Start with a consistent sheet structure
OCR is much easier to operationalize when every invoice lands in the same column structure. Decide whether you need one row per invoice, one row per line item, or two linked sheets. A summary sheet works well for payment tracking; a line-item sheet is better for spend analysis.
The free invoice column template builder can generate a starting header row.
Invoice OCR workflow
- Collect a clear source. Use the original PDF when possible. For photos, keep the invoice flat, fill the frame, and avoid shadows over totals or line items.
- Select spreadsheet extraction. Upload the invoice in Sheets mode. The OCR engine analyzes both table layout and common invoice fields.
- Verify identifiers first. Check vendor name, invoice number, purchase order, and dates. These fields drive deduplication and matching.
- Review financial values. Confirm currency, subtotal, tax, discounts, shipping, and total against the source document.
- Normalize the output. Apply consistent vendor names, date formats, account categories, and number formats before importing into another system.
One row per invoice or one row per item?
| Layout | Best for | Tradeoff |
|---|---|---|
| One row per invoice | Payment status, due dates, vendor totals | Does not preserve detailed purchases |
| One row per line item | Spend analysis, inventory, cost allocation | Invoice-level fields repeat across rows |
| Summary + items sheets | Auditable finance workflows | Requires a shared invoice ID between sheets |
Prevent duplicate invoices
Create a unique key from vendor name, invoice number, and invoice date. Before appending a new row, check whether that key already exists. Keep the source filename or a Drive link in the sheet so reviewers can trace each row back to the original document.