The task: get the rows without retyping the invoice
A supplier invoice arrives as a PDF, but your cost sheet or purchase register needs an item code, description, quantity, price and amount on each row. Copying the page as one block loses that structure. Typing every row again adds a second opportunity to change an identifier or decimal place.
ParseMill’s current invoice workflow extracts an aligned native-text table with confirmed column headers. It creates reviewable rows with source evidence, then exports Excel or CSV. It does not read arbitrary invoice layouts or decide which charges should be paid.
Check these conditions before uploading
| Check your PDF | A suitable starting point | Requires a different approach |
|---|---|---|
| Text | Words can be selected and copied | Scans or photographed pages need OCR first |
| Headers | Exact single-word labels, such as Sample or Result | Merged or multi-line column headings |
| Rows | Values align under their columns | Nested tables, prose specifications, or diagram labels |
| Continuation | The header repeats on each table page | Continuation pages without a recognisable header |
The invoice preset begins with Item, Description, Quantity, Price, Amount. Replace them with the exact single-word labels in your own PDF. Headers such as Unit price, a description wrapped onto several lines, or a scan require another approach. Keep one invoice per job.
Inspect four line items and one deliberate zero
The public worked example starts with a synthetic supplier invoice. Its four lines include Copper pipe, Brass elbow, Filter cartridge and Delivery. The delivery price and amount are explicit 0.00; those values must survive as zeros rather than becoming missing data. Item 00104 must retain its leading zeros.
Download the input and precomputed output before creating an account. The example shows this fixture’s behavior, not a claim about all supplier invoices. The evaluation includes separate fixtures and known failure cases.
One invoice, a checked export, then repeat
- Choose one representative invoice. Confirm selectable text, aligned rows and matching headers. Keep the source PDF.
- Use the invoice preset. Check the column names and available processing allowance before starting.
- Review the result. Open source references and compare item codes, quantities and decimals. Check the complete invoice for missing rows and repeated headers.
- Check business meaning separately. Reconcile expected totals, credits, tax and currency using your own accounting process. A successful extraction does not validate the invoice.
- Export and reuse. Download the reviewed revision as Excel or CSV and save the column set for another suitable document. Review each new invoice independently.
The Excel output contains Data, Sources and Exceptions sheets. Exported values are text, so identifiers and decimal precision remain inspectable. Convert to numbers only in a separate checked calculation step.
Make the next action the same task
Open the invoice workflow with these columns prefilled. You can run the synthetic sample in the workspace or process your own suitable PDF when your account has processing access. The current upload limits are 20 MB and 300 pages per PDF.
Working with test reports instead? The supplier-register workflow uses the same source review and export steps with a different schema.
Sources and evidence
- ParseMill synthetic supplier invoice (CC0) · Checked 2026-09-28
Public examples are synthetic. Each example and evaluation describes its tested inputs and limitations. This page is authored by ParseMill.