Multi-page table repair
Merge PDF tables across pages without duplicating rows
Combine repeated table fragments while keeping page references, repeated headers, and continuation rows under control.
Keep page identity until the table is checked
A table spread across pages often becomes several output fragments. Before combining them, attach the source filename, page number, and table order to each fragment. Those references make it possible to locate a missing or duplicated record after the layout has been flattened into one worksheet.
Save an untouched extraction. Work on a separate version so that a cleanup rule can be reviewed or reversed. Start with two adjoining pages that contain a real table boundary. A method that works only on the first page does not answer the main question of whether the records continue correctly.
Decide which fragments belong together
Compare the meaning of the headings, the units, and the kind of record in each table. Similar column widths do not establish that two tables share a schema. A supplier might place a delivery-charge table beneath its product list, or change from retail to trade pricing partway through the document.
Microsoft's Power Query PDF connector documents options for multi-page tables and cautions about large or multi-line structures. Treat automatic collection as a starting point. Inspect whether it has joined the correct sections and preserved the association between identifiers and amounts. Do not assume that fewer output tables necessarily means a better result.
Normalize headers before appending
Give corresponding columns the same names and types. Keep distinct concepts distinct: list price is not discounted price, and invoice quantity is not pack size. Microsoft documents that Power Query appends tables by matching header names rather than relative column positions. This makes consistent naming a practical control.
Remove repeated page headers only when they match an identified header pattern. A row containing a word such as Total is not automatically disposable; it may be an actual product description or a summary needed for reconciliation. Keep a record of filtered rows during the pilot so that accidental deletions are visible.
Handle continuation rows carefully
Inspect descriptions or records split at page breaks. Join fragments only when source evidence establishes that they represent the same item. A blank SKU can indicate a continuation, but it can also indicate a note, an adjustment, or a recognition failure. The rule should use more than one visual clue.
Compare the last complete record before the break with the first complete record after it. Check whether a line has been repeated by the extraction process or deliberately repeated in the source. Flag suspected duplicates for review rather than deleting every repeated SKU, because multiple price breaks or invoice lines can legitimately share an identifier.
Prove the combined table is complete
Count the source records for the sample pages and compare them with the combined result. Verify the first and last data rows on every page. Where meaningful, reconcile numeric summaries, while remembering that equal totals can hide offsetting errors or misplaced amounts.
Only then remove unnecessary visual spacing and prepare the final table. Keep provenance columns in the working version even if a customer-facing export omits them. A compact worksheet is useful; a compact worksheet whose rows can be traced back to the source is far easier to maintain when the supplier changes its layout.
Sources
Evidence status: methodology. No unverified accuracy, savings or traffic claim is made.