Keep invoice subtotal and tax summary out of line-item rows

Where the extra work starts

Keep summaries separate from the purchased rows so the spreadsheet is ready for review instead of another round of manual cleanup.

Upload a representative supplier invoice, inspect its extracted fields and line items beside the PDF, then download Excel. Put the awkward invoice in your trial too: the useful result is less total work after corrections.

Retain summary information as document context where needed. Excluding it from a line-item register does not mean ignoring it or making a judgment about its accounting treatment.

A practical way to get it right

Inspect the transition from the final item to the summary area. Look for labels, spacing, rules and changes in which columns are populated. Record the last actual item and the summary elements that follow. Use the source's meaning as well as visual position; a delivery charge may be a genuine item line on one document and a document-level adjustment on another.

Create an expected-row list before comparing extraction output. Check the final few rows especially carefully. If a summary was captured as an item, record the exclusion reason and source location rather than silently deleting it without explanation.

Keep any arithmetic comparison with totals separate from the item transcription. Do not invent quantity one or an item code to make a summary line fit the row schema. Preserve ambiguity for the reviewer when the layout does not make the distinction clear.

OpenPeppol: Invoice syntax separates document totals and allowances from individual lines. Use that structural distinction as an aid when documenting your PDF’s summary exclusions.

What this looks like on a real task

Synthetic example. The people, references and figures below are invented to demonstrate the method.

The five printed rows shown contribute two item rows to this register. The earlier item rows are outside this excerpt, so the table is not a full invoice count. Copy the printed-row and treatment columns and identify the complete expected item range in your own source before validating an export.

Catch the cases that create rework

Some documents repeat a carry-forward subtotal at page boundaries. Do not treat it as an additional item or assume it is the final document total. Inspect the next page and preserve the relationship. A blank item-code cell alone is not enough to classify a row, since legitimate descriptions or charges can also lack codes.

If extraction consistently cannot distinguish the boundary, use a controlled manual review or another method rather than accepting an inflated row set.

OpenPeppol: Invoice line syntax describes individual line fields. A summary amount without a corresponding item relationship should not be forced into those fields merely to preserve a rectangular table.

Decide whether the change is worth keeping

Check both directions: every source item has an output row, and every output row maps to a genuine source item. Keep a short exclusions list for summaries and repeated headers. That bidirectional check catches both missing items and invented extras, while leaving document-level interpretation with the responsible reviewer.