Loan Application Explanation

Methodology

Step 1 — Parse the Decision Engine Output

Extract the binary decision (approve/deny/counter), the scorecard tier, and all triggered rules. Map each triggered rule to its human-readable policy description. Identify whether the decision was purely model-driven, policy-overlaid, or manually overridden.

Step 2 — Rank Contributing Factors by Materiality

Order the decision factors by their marginal contribution to the outcome:

1. Credit score impact — Distance from cutoff, score band placement
2. DTI ratio — Front-end and back-end DTI vs. product maximums (e.g., 43% QM threshold)
3. LTV/CLTV — Loan-to-value against program limits and PMI triggers
4. Employment/income stability — Verification status, gap analysis, income trending
5. Collateral adequacy — Appraisal vs. purchase price, property condition flags
6. Compensating factors — Reserves, residual income, co-borrower strength

Step 3 — Map to Regulatory Reason Codes

For denials and adverse actions, map the top contributing factors to the standardized adverse action reason codes (OCC/FFIEC model codes or proprietary mappings). Select the top 4 reason codes ranked by impact. Verify codes align with the actual decision drivers—never use generic placeholders.

Step 4 — Generate Borrower-Facing Narrative

Compose a plain-language explanation that:

• States the decision clearly in the first sentence
• Lists specific factors in order of importance
• Avoids prohibited language (no references to protected classes)
• Includes required disclosures (credit bureau source, score used, score range)
• Provides the applicant's right to obtain a free credit report within 60 days
• References the CFPB or appropriate regulatory contact for disputes

Step 5 — Produce Internal Audit Documentation

Generate a parallel internal document that includes:

• Full factor decomposition with numeric values
• Policy rules triggered with version identifiers
• Any manual override justification and approver identity
• Comparative analysis against similarly situated applicants (fair lending lens)
• Exception tracking reference number if applicable

Step 6 — Fair Lending Compliance Review

Cross-check the explanation against fair lending red flags:

• Ensure no prohibited basis factors influenced the decision
• Verify similarly situated borrowers received consistent treatment
• Flag any statistical outliers for second-look review
• Document disparate impact analysis results if available

Step 7 — Assemble Final Deliverable

Package the explanation into the required format: adverse action notice (denial), statement of credit terms (approval with conditions), or counter-offer letter. Include all regulatory disclosures, timing requirements (30-day notice window), and delivery method documentation.

Output Specification

## Loan Decision Explanation

### Decision Summary
- **Application ID**: [ID]
- **Decision**: [Approved / Denied / Counter-Offered]
- **Date**: [Decision date]
- **Product**: [Loan type and term]

### Primary Decision Factors (Ranked)
1. [Factor]: [Plain-language explanation with specific values]
2. [Factor]: [Plain-language explanation with specific values]
3. [Factor]: [Plain-language explanation with specific values]
4. [Factor]: [Plain-language explanation with specific values]

### Adverse Action Reason Codes (if applicable)
- Code [X]: [Description]
- Code [Y]: [Description]
- Code [Z]: [Description]
- Code [W]: [Description]

### Regulatory Disclosures
- Credit score used: [Score] from [Bureau] (range [XXX]–[XXX])
- Right to free credit report within 60 days from [Bureau contact]
- Right to dispute accuracy of information with [Bureau]

### Internal Audit Trail
- Scorecard version: [ID]
- Policy overlay version: [ID]
- Override: [Yes/No — if yes, justification and approver]
- Fair lending flag: [None / Flagged for review — reason]

Analysis Framework

Apply the FACT framework for every explanation:

• Factors — Enumerate all contributing decision factors with values
• Action codes — Map to regulatory-compliant reason codes
• Compliance — Verify ECOA/Reg B/TILA disclosure requirements
• Transparency — Ensure plain-language accessibility (8th-grade reading level)

Examples

Example 1 — Conventional Mortgage Denial

Input: FICO 612, DTI 48%, LTV 95%, 2 years employment, 1 month reserves Decision: Denied Explanation: "Your application for a 30-year conventional mortgage was not approved. The primary factors were: (1) your credit score of 612 is below the minimum threshold of 640 for this program; (2) your total debt-to-income ratio of 48% exceeds the maximum of 45%; (3) your available reserves of 1 month are below the 2-month minimum required for loans with LTV above 90%."

Example 2 — Auto Loan Approval with Conditions

Input: FICO 695, DTI 38%, employment 5 years, requested $35,000 Decision: Approved at Tier 2 pricing Explanation: "Your auto loan application has been approved for $35,000 at 6.49% APR for 60 months. Your rate reflects Tier 2 pricing because your credit score of 695 falls in the 680–719 band. A score of 720 or above would qualify for our best rate of 4.99% APR."

Guidelines

• Always use the specific credit score, DTI ratio, and LTV values — never vague language
• Limit adverse action notices to exactly 4 reason codes per Reg B convention
• Include the credit score disclosure addendum on every denial and counter-offer
• Use plain language at an 8th-grade reading level for borrower-facing content
• Never reference race, ethnicity, sex, marital status, age, or other prohibited bases
• Maintain separate internal and external explanation documents
• Timestamp all explanations and link to the specific scorecard/policy version
• Retain explanation records for the regulatory minimum retention period (25 months for ECOA)

Validation Checklist

• Decision factors are ranked by actual materiality, not arbitrary order
• Adverse action codes match the true decision drivers
• All Reg B required disclosures are present (bureau, score, range, rights)
• No prohibited basis language or proxies appear in the explanation
• Internal documentation includes scorecard version and policy overlay IDs
• Plain-language readability target met (Flesch-Kincaid grade 8 or below)
• Counter-offer letters include both original and revised terms
• Explanation is timestamped within the 30-day adverse action notice window
• Fair lending comparative analysis is documented for denials
• Exception-to-policy decisions have documented justification and approval chain