Financial Validation

Overview

LoanPilot performs automated financial validation by synthesizing data from multiple sources—Bank Statements, GST Returns, and Financial Statements (P&L/Balance Sheets). This process ensures that the data used for underwriting is mathematically consistent and verified across different reporting formats.

The validation engine uses AI-driven extraction paired with strict schema enforcement (Zod-based) to detect discrepancies, fraud risks, and repayment capacity.

Document-Level Validation

Each financial document undergoes an analysis phase where key metrics are extracted and validated for logical consistency.

Bank Statement Analysis

Validates the cash flow health of the borrower.

• Period Integrity: Ensures the analysis covers the requested start and end dates.
• Balance Validation: Calculates daily balances to identify negative balance trends.
• Fraud Detection: Evaluates fraud_risk_score based on cash deposit percentages and transaction anomalies.
• Repayment Capacity: Automatically identifies existing EMIs and calculates the emi_to_inflow_ratio and net_disposable_income.

Financial Statement Analysis

Processes Balance Sheets, Profit & Loss statements, and ITRs.

• Mathematical Consistency: Validates that Total Assets = Total Liabilities + Total Equity.
• Ratio Benchmarking: Calculates critical performance indicators:
  • Liquidity: Current Ratio, Quick Ratio.
  • Solvency: Debt-to-Equity, Debt-to-Assets.
  • Profitability: Net Profit Margin, Gross Profit Margin, ROE, ROA.
  • Coverage: Debt Service Coverage Ratio (DSCR).

Cross-Document Reconciliation

The Reconciliation engine is the core of the validation layer. it cross-references data between disparate documents to ensure the borrower's reported figures match their actual cash flow.

Key Reconciliation Checks

| Check Type | Logic | | :--- | :--- | | GST vs. Bank | Compares total credits in bank statements against outward supplies reported in GST returns. | | GST vs. P&L | Cross-references annual revenue in the Profit & Loss statement with the sum of GST returns for that fiscal year. | | P&L vs. Bank | Validates reported revenue and operating expenses against actual bank inflows and outflows. |

API Usage

Triggering Analysis

To run a financial validation, POST to the respective analysis endpoint.

Endpoint: POST /api/analyze/financial-statement

Request Body:

{
  "documentId": "uuid",
  "applicationId": "uuid",
  "organizationId": "uuid",
  "statementType": "profit_loss", // Options: balance_sheet, profit_loss, itr
  "aiProvider": "openai" // Optional
}

Response Schema: The system returns a structured analysis object including calculated ratios and a confidence score:

{
  "success": true,
  "data": {
    "period_year": 2023,
    "profit_loss": {
      "revenue": 500000,
      "net_income": 75000
    },
    "ratios": {
      "net_profit_margin": 15.0,
      "current_ratio": 1.8,
      "dscr": 2.1
    },
    "confidence_score": 0.98
  }
}

Performing Reconciliation

To trigger the mathematical cross-check between all uploaded and analyzed documents.

Endpoint: POST /api/analyze/reconciliation

Request Body:

{
  "applicationId": "uuid"
}

Validation Flags and Confidence Scores

Every validation result is accompanied by metadata that should be reviewed by the underwriter:

1. Confidence Score: A float between 0 and 1 representing the AI's certainty in the extracted data. Scores below 0.80 are typically flagged for manual review.
2. Fraud Flags: Boolean indicators (e.g., has_fraud_flags) triggered by suspicious patterns like high cash-to-revenue ratios or regular round-sum transfers.
3. Volatility Metrics: The cashflow_volatility index measures the stability of monthly inflows, aiding in risk assessment for seasonal businesses.