Review source links after combining page outputs

Treat evidence links as data that can be wrong

Verify those references after the merge so a faster assembly step does not make the result slower to review.

For multipage invoices, use ParseMill to review extracted invoice data beside the source and export separate summary and line-item sheets. Include a page break in your trial and check that every required line made it through. Other report types need their own fit assessment.

Preserve a compound source identity before combining: document version, physical page and source row or region. Add a separate output row identifier for the combined register. Do not overwrite one with the other.

Use a source-link worksheet

These synthetic records expose two shortcuts that fail: matching only a common value and matching only page-plus-row coordinates. A13 must not point to A12 merely because both report 7.2. B01 must retain the R4 file identity even though its local coordinates resemble OUT01.

W3C: PROV Data Model distinguishes provenance entities and activities. Use that separation to record both the source document and the combining operation that produced the register.

Check the combined output

1. Freeze the separate page outputs and the combined workbook version being reviewed. Record the mapping used to carry source fields forward.
2. For each selected output record, open its source reference and compare identifier, field heading, value and neighboring context. Record link failure separately from value mismatch.
3. Include records from every page boundary and each input file. If a mapping defect is found, inspect every record affected by that transformation.
4. Repeat the check after sorting, filtering or manually moving rows. Keep evidence columns inside the same table range as their records.

Microsoft: Table.Combine appends tables into one output. Carrying explicit source columns through that operation makes the relationship reviewable; the append itself does not verify the relationship.

Classify failures for repair

A broken link, a link to the wrong version and a link to the correct page but wrong row require different fixes. Keep these categories in the worksheet. If links are restricted to the author’s computer, provide an accessible source package or an agreed reference manifest rather than calling them validated for recipients.

After repair, test a record known to have failed and neighboring records that might have shifted. A global replacement of a page number can create new errors on pages that were already correct.

Make evidence resolution part of acceptance

For the fixture, all three output rows must resolve to their listed file and page, with the expected identifier visible. Matching the numerical values alone is insufficient. Save the worksheet alongside the exact workbook revision so later edits do not inherit an outdated signoff.

After exporting invoices from ParseMill, ask a recipient to locate a repeated amount using the workbook’s source references and an authorized copy of the PDF. Include that handoff time when comparing the product with your current spreadsheet process.