You cannot migrate what you have not measured. An audit is a full inventory of the source AtoM instance - descriptions, authority records, terms, digital objects and their relationships - while mapping is the crosswalk that decides where every field lands in Heratio. Skipping either step is the single most common cause of silent data loss during migration.

Treat this as a records exercise governed by ISO 16175 (principles for records in digital environments) and the description standards themselves: know your source model before you touch the target.

Inventory the source

  • Count information objects by level of description, and note the depth of your hierarchy (fonds, series, file, item).
  • Enumerate authority records (ISAAR-CPF actors), repositories (ISDIAH), and every taxonomy term and place/subject access point.
  • List digital objects, their derivatives (reference, thumbnail, master) and total storage footprint.
  • Capture relationships: parent/child hierarchy, actor-to-record links, and the generic relation table entries.

Build the field-level crosswalk

Map each AtoM field to its Heratio destination. Most ISAD(G) elements map one-to-one, but decide deliberately how EAD-exported structures, Dublin Core equivalents and RiC-O entities should be represented. Where a target field does not exist, the crosswalk must say explicitly whether the value is transformed, merged, held in a note, or intentionally dropped - a documented drop is a decision, an undocumented one is a defect.

In practice. Keep the crosswalk in a shared spreadsheet with columns for source element, target element, transformation rule, and owner. Add a "sample value" column and paste a real record into it - abstract mappings look correct until you test them against messy production data with empty fields, HTML in notes and inconsistent date formats.

Assess data quality early

Profile the source for the problems that will bite during import: duplicate authority records, inconsistent date encodings, orphaned digital objects, non-unique slugs and access points that differ only by casing or whitespace. Quantify each issue now so the transform lesson has a concrete backlog rather than surprises.

Scope and sign-off

Confirm what is in and out of scope - draft records, private descriptions, deleted-but-retained items - and get the collection owner to sign the mapping. This baseline is what you validate against after cutover.

Key takeaways

  • Audit produces a complete inventory; mapping produces a field-level crosswalk with an owner for every rule.
  • Every dropped or merged field must be a documented decision, never a silent loss.
  • Profile data quality now so remediation is a planned backlog, not a cutover surprise.

Check your understanding

Last modified: Saturday, 1 August 2026, 11:10 AM