1. Inspect the source and the intended columns

The source is a CC0 synthetic fixture generated by ParseMill for testing. It contains no customer data and is not a real supplier certificate. Both pages repeat the exact headers Sample, Test, Result, Unit, Method.

2. Inspect the extracted register

Five data rows, two source pages. Values below are the raw extracted strings. The source page numbers are displayed from 1; the downloadable JSON uses zero-based page indexes.

Precomputed output · synthetic data · all cells unreviewed
SampleTestResultUnitMethodSource PDFReview
S-001Tensile strength42.60MPaISO 527Page 1Unreviewed
S-001Elongation18.2%ISO 527Page 1Unreviewed
S-002Moisture< 0.0100%ISO 15512Page 1Unreviewed
S-002Residue0mgInternalPage 2Unreviewed
00042Density1,234g/cm3ISO 1183Page 2Needs interpretation
How to check a source reference

Open the page link on the Moisture row. Find Sample S-002 and read the Result and Unit cells together: < 0.0100 and %. Check the qualifier, all four decimal places and the column alignment. A link to the right page does not by itself prove the value was extracted correctly.

The signed-in workspace highlights individual source words. This precomputed example provides page links and the raw JSON evidence for inspection.

3. Review what the parser cannot decide

Needs review: Density. The raw result 1,234 is preserved. The report deliberately leaves the comma’s meaning ambiguous. The parser marks numeric interpretation unsupported; it does not choose between 1.234 and 1234.

The other rows also remain unreviewed. Compare the moisture qualifier < 0.0100, explicit residue value 0, leading-zero identifier 00042, and units with the PDF. Source-page links locate the page; they are not an approval of the result.

The signed-in workspace offers cell-level source highlights and saved corrections. This public example uses a static table and source-page links, so it can be inspected before registration.

4. Download the precomputed output

The workbook contains Data, Sources and Exceptions sheets. Public sample source links are adapted to the downloadable fixture PDF; private workspace exports require access to their original source. All values are still unreviewed. CSV import can change numeric-looking identifiers, so verify 00042 after opening it.

5. Use the workflow for your own document

Invited users can open a workspace prefilled with these five column names, add a suitable PDF, review the result and export it. A new document is a new extraction and needs its own checks. Save the columns if you process more reports with the same layout.

Processing requires an invited account and an active allowance. Files may be up to 20 MB and 300 pages. If you do not yet have access, request an invitation; a request does not create an account.

Current product scope. ParseMill reads native-text, aligned tables with exact, single-word column headers. It does not currently provide OCR, arbitrary layout understanding, unit conversion, or automatic comparison across supplier documents.

Continue the task

Try your own suitable PDF.

Open the matching workflow with the sample columns prefilled. Verify your email to create an account; eligible workspaces receive 10 processing pages for 7 days.

Use these columns in ParseMillSee free-trial details

Sources and evidence

Public examples are synthetic. Each example and evaluation describes its tested inputs and limitations. This page is authored by ParseMill.