Keep a table section separate when its row meaning changes

Read the section title as part of the schema

Keep that change visible so combining tables does not create a register that looks consistent but mixes incompatible records.

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.

Write a one-sentence definition of a row for each section before combining the data. Include the entity measured, the observation type and any stated basis. If those definitions differ, keep separate tables or add an explicit section and observation-type field. Do not let the convenience of one worksheet erase a distinction that matters to interpretation.

Record the section boundary

Microsoft: Table.Combine appends input tables. W3C: PROV Data Model provides a model for recording origins; your section reference is a practical way to retain that context.

1. Locate the section title and its physical page range. Record the exact title and any explanatory sentence defining the table.
2. Identify what one row represents: an individual sample, a repeat test, an average, a limit or another stated object. Keep uncertainty about the role explicit.
3. Compare field definitions, units, method labels and result basis across sections. Same spelling is not enough if the definition differs.
4. Choose separate outputs when the row definitions cannot be reconciled. If a combined register is useful, preserve section identity and the confirmed row role on every contribution.

Use a section-boundary log

Synthetic example: these sections reuse the same three headings in one report.

A row reading “G1, 12.0, MPa” cannot be interpreted correctly without its section. In the first table it could be an observation; in the second it could summarize several observations; in the third it may be a reference value. Appending all three as ordinary measurements would create a misleading analysis even if every character was copied correctly.

Handle reused identifiers carefully

A sample label can restart in each section. A key built only from Sample may therefore collide across tables that describe different populations. Use the report version, section reference and local row identity together until a qualified reviewer confirms a stronger relationship. Do not deduplicate the collision automatically.

A section title may appear on a preceding divider rather than directly above the table. Preserve that source location alongside the table range. If the title is inferred from the file name instead of stated in the document, label it as an unconfirmed interpretation and ask the report owner before using it to classify results.

Check the register by meaning as well as count

Review a row from each section and ask whether its output still communicates what was measured or summarized. Confirm that reference values cannot be mistaken for observations in a downstream filter or chart. Check that all retained rows carry a section reference and that any omitted section is named in the coverage statement.

The public material-register example gives you source and output to inspect before a broader document-workflow discussion. Use the section differences in your own reports to define that discussion. Supplier-report work needs an agreed scope; the self-service trial is for invoice PDFs.