Identify what counts as an item
Repeated headers, continuation text, subtotals and notes can look like table rows. Define which rows belong in the output for your document class. A wrapped description may occupy several visual lines but represent one item. Conversely, identical descriptions can belong to different real items. Do not remove apparent duplicates solely because their text matches; use the document context and an appropriate identifier where one exists.
Use a known small example
Consider a fictional table with A01, two boxes; B02, seven pieces; and a subtotal line. The expected output has two item records, not three. A parser might return two records yet pair the seven with A01 and the two with B02. That passes a count check and fails the task. Keep expected code, quantity and unit pairings in the acceptance fixture so a reviewer can locate the mismatch.
Inspect page boundaries
Check the final item on one page and the first on the next, along with repeated headers and carried descriptions. Record any source row that cannot be confidently linked to an output row. Do not use a grand total as the sole test: multiple errors can offset each other. A useful review combines counts, identities, field-level expectations and the original document, while recognizing that not every source supplies a unique row identifier.
Route unresolved structure deliberately
If the method cannot consistently separate item rows from notes, keep that document class in review. A manual correction should retain what changed and why; silently overwriting the output hides the evidence needed to improve the route. Continue to exception queues to assign ownership, or to the field contract if different reviewers disagree about the correct structure. Do not connect a parser directly to purchasing or accounting decisions on this guide alone.
Sources used for this page
These records support the facts and comparisons above. Merchant-controlled records are labelled so you can separate product claims from independent evidence.
- Parsio: documented parsing options — Merchant documentation · parsio.io · Merchant-controlled · checked 2026-09-30
- Microsoft: Power Query PDF connector — Platform documentation · learn.microsoft.com · Merchant-controlled · checked 2026-09-30