Inventory sensitive content locations
List the fields and free-text areas likely to contain names, contact information, account references, confidential terms, or signatures. Do not limit the review to a spreadsheet column list. Scans, headers, attachments, embedded images, and document metadata can contain information that is absent from an extracted text file.
Specify the intended treatment
For each category, choose exclusion, masking, replacement, restricted access, or retention subject to project review. Explain whether relationships must remain usable after transformation. Replacing every customer identifier independently can break linked-document tasks, while a consistent project-scoped replacement can preserve necessary associations without publishing the original value.
Check the resulting artifacts
Review both machine-readable outputs and visible document pages. Confirm that a hidden text layer does not retain values removed from an image and that a transformed attachment matches its manifest entry. Record review coverage and unresolved findings; a processing step alone does not establish that the entire collection is suitable.
Resolve scope before delivery
Agree on who reviews exceptions, which artifacts are withheld, and how revisions are versioned. This is a workflow specification, not a declaration that a source is cleared for every purpose. Bring the intended task and handling requirements to Fiund so each potential source can be assessed within that scope. Choose a pilot document with a visible name in both the page image and the extracted text. Check all intended outputs after the proposed treatment, including previews and annotations. This exercise makes the review concrete and helps identify derivative artifacts that otherwise fall outside a narrowly defined masking or exclusion step.
Review checklist
- Inspect text, images, and metadata
- Define treatment by field category
- Verify relationships after transformation
- Record review scope and unresolved exceptions