Invoice line item extraction: preserve rows and reconcile totals

Header fields and line items need different structures

Invoice line item extraction captures the repeated products or services inside an invoice: descriptions, quantities, unit prices, and printed line amounts. A correct invoice total does not prove that all item rows were captured, and a complete-looking table does not prove the rows still align.

Decide whether your destination needs one record per invoice or one record per item. In an item table, repeat an invoice identifier on each row so two items from one invoice remain distinguishable from two separate invoices.

Parzable supports custom field descriptions, and the homepage describes line-item capture. Its current schema editor offers scalar field types rather than a dedicated nested item-table editor. The batch CSV has one row per completed file; complex values may be serialized inside a cell. Test the result shape before relying on an item-by-item import.

A worked item table

Fictional example: invoice DEMO-104 contains two items. Both have to retain their own quantity and unit price. The following table is a target layout you can use for review, not a promise that Parzable exports this exact layout.

The line amounts sum to 75.00. If the source also prints tax of 15.00 and an invoice total of 90.00, those values can be reconciled separately. Do not add the tax to each item unless the source and target layout require that representation.

Download the fictional line-item CSV to inspect that target layout. It is independent of the app’s native batch export.

Describe the items, then inspect the actual result

For an exploratory schema, a string field named Line items can use a description such as: “List every printed product or service row in order, retaining description, quantity, unit price, and printed line amount. Include continuation pages. Do not calculate missing values.” This requests the information, but it does not create a typed array or guarantee a separate CSV row for each item.

Run a short invoice first. Inspect the result in the interface and in the downloaded CSV. If your receiving system requires a normalized item table, transform and review the result in a separate staging sheet. Do not assume that a string field containing several items will import as several records.

Keep the source invoice beside the output while reviewing. Wrapped descriptions can look like extra rows, and repeated table headers on a second page can look like items.

Four checks for repeated rows

1. Count: compare source item rows with extracted items, excluding table headers and carried-forward totals.
2. Alignment: confirm each quantity, price, and amount still belongs to the correct description.
3. Continuation: inspect the last item before and first item after each page break for omissions or repeats.
4. Arithmetic: compare printed line amounts with the invoice summary, accounting for discounts and charges shown separately.

When the shape or row count is wrong, hold that output for manual review. A small successful example is a reason to test harder cases, rather than evidence of universal support for every supplier’s table.