Archival Data
Migration Guide

Introduction: Migration is More Than Just "Copy-Paste"

As an archivist, you know that your data is your most valuable asset. Moving that data from a legacy system (like Archivists' Toolkit, older DB textworks, or even Excel spreadsheets) to a modern platform like AtoM (Access to Memory) or ArchivesSpace is rarely a simple "export/import" job.

It is an archival project, not just a technical one.

At USA Archives Hosting, we have overseen migrations for complex, multi-repository institutions. We wrote this guide to help you understand the path ahead—and how to avoid the common pitfalls that lead to data loss or "dirty" descriptions.

Phase 1: The Audit (Know Your Data)

Before you install any software, you must audit your current metadata.

• Inconsistent Dates: Do you have dates entered as "circa 1900," "c.1900," and "1900?" Modern systems rely on ISO 8601 standards (YYYY-MM-DD). These need to be normalized.
• Orphaned Records: Identify records that have lost their link to a parent collection or series.
• Controlled Vocabularies: Are your subject headings free-text or linked to a standard like LCSH? Migration is the perfect time to reconcile these.

Phase 2: The Mapping Strategy

This is where the "Archivist Advantage" comes in. You need to map your old fields to the new standards (ISAD(G) or DACS).

If moving to AtoM:

AtoM relies heavily on the slug (URL) and reference code.

• Tip: Ensure your reference_code column in your CSV import is unique across the entire database, not just the collection.
• Hierarchy: AtoM builds hierarchy (Fonds -> Series -> File) via the parentId column. If this link is broken in your spreadsheet, your tree structure will collapse.

If moving to ArchivesSpace:

ArchivesSpace is stricter about agent links (People/Corporate Bodies).

• Tip: You cannot just type a name into a "Creator" field. You often need to create the Agent record first, obtain its URI, and link it.
• Dates: ArchivesSpace splits dates into "Expression" (what the public sees, e.g., "Spring 1940") and "Begin/End" (machine-readable, e.g., "1940-03-01").

Phase 3: The "Cleanup" (Sanitization)

We use custom Python scripts to clean data before it ever touches the new server. Common tasks include:

1. Character Encoding: Fixing "Mojibake" (broken characters like é instead of é).
2. Whitespace Removal: Stripping trailing spaces that cause search errors.
3. HTML Stripping: Removing old formatting tags (<b>, <font>) that break modern themes.

Phase 4: The Test Import (Sandbox)

Never import directly into production. We set up a "Sandbox" environment for our clients. We run the import, and then you (the archivist) browse the collection.

• Does the hierarchy look right?
• Are the digital objects attached to the correct descriptions?
• Did the notes fields migrate?