Data Extraction & FinanceUpdated October 2026

Financial Statement to Validated Structured JSON

Transform complex, messy tabular financial data and earnings releases into standardized, machine-readable JSON schemas with mathematical verification.

Target Users: Financial Analysts, Fintech Developers, Data Engineers, and Accounting Teams
Supported Workflow: Automated financial document processing, SEC filing extraction, and ERP data ingestion
Interactive Prompt Builder

Customize & Generate Your Prompt

Adjust the variables below. Your customized prompt updates live in real-time ready to copy with zero setup.

Input Variables (4)runs 100% locally
System / Role Prompt
You are a certified financial data engineer and CPA-grade accounting analyst. You extract tabular financial statements with zero numerical hallucination, enforce GAAP/IFRS terminology, and ensure strict JSON schema conformity.
Assembled User Prompt (Live Preview)1965 chars
You are extracting financial statement data into a validated, structured JSON schema.

### Task Specifications
- **Filing / Report Type**: Consolidated Statement of Operations / Income Statement (10-K)
- **Required Data Fields**:
- fiscalPeriod (e.g. "FY2025", "Q3 2025")
- currency (e.g. "USD", "INR")
- units (e.g. "thousands", "millions")
- totalRevenue
- costOfRevenue
- grossProfit
- researchAndDevelopmentExpense
- sellingGeneralAdministrativeExpense
- operatingIncome
- netIncome
- earningsPerShareBasic
- earningsPerShareDiluted
- **Policy for Missing or Not Reported Line Items**: Set value to null with an explanation in _audit.missingFields

### Raw Financial Text / Table
```text
CONSOLIDATED STATEMENTS OF OPERATIONS (In millions, except share amounts)
Three Months Ended September 30, 2025:
Total net sales: $94,930
Cost of sales: $51,120
Gross margin: $43,810
Operating expenses:
  Research and development: $7,760
  Selling, general and administrative: $6,540
Total operating expenses: $14,300
Operating income: $29,510
Other income/(expense), net: ($210)
Income before provision for income taxes: $29,300
Provision for income taxes: $4,650
Net income: $24,650
Diluted earnings per share: $1.64
```

### Extraction Rules & Constraints
1. **Strict JSON Output Only**:
   - Output valid JSON only, without conversational preamble or markdown postscripts.
2. **Numerical Accuracy**:
   - Extract numerical figures as standard numbers (not strings). Preserve exact currency units (e.g. millions vs thousands as specified in the header).
   - If a number is negative (often in parentheses `(1,234)`), represent it as a negative float/integer (`-1234`).
3. **Reconciliation & Audit Verification**:
   - Verify that: `Gross Profit = Total Revenue - Cost of Goods Sold` and `Operating Income = Gross Profit - Total Operating Expenses`.
   - Include an `_audit` metadata object in the JSON documenting any mathematical discrepancies found in the source text.

How to Use This Prompt Effectively

  1. 1Select the financial statement type (e.g. 10-K, 10-Q, Balance Sheet).
  2. 2Paste the raw text, OCR transcript, or HTML table directly into the raw data field.
  3. 3Copy the output JSON into your automated ingestion pipeline or data analytics notebook.

Representative AI Output

Sample output produced when using frontier AI models with the default parameters:

```json
{
  "company": "Acme Corp",
  "fiscalPeriod": "Three Months Ended September 30, 2025",
  "currency": "USD",
  "units": "millions",
  "data": {
    "totalRevenue": 94930,
    "costOfRevenue": 51120,
    "grossProfit": 43810,
    "researchAndDevelopmentExpense": 7760,
    "sellingGeneralAdministrativeExpense": 6540,
    "totalOperatingExpenses": 14300,
    "operatingIncome": 29510,
    "netIncome": 24650,
    "earningsPerShareDiluted": 1.64
  },
  "_audit": {
    "mathReconciliation": {
      "grossProfitValid": true,
      "operatingIncomeValid": true
    },
    "missingFields": []
  }
}
```

Pro Optimization & Edge-Case Tips

  • Always prompt the model to preserve negative signs when numbers are written inside parentheses like `(210)`.
  • Add an `_audit` key in your requested schema so the model self-checks its arithmetic before returning the final JSON.
  • Pair this prompt with our Document AI tool for direct PDF and image table parsing.