenced by repayment history, credit bureau records, references, length of business operation and conduct of existing accounts.

Capacity — the ability to service the debt from income or business cash flow. The most quantitative of the five, and the one that determines the loan amount.

Capital — the applicant's own stake. A borrower contributing equity to a purchase, or holding retained earnings in a business, has more to lose from default.

Collateral — assets pledged as a secondary repayment source if capacity fails.

Conditions — the purpose of the loan and the external environment: sector outlook, seasonality, currency exposure, regulatory change, concentration of the borrower's own customers.

A sixth is often added in practice: Compliance — identity verification, sanctions and politically exposed person screening, and the anti-money-laundering checks that must clear regardless of credit quality.

What underwriters actually assess

Identity and legal standing. Verified identity documents, business registration, tax registration, and authority of the signatory to borrow on behalf of an entity.

Income and cash flow. Payslips and bank statements for salaried applicants; bank statements, sales records, stock turnover and supplier records for traders and small businesses. For informal businesses with no financial statements, underwriters reconstruct cash flow from observation and from the applicant's own records, then discount it.

Existing debt. Bureau reports, declared obligations, and evidence of borrowing from informal sources. Undisclosed parallel borrowing is a leading cause of default.

Affordability. Disposable income after living costs and existing obligations, not just gross income. Many jurisdictions now impose a statutory affordability test.

Purpose and use of funds. Whether the stated purpose is plausible, productive, and consistent with the requested amount and tenor.

Security. Ownership, valuation, condition and enforceability of any pledged asset, plus the standing of any guarantor.

Fit with policy. Whether the application falls inside the lender's stated appetite for sector, tenor, ticket size, geography and product.

Key underwriting metrics

Debt-to-income (DTI) — total monthly debt obligations divided by gross monthly income. Screens consumer affordability.

Debt service coverage ratio (DSCR) — net operating income divided by total debt service. A DSCR of 1.0 means income exactly covers repayments with no margin; lenders typically require a cushion above that.

Loan-to-value (LTV) — loan amount as a percentage of appraised collateral value. Lower LTV means a larger buffer against value decline.

Instalment-to-income ratio — the proposed repayment as a share of income, used where full DTI data is unavailable.

Credit score — a statistical estimate of default probability, from a bureau or an internal scorecard.

Probability of default, loss given default and exposure at default — the components of expected loss, used by larger institutions to price and provision.

Manual, automated and hybrid underwriting

Manual (judgmental) underwriting relies on an experienced officer weighing the evidence. It handles unusual cases, thin credit files and informal income well. It is slow, expensive per file, and inconsistent between officers — two underwriters can reach different decisions on identical facts.

Automated underwriting applies rules and statistical scorecards to structured data, returning a decision in seconds. It is consistent, cheap at volume and auditable. It fails on cases outside the data it was built from, and it can encode historical bias if the training data reflected it.

Hybrid underwriting auto-decides the clear approvals and clear declines, and routes the middle band and all exceptions to a human. This is the dominant model, because the marginal cases are both the smallest share of volume and the largest share of risk.

The choice is not purely technical. Automated decisions at scale attract regulatory attention around explainability and fairness, and many jurisdictions give applicants a right to an explanation of an adverse decision.

The underwriting process

1. Intake and completeness check — confirm the application is complete and the mandatory documents are present.
2. Identity and compliance screening — verify the applicant and clear regulatory checks before analytical work begins.
3. Data gathering — bureau reports, bank statement analysis, financial statements, references, site or business visit.
4. Verification — confirm income, employment, business existence and asset ownership independently of the applicant's assertions.
5. Analysis — build the cash flow picture, calculate ratios, stress the numbers against plausible downside scenarios.
6. Security assessment — value and verify collateral; assess guarantors.
7. Policy check — test the application against credit policy, exposure limits and concentration caps.
8. Recommendation — approve, decline, or approve with conditions, amount, tenor, pricing and covenants stated.
9. Approval — authorisation by whoever holds the delegated limit for that exposure size.
10. Documentation and disbursement conditions — the conditions precedent that must be satisfied before funds are released.

Underwriting policy and delegated authority

A credit policy sets out what the lender will and will not do: eligible borrower types, maximum tenor, ticket size bands, acceptable security, required ratios, prohibited sectors and exposure limits per borrower or group.

Delegated authority assigns approval limits by role, so that larger exposures require more senior authorisation and the largest go to a credit committee. The design principle is separation: origination, underwriting and approval should not all sit with one person, because an officer who both recommends and approves has no check on optimistic assessment.

Exceptions and overrides — cases approved outside policy — should be permitted but tracked. Override rates and the subsequent performance of overridden loans are among the most useful diagnostics a lender has, because a rising override rate usually precedes a rising default rate.

Underwriting in microfinance and SME lending

Where borrowers lack audited accounts, formal payslips or registrable assets, classical underwriting inputs simply do not exist. Microfinance institutions (MFIs) and small business lenders substitute other methods:

• Cash flow reconstruction — building a simple income statement from the borrower's own record-keeping, stock counts and observed sales during a business visit.
• Character-based assessment — community references, group membership, standing among suppliers and customers.
• Progressive lending — starting with a small loan and increasing the limit as repayment history builds, letting behaviour substitute for documentation.
• Group screening — where a group's members guarantee each other, they perform selection that the lender cannot, since they know each other's circumstances directly.
• Alternative data — mobile money transaction history, airtime purchase patterns, utility payments and supplier ledgers.

These methods trade documentary certainty for local knowledge, and they depend heavily on loan officer judgement and turnover in the field.

Common underwriting failures

• Verification skipped under volume pressure. Stated income accepted because checking it slows throughput.
• Undetected parallel borrowing. The borrower's true obligations exceed what any single lender can see.
• Correlated risk ignored. A portfolio of applicants who all sell into the same market, or all depend on one employer, passes individually and fails together.
• Collateral treated as capacity. Approving a loan the borrower cannot service because the security looks strong. Enforcement is slow, costly and uncertain; it is not a repayment plan.
• Policy drift. Standards loosened incrementally during growth pushes, with no single decision visibly crossing a line.
• Stale scorecards. Statistical models built on pre-shock data applied unchanged after conditions have changed.
• No feedback loop. Underwriting decisions never compared against subsequent performance, so the criteria never improve.

Regulatory considerations

Most jurisdictions now regulate credit decisioning in some form. Common requirements include a documented affordability assessment before advancing consumer credit, prohibition of discrimination on protected characteristics, disclosure of the effective interest rate and total cost of credit, retention of the decision file for a defined period, and a right for declined applicants to be told the principal reason. Where decisions are automated, additional obligations around explainability and human review frequently apply.

Frequently asked questions

How long does underwriting take? Anywhere from seconds to several weeks. Automated consumer decisions return instantly; a secured business facility requiring valuation, site visits and committee approval commonly takes two to six weeks.

What is the difference between pre-qualification and underwriting? Pre-qualification is an indicative assessment based on unverified information supplied by the applicant. Underwriting is the verified assessment that produces a binding decision.

Can an underwriting decision be appealed? Most lenders operate a reconsideration process where the applicant supplies additional evidence. Some jurisdictions also grant a statutory right to request review of an automated decision by a human.

Does a decline damage the applicant's credit record? The application enquiry is usually recorded and visible to other lenders; the decline outcome itself often is not. Repeated enquiries in a short window are generally read as a negative signal.

Who is responsible if an underwritten loan defaults? Individual defaults are expected and priced for. Accountability attaches to patterns rather than single cases, which is why override rates, vintage analysis and portfolio at risk by originating officer are monitored.