Review a continuation page whose column widths changed

Start with meaning, not horizontal position

Review the transition before accepting the combined output so a layout change does not become a hidden row-association error.

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.

Microsoft: PDF connector describes importing detected PDF tables. Detection provides a starting point for review; it does not establish that each detected column has the intended meaning.

Use this synthetic width-change fixture

These invented rows describe a test fixture, not a supplier document or a measured product result. Page 2 allocates more room to Description. The expected mapping follows the heading and visible cell boundary, even where the old horizontal coordinate now falls inside another field.

A numeric-only check would miss a displaced qualitative result. Include A9 when checking the new page, because it exercises a different value shape from the two density rows.

Build a column-by-column check

1. Save separate page outputs before combining them. Add physical page and source row identifiers while their origins remain clear.
2. For the first, middle and last visible rows on the changed page, compare every output field with its printed heading and cell. Expand to the entire page if any association is wrong.
3. Write an explicit mapping from each detected column to the agreed field name. Keep an unexpected extra column for investigation instead of discarding it silently.
4. Reopen the combined output and verify the corrected mapping on both pages. Confirm that the earlier page did not inherit the later page’s offset.

Microsoft: Table.SelectColumns documents selecting named columns in a specified order. That operation can express an already verified mapping; selecting a familiar name cannot repair a wrongly assigned cell by itself.

Failures worth recording

Do not split a long description solely because it crosses the previous page’s column boundary. Conversely, do not join a unit into the description just because the whitespace is narrow. Watch for a heading that spans two subcolumns: a visually small change may introduce a genuinely new field rather than merely a different width.

Record each correction as source page, source row, old output field, correct field and evidence. If a cell boundary is unreadable, preserve the raw text and mark the association unresolved. A clean-looking table is not a reason to guess.

Accept the joined output deliberately

For this fixture, acceptance means three records remain, A8 has result 1.3 with its own unit, and A9 retains Uniform as a result. No value should move to a neighboring record. Retain the mapping note with the export so a later refresh can be checked against the same boundary.

For an invoice buying test, choose a supplier invoice whose column widths change after a page break. Review the first and last line on the changed page, then open the workbook. Use the material fixture above as an acceptance question for a separately scoped report workflow.