Invoice Data Extraction
This test procedure extracts key information from invoice documents for accounts payable processing.
Objectives
Accurately identify invoice metadataExtract all financial figuresFlag any discrepancies or missing informationDocument Handling
Process each invoice individuallyIf a document contains multiple invoices, extract each separatelyFor illegible sections, mark as [ILLEGIBLE]Quality Standards
All amounts must include currency symbolsDates must be validated for logical consistencyVendor information should match against known vendor list when possibleDetailed Field Extraction Rules
Invoice Number Extraction
The invoice number is a critical identifier. Follow these steps:
Check the top-right corner of the document firstLook for labels such as "Invoice #", "Invoice No.", "Inv Number", or "Reference"If multiple potential invoice numbers exist, prefer the one in the headerInvoice numbers may contain letters, numbers, and hyphensDo not confuse with PO numbers or order referencesDate Field Processing
Multiple dates may appear on an invoice:
Invoice Date: When the invoice was createdDue Date: When payment is expectedService Date: When services were renderedShip Date: When goods were shippedFor each date field:
Convert to MM/DD/YYYY formatValidate that dates are logical (e.g., due date after invoice date)If only partial date is visible, mark as [PARTIAL: visible portion]Amount Calculations
When extracting monetary amounts:
Always include the currency symbol or codePreserve decimal precision as shown on documentFor line items, verify that: quantity x unit price = line totalSum of line totals should equal subtotalSubtotal + tax + fees - discounts = totalTax Handling
Tax information varies by jurisdiction:
Look for tax rate percentagesIdentify tax type (sales tax, VAT, GST, etc.)Some invoices show multiple tax linesTax-exempt items should be notedVendor Information
Extract complete vendor details:
Company name (official registered name)Address (street, city, state/province, postal code, country)Phone numberEmail addressTax ID / VAT number if presentBank details if shownCustomer/Bill-To Information
Similarly extract recipient details:
Company or individual nameBilling addressShipping address (if different)Customer account numberContact informationEdge Cases and Special Handling
Multi-Page Invoices
For invoices spanning multiple pages:
Look for "Page X of Y" indicatorsContinuation markers like "Continued on next page"Running totals vs. final totalsEnsure all line items are capturedUse final page total as the authoritative amountCredit Notes and Adjustments
Credit notes require special handling:
Negative amounts should be preserved as negativeReference to original invoice should be capturedReason for credit if statedForeign Currency
When invoices show foreign currency:
Extract amount in original currencyNote exchange rate if providedDo not perform currency conversion unless explicitly instructedHandwritten Annotations
If handwritten notes appear on the invoice:
Note their presence in a comments fieldDo not treat handwritten amounts as authoritativeFlag for human review if annotations conflict with printed dataValidation Rules
Automatic Checks
The following validations are performed automatically:
Date format consistencyMathematical accuracy of totalsRequired field completenessCharacter encoding validityManual Review Triggers
Flag for manual review when:
Confidence score below 85%Any field marked as illegibleCalculated totals don't match stated totalsDocument quality is poorMultiple invoices detected in single document