Human review of a credit application using verified documents, lending policy, repayment analysis, and documented judgment rather than an automated result alone.
Manual underwriting is a credit-review process in which a human underwriter evaluates an application, verifies supporting evidence, applies lending policy, and documents a decision rather than relying on an automated result alone. The review can supplement an automated underwriting system, resolve a referral, or serve as the primary method for complex consumer and business credit.
Manual does not mean informal or standard-free. A defensible manual decision still needs consistent criteria, verified data, approval authority, fair-lending controls, and a traceable reason for approval, counteroffer, or denial.
Record the amount, purpose, maturity, payment schedule, rate structure, collateral, guarantors, and conditions. Creditworthiness cannot be assessed independently of the proposed obligation.
Possible evidence includes:
The applicable product and jurisdiction determine what a creditor may request and use.
The underwriter identifies the primary repayment source and tests whether it is reliable enough for the term. Collateral can support recovery but does not replace recurring cash flow.
The application is compared with eligibility rules, risk limits, documentation standards, pricing, collateral margins, and exception authority.
The file should show verified facts, calculations, assumptions, exceptions, conditions, approvers, and the actual reasons for the outcome.
Assume a borrower applies for a loan with a proposed monthly payment of $900. The application reports:
| Item | Monthly amount |
|---|---|
| Base gross income | $5,000 |
| Claimed average bonus income | $1,000 |
| Existing monthly debt payments | $1,300 |
| Proposed loan payment | $900 |
Using base income alone:
DTI = ($1,300 + $900) / $5,000 = 44%
If the lender’s policy permits stable bonus income and the underwriter verifies an acceptable history and likely continuation, the analyzed income could be $6,000:
DTI = ($1,300 + $900) / $6,000 = 36.7%
The arithmetic does not decide the application. The underwriter must determine whether the bonus is eligible, whether all debts are included, whether the payment can change, and whether other policy requirements are met. Unverified bonus income is not a compensating factor.
| Area | Evidence | Decision question |
|---|---|---|
| Character and repayment history | Credit file, bank or trade history | Has the borrower generally met obligations as agreed? |
| Capacity | Income or operating cash flow, debt service | Can the borrower make the proposed payments? |
| Capital and liquidity | Savings, equity, business capital | Is there a buffer against disruption? |
| Collateral | Valuation, ownership, lien search | What recovery support exists if repayment fails? |
| Conditions | Purpose, industry, economy, loan structure | What external or transaction risks can change performance? |
| Documentation quality | Reconciliations and third-party records | Are material facts complete and reliable? |
These areas resemble the 5 Cs of Credit, but actual policy can use another framework.
| Feature | Manual underwriting | Automated underwriting |
|---|---|---|
| Decision mechanism | Human applies policy and judgment | System applies rules, scorecards, or models |
| Best suited to | Complex, referred, nonstandard, or document-heavy cases | High-volume cases with structured data |
| Main strength | Can evaluate documented context | Speed and consistent calculation |
| Main risk | Inconsistent discretion or weak documentation | Data, model, threshold, or explainability errors |
| Control need | Authority, second review, exception tracking | Validation, monitoring, governance, override controls |
Many lenders combine the two. An automated system can calculate and screen, while a person reviews exceptions or validates unusual data.
A policy exception occurs when an otherwise considered application falls outside a stated standard. An override changes or rejects an automated or delegated result.
A strong exception record identifies:
An exception cannot cure a legal prohibition, missing mandatory eligibility, unsupported income, fraud indicator, or absent authority.
Under U.S. Regulation B, a creditor generally may consider obtained information when evaluating credit, but it cannot use information barred by the regulation or discriminate on a prohibited basis. Human discretion can create inconsistent outcomes if policies, training, review, and exception controls are weak.
When adverse action rules apply, the creditor must provide the required notice and actual principal reasons. A vague label such as “failed manual review” does not explain the decision if the real reason was insufficient verified income, delinquent obligations, or another specific factor.
Rules differ for consumer, mortgage, small-business, and other credit. This page does not determine the notice required for a particular application.
Manual review can capture context that structured models miss, but it can also introduce inconsistency, cognitive bias, operational delay, and recordkeeping failures. Complex files can still contain incomplete or fraudulent evidence.
This page is educational and is not personalized lending, legal, fair-lending, mortgage, business, or financial advice. Actual underwriting depends on product rules, lender policy, verified facts, and current law.