Read a PDF table printed in side-by-side continuation blocks

Identify whether the page contains parallel or continuing blocks

Check the continuation rule so the spreadsheet follows the records instead of simply reading across the page.

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.

The visual arrangement can defeat a simple left-to-right text stream. A copied line may interleave one record from each block, pairing an identifier on the left with an amount on the right. Keep the blocks separately labeled until their relationship is established. A clean spreadsheet preview is not evidence that the reading sequence is correct.

Build a block-order map

Microsoft: PDF connector supports importing tables detected in a PDF. Microsoft: Table.AddIndexColumn can retain positions within each imported block; those positions need a block reference to remain meaningful.

1. Label each visible block using physical page and location, such as P2-left and P2-right. Record the headings and first and last identifiers.
2. Decide whether the blocks are independent or sequential. Look for explicit numbering or a caption, and record the source evidence for the decision.
3. Inspect a complete row in each block. Confirm that all fields come from the same visual row rather than a line spanning both blocks.
4. If the sequence is supported, append the blocks in the reviewed order while retaining block and local-row references. If not, transcribe manually or retain separate outputs with an unresolved relationship.

Compare two possible readings

Synthetic example: one page displays a numbered list in two vertical blocks.

The intended order is R01 through R06 when the source explicitly continues the numbered list from left block to right block. A horizontal reading could instead produce R01, R04, R02, R05, R03, R06. Worse, it could combine R01 with the amount belonging to R04. Retain local block positions so both ordering and field association can be checked. If the source labels the right block as another report, no sequential append is justified.

Do not repair ambiguity with a convenient sort

Sorting the extracted identifiers may hide an interleaving error without correcting the associated values. Before sorting, verify that each identifier still has its own description, amount and unit. A valid-looking ascending sequence can contain rows assembled from the wrong block.

Also check footnotes that span the full page width and captions repeated above both blocks. Such text may apply to both tables but is not itself another row. Record the note relationship separately. When block boundaries or field positions are ambiguous, preserve an unsupported-layout outcome instead of guessing a flat table.

Verify the block handoff

Count source records by block and reconcile them with the retained output. Check the last record in the first block, the first record in the second, and every field around that boundary. A reviewer should be able to move from any output row back to one block and one visible source row.

The public synthetic register demonstrates a simpler supported table shape. It does not prove support for side-by-side continuation layouts. Use a manual block map or a more suitable tool when the current extraction workflow cannot preserve that structure reliably.