The Problem
The National Archives and Records Administration (NARA) uses a multi-stage quality gate system for digitized records entering the National Archives Catalog. Before any digital object reaches the transformation stage, it must pass compliance checks similar to a preflight checklist. These checks verify image file standards, authority rule conformity, and file-naming conventions.
When digital objects fail these checks, they're rejected before transformation work begins. Non-compliant records can't proceed to cloud upload, metadata pairing, or the machine-readable transformation that makes records queryable and portable across platforms.
Process Overview
The process follows a strict sequence:
- Cataloging and digitization: Records are arranged, described, and digitized at custodial units across NARA.
- Digital object packaging: Staff package digital objects with metadata, following technical and authority requirements.
- Transfer to Office of Innovation: Packaged data sets move to post-processing.
- Compliance checkpoint: All incoming digital objects undergo verification against technical standards.
- Rejection or advancement: Non-compliant objects stop here; compliant objects proceed to image file copying, cloud upload, and metadata pairing.
- Transformation: Data restructures from human-readable to machine-readable format.
- Import: Structured data enters the National Archives Catalog.
The compliance checkpoint between steps 4 and 5 is a hard stop. If your file-naming convention doesn't convey the intended structure, or if your image files don't meet technical specifications, the entire data set returns to the custodial unit.
Common Control Failures
Organizations that skip or inadequately implement pre-import compliance checks face three control failures:
Missing technical validation at the source: If custodial units don't verify image file standards before packaging, they discover failures only after transfer. This creates rework loops and delays access.
Inadequate authority control verification: Controlled vocabulary compliance is essential for pairing digital objects with their descriptions correctly. When authority rules aren't enforced at the packaging stage, the compliance checkpoint rejects the entire batch.
Weak file-naming governance: File names aren't arbitrary. They carry structural instructions: "these nine image files form a single nine-page document." If your naming convention doesn't encode this logic correctly, the transformation process can't execute. The system won't guess your intent.
Compliance Standards
ISO 14721 (the OAIS reference model) defines the framework NARA's process implements. The standard requires that Submission Information Packages meet specific technical and semantic requirements before they can become Archival Information Packages.
Specifically:
Preservation Description Information must be complete and compliant: This includes technical metadata about file formats, fixity data, and provenance information. The compliance checkpoint verifies this metadata exists and conforms to defined schemas.
Controlled vocabularies must be applied consistently: ISO 14721 requires that descriptive metadata use authority-controlled terms. Random or inconsistent terminology breaks the semantic relationships that make records discoverable.
Submission agreements must be honored: The standard assumes a documented agreement between the producer (custodial unit) and the archive (Office of Innovation) about what constitutes a valid submission. The compliance checks enforce this agreement. There's no negotiation at the checkpoint, either your submission meets the documented requirements or it doesn't.
The Generally Accepted Recordkeeping Principles reinforce this approach. The Principle of Availability states that organizations must maintain records in a manner that ensures timely, efficient, and accurate retrieval. You can't retrieve what you can't import, and you can't import what fails compliance.
Action Items for Your Team
Integrate compliance verification into your digitization workflow: Don't wait until you're ready to transfer records to discover your image files don't meet technical standards. Create a checklist based on your receiving system's requirements and run it before you package.
Document your file-naming conventions as executable instructions: Your naming scheme should explicitly encode structure: page sequence, document grouping, relationship to parent series. Test whether someone unfamiliar with your collection could reconstruct the intended structure from the file names alone.
Implement authority control at the point of description: If you're cataloging records that will eventually become digital objects, apply controlled vocabulary from the start. Retrofitting authority-compliant terms into existing descriptions creates delays and errors.
Treat metadata as a deliverable, not documentation: The metadata you create during cataloging becomes the Preservation Description Information that pairs with digital image files. If it's incomplete or non-compliant, your digital objects can't transform into machine-readable records. Metadata quality directly determines import success.
Test your transformation logic before full-scale digitization: If you're planning to structure a collection into preservation originals and public access copies, or to group image files into multi-page documents, run a pilot batch through the compliance checkpoint. Discover structural errors on 50 records, not 50,000.
Understand that machine-readable means queryable and portable: The transformation from human-readable to machine-readable isn't academic. It's what allows your records to be queried granularly, compared across collections, and shared via API to external platforms. If your records can't transform, they can't interoperate.
The compliance checkpoint isn't bureaucracy. It's the last point where you can prevent records from entering a system in a state that makes them un-queryable, un-portable, and effectively unusable for anything beyond linear browsing. Every organization managing digital records faces this same gate, whether you call it a compliance check, a validation layer, or a pre-ingest review. Build your controls to pass it on the first attempt.
