Archive a completed extraction with enough context to reproduce it

Preserve the chain of work

Keep the source version, output and review context together so a later question does not become a new extraction project.

ParseMill exports reviewed invoices to Excel with separate summaries, line items, sources and review status. You can import an XLSX or CSV template and save a client setup. Run a real invoice through to the final workbook and check it in the application that receives your work.

Follow the organization’s retention and access rules for source documents. An archive manifest can reference an authorized repository rather than duplicating sensitive files into an uncontrolled folder.

Copy an archive manifest

The identifiers are illustrative. Fill the manifest with actual paths or stable references, and test that the authorized reviewer can resolve them. Include dates as factual event dates rather than deriving them from filenames alone.

W3C: Data on the Web Best Practices recommends version information and history. A completed extraction benefits from the same explicit relationships between its successive artifacts.

Assemble and verify the archive

1. Identify the exact source and delivered output versions after final review. Record hashes where byte-level identity matters.
2. Save configuration, schema, transformation rules and execution information available for the run. Document external service versions or unknown dependencies honestly.
3. Retain correction logs, unresolved exceptions and the scoped signoff. Link each to the output revision it actually covers.
4. Perform a retrieval check: open the manifest, resolve the authorized source and final export, and reproduce a selected record’s evidence trail. Record any missing dependency.

W3C: Provenance model primer models entities, activities and agents. Use those distinctions to explain which artifacts existed, which actions transformed them and who reviewed the result.

Be precise about reproducibility

A cloud service may change, a model version may be unavailable or a manual interpretation may require human judgment. Record those limits rather than promising an identical rerun. The archive can still preserve the original output and evidence needed to assess the completed work.

Do not archive only the cleaned workbook and delete the original export. The baseline helps distinguish parser behavior from later corrections if an issue is discovered. Keep access scoped to the people authorized to inspect the underlying material.

Check the archive is usable

For the fixture, a reviewer can follow Workbook 27C back to Register 27B, Export 27A and Report R51 revision 3, including the two exceptions. The configuration and review record explain what was done and what remains outside scope.

Use this manifest when completing a ParseMill extraction project. It turns a finished file into a reviewable record of work that your team can revisit when sources change or questions arise.