Metadata standards for records, without the jargon

Metadata is the information that makes a record findable, understandable, and trustworthy without opening it. A folder of ten thousand files named IMG_0001 through IMG_9999 has almost no metadata, and so it has almost no value, because nobody can locate anything or say what it is. The work of description is what turns a pile of files into a collection someone can use. This guide covers the standards worth knowing and the smallest set of fields that will actually help.

Two different jobs, often confused

People say "metadata" as if it were one thing. In records work it does at least two distinct jobs, and mixing them up causes most of the confusion in the field.

Descriptive metadata tells you what a record is: its title, who created it, when, what it is about, what language it is in. This is what supports discovery. It answers "is this the thing I am looking for."

Preservation metadata tells you what has happened to a record over time: what format it is, what checksum it had at acquisition, who transferred it, what migrations or fixity checks were run on it and when. This is what supports authenticity and long-term care. It answers "can I trust this, and do I know its history."

A record needs both. Descriptive metadata without preservation metadata gives you a findable object you cannot vouch for. Preservation metadata without description gives you a trustworthy object nobody can find.

Dublin Core: the common vocabulary

Dublin Core is the most widely adopted general standard for descriptive metadata, and it is the right starting point for almost anyone. Its strength is that it is small and easy. The core set defines fifteen elements, all optional and all repeatable, that cover the common ground of describing almost any resource: Title, Creator, Subject, Description, Publisher, Contributor, Date, Type, Format, Identifier, Source, Language, Relation, Coverage, and Rights.

Because those fifteen elements are so widely understood, describing your material in Dublin Core terms means other systems can ingest and read it later. You are not inventing a private scheme that only your current software understands. For many small collections, a clean Dublin Core record per item is genuinely enough, and it is far better than the elaborate scheme you never finish filling in.

Dublin Core will not carry the fine-grained detail that a specialized domain needs. Archival description often uses EAD (Encoded Archival Description) to represent the hierarchy of a collection, and library cataloguing has its own deep standards. But as a shared baseline that keeps you interoperable, Dublin Core is hard to beat, and you can always map its fields into a richer scheme later.

PREMIS: documenting what happened

PREMIS (Preservation Metadata: Implementation Strategies) is the standard for the preservation side. Where Dublin Core describes the resource, PREMIS is built to record the things that matter for keeping a digital object authentic and usable over time. It is organized around a few entities: Objects (the files themselves and their technical characteristics), Events (things that happened, such as ingest, a fixity check, a format migration, or a validation), Agents (the people or software that carried out those events), and Rights (what you are permitted to do with the object).

The center of gravity in PREMIS is the Event. Every time you do something to a digital object, that action becomes a recorded event with a date, an outcome, and the agent responsible. Over years, this builds a documented history: this file was ingested on this date with this checksum, verified on these later dates, migrated from one format to another by this tool on this date, and the migration succeeded. That history is exactly what lets you answer, credibly, "how do you know this file is what it claims to be." You do not have to implement the full PREMIS data model to benefit from its thinking. The habit of recording every action on an object as a dated, attributed event is the valuable part, and it scales down to a simple log.

A minimum field set you can adopt this week

Standards can stall you if you treat adoption as all-or-nothing. It is not. Here is a small, defensible set of fields that captures both jobs and maps cleanly onto Dublin Core and PREMIS thinking. Record these for each item or each logical group, and you are already ahead of most collections:

1. Identifier: a unique, stable ID you assign and never reuse.
2. Title: a human-readable name for the record.
3. Creator: who made it.
4. Date: when it was created, in ISO 8601 format (YYYY-MM-DD), which sorts and parses unambiguously.
5. Description: a sentence or two on what it is and why it is kept.
6. Format: the file format, ideally by a precise identifier rather than just an extension.
7. Source of transfer: where it came from and who handed it over.
8. Date of acquisition: when you took custody.
9. Checksum: the SHA-256 hash at acquisition (this ties your description to the fixity routine).
10. Rights: who owns it and what you may do with it.

The first six lean descriptive, and the last four lean toward preservation and provenance. Together they let a stranger find the record, understand it, verify it, and know whether they are allowed to use it. Start there, keep it consistent, and enrich later. Consistency across every record matters more than depth on any single one, because search and sorting only work when the same field means the same thing everywhere.

Practical cautions

A few things save real pain later. Use controlled, consistent values where you can, so that "photograph," "photo," and "Photograph" do not fragment into three categories that no search will unite. Prefer explicit date formats, because a date written as 03/04/05 is genuinely ambiguous across the world and ISO 8601 removes the guesswork. Keep the metadata alongside the objects and included in your fixity manifest, so the description travels with the material and is itself protected from silent loss. And write down your own rules, even a short set, so the next person describes things the way you did rather than starting a competing convention.