Invoice to CSV: choose columns and avoid broken imports

Choose the row structure before converting

An invoice-to-CSV workflow turns information from a PDF or image into named columns. Start by deciding what one row represents. For a purchase register, one row per invoice is usually convenient. For product analysis, one row per line item is a different structure and needs its own review.

Parzable’s batch CSV export creates a row for each completed file, with Filename, Processed At, and your schema’s fields. It does not automatically turn an arbitrary nested line-item result into a separate item table. Keep invoice-level and item-level datasets distinct.

A useful first schema is Vendor, Invoice number, Invoice date, Currency, Subtotal, Tax amount, and Total. Add fields because the destination needs them, rather than extracting everything and deciding later.

What makes a CSV import correctly?

CSV is plain text with records and fields. Under the common conventions documented in RFC 4180, a field containing a comma, quote, or line break is enclosed in double quotes; an embedded double quote is doubled. Those characters are part of the file format, not invoice data to delete.

"Vendor","Invoice number","Currency","Total"
"Example Supplier, Ltd","000482","USD","1248.50"
"Example Studio","INV-0083","EUR","240.00"

These fictional rows demonstrate why splitting every line at every comma is unsafe: the first vendor name contains a comma. Use a spreadsheet’s CSV importer or a CSV parser, not a text replacement that removes quotation marks.

Download the example CSV and check its preview before using your own invoices. Select UTF-8 when your importer offers an encoding choice, and check supplier names with accents or non-English characters.

Create and check a Parzable export

1. Create a schema in Schemas. Use stable names that you can map to the destination’s columns.
2. In Extract, choose the schema and upload the invoice files.
3. Wait for processing to finish. Note any errors; an export of completed results may leave failed files out.
4. Download the CSV and compare the result count with the completed file count.
5. Import into a staging sheet, review it, and map it to the receiving system’s required layout.

CSV compatibility does not prove compatibility with a particular accounting system’s bill importer. Those importers may need specific vendor identifiers, account codes, tax codes, and column names. Use the receiving system’s own template and validate a small import before scaling.

Treat missing and transformed values explicitly

The extraction prompt instructs Parzable to return N/A when a requested value is unavailable. Check that value before importing into a numeric or date field. Do not replace missing tax, discounts, or quantities with zero unless a reviewer has established that zero is correct.

Preserve invoice numbers as text. Keep currency in a separate column when combining suppliers. Retain the raw date if its meaning is uncertain. If a description starts with a spreadsheet formula character, import it as text rather than allowing it to execute as a formula.

Finish with a simple reconciliation: source files received, files completed, files failed, rows exported, and rows approved. Those counts explain what your CSV contains and what still needs attention.