Preserve a batch or lot identifier that sits outside the table

Outside the table does not mean global context

Keep it with the observations so the register can still answer which material the results describe.

Our public material-register example shows a synthetic report beside downloadable output. Use it to inspect how values and source references travel together. The self-service ParseMill trial is for invoice PDFs; supplier test reports and engineering review need a separate fit discussion.

Keep batch, lot, sample and report identifiers in separate fields unless the source explicitly equates them. A familiar code format or a nearby label is not enough to infer that they represent the same thing.

Use a batch-context worksheet

The identifiers are invented. Preserve leading zeros, punctuation and any uncertainty in the transcription. B3 should not become an accepted batch value for nearby rows until both its characters and scope are supported by the source or a documented clarification.

Microsoft: Keeping leading zeros and large numbers describes text-preserving import for numeric-looking codes. Treat a lot identifier as text even when most of its characters are digits.

Capture context before propagation

1. Inventory every batch or lot label across the source, including covers and section headings. Record physical page and precise location for each occurrence.
2. Transcribe the identifier exactly and distinguish the label’s role from sample or report identifiers. Keep uncertain characters explicit.
3. Determine the governed observation range from source wording or verified structure. Create a context-to-observation mapping rather than an undocumented fill-down.
4. Check records at each boundary and each new report segment. Reopen the export to verify that the identifier remains text and attached to the correct rows.

W3C: Provenance model primer supports explicit provenance relationships. The mapping should preserve the source context record and the reviewer’s evidence for assigning it to observations.

Watch unsupported automatic propagation

A batch on the first page may describe only the first listed material. A header can persist across a reused template even after a section changes. Inspect those possibilities before treating the value as document-wide context.

If several batch labels conflict, retain them with their locations and request clarification. Do not choose the first or most frequent value automatically. Record which observations are held from batch-based matching while the conflict remains unresolved. Include the conflicting labels and their page locations in the clarification request so the issuer can resolve the specific boundary.

Check the assignment is reproducible

For the fixture, S1–S4 receive B1’s confirmed lot and S5–S7 receive B2’s. B3 stays unresolved rather than contaminating the next rows. A reviewer should be able to follow every accepted assignment to the exact source label.

For a supplier-report fit assessment, include out-of-table lot labels and the observation ranges they govern. Agree which context would be extracted and which assignments need manual review before a pilot. Handwritten or uncertain labels are not a capability demonstrated by the invoice trial.