Data Catalogue Metadata Standard Template (DCAT 3)

Data Catalogue Metadata Standard: [Organisation name]

Standard for what every dataset, data service and data product must say about itself in the data catalogue: the minimum metadata set and who fills each field, the mapping to the W3C Data Catalog Vocabulary (DCAT) Version 3 (W3C Recommendation, 22 August 2024), catalogue roles, certification badges with entry and exit criteria, completeness scoring, harvesting and review. Replace every [bracketed] value and every row marked Example. Worked examples use Corvid Commerce, a fictional SaaS retailer with a lakehouse, dbt transformations and domain-owned data products.

Document control

Approvals

Revision history

Contents

• 1. Purpose and scope
• 2. Principles
• 3. Asset types and identifiers
• 4. Minimum metadata set
• 5. Extended metadata by asset type
• 6. Catalogue roles and responsibilities
• 7. Certification badges
• 8. Completeness and quality scoring
• 9. Harvesting, publication and exchange
• 10. Lifecycle, review and deprecation
• 11. Compliance, exceptions and metrics
• 12. Glossary and references
• Standard and licence note

1. Purpose and scope

1.1 Purpose

[Two to four sentences on the outcome: consumers find the right data, understand it and know whether they can trust and use it.]

Example: Analysts at Corvid Commerce spend part of every request finding out which of three order tables is current and who can approve access. This standard makes every shared dataset findable by business name, states who owns it, what it contains, how fresh it is, whether it holds personal data and whether it is certified, so that a consumer can decide to use it without asking around.

1.2 Scope

An asset is shared when anyone outside the owning team reads it. Every shared asset must meet the minimum set in section 4. Private assets may be listed but are hidden from default search.

2. Principles

3. Asset types and identifiers

Rules:

1. Identifiers are assigned once and never reused, even after the asset is retired.
2. The physical location (database, schema, table, bucket path) is a property of the distribution, not part of the identifier.
3. A data product record links to its datasets and services with dcat:qualifiedRelation or the tool equivalent, so that consumers see the parts behind the product.

4. Minimum metadata set

Fields marked M are mandatory for every shared asset. R means recommended; it counts towards the completeness score in section 8.

Example: urn:corvid:dataset:sales:orders is a Silver dataset that holds personal data, uses no external data and is not itself a data product. For it, 16 fields are mandatory (META-001 to META-007, META-010 to META-012, META-014, META-015, META-019, META-021 to META-023) and 8 are recommended (META-008, META-009, META-013, META-016 to META-018, META-020, META-024). All 16 mandatory fields are filled; 6 of the 8 recommended fields are filled (licence and temporal coverage are empty), which gives the scores in section 8.

5. Extended metadata by asset type

6. Catalogue roles and responsibilities

6.1 RACI for catalogue activities

R: responsible, A: accountable, C: consulted, I: informed.

7. Certification badges

Rules:

1. Only the certification workflow sets a badge; owners cannot set it by editing the record.
2. A downgrade is automatic when the exit trigger is detected and notifies the owner, the steward and the consumers who follow the asset.
3. A badge is shown with its date and the evidence link (check results, sign-off ticket).

8. Completeness and quality scoring

Completeness of one record:

• Mandatory completeness = filled mandatory fields divided by mandatory fields that apply to the asset type.
• Recommended completeness = filled recommended fields divided by recommended fields that apply.
• Record score = 0.7 times mandatory completeness plus 0.3 times recommended completeness, shown as a percentage.

Example: For the Orders record in section 4 the mandatory completeness is 16 of 16 (100%) and the recommended completeness is 6 of 8 (75%), so the record score is 0.7 x 100 + 0.3 x 75 = 92.5%.

A field counts as filled only when it passes the validation rule in section 4. Placeholder text such as "n/a" or a copy of the title fails validation.

9. Harvesting, publication and exchange

Rule of single writer: each field has exactly one writing source. When a person edits a harvested field, the edit is stored as an override with a reason and an expiry date.

9.1 Exchange format

Records are exported in DCAT 3 (RDF, serialised as JSON-LD or Turtle) so that a data portal or a partner catalogue can harvest them. Where the organisation publishes open data in the EU, follow the DCAT application profile used by the target portal (DCAT-AP 3.0.1 at the time of writing; check the current release at SEMIC).

{
  "@context": {
    "dcat": "http://www.w3.org/ns/dcat#",
    "dcterms": "http://purl.org/dc/terms/"
  },
  "@id": "urn:corvid:dataset:sales:orders",
  "@type": "dcat:Dataset",
  "dcterms:title": "Orders",
  "dcterms:description": "One row per customer order placed on any Corvid Commerce storefront, with status, totals and channel. Example record for this template.",
  "dcat:keyword": ["orders", "sales", "checkout"],
  "dcterms:accrualPeriodicity": {"@id": "http://purl.org/cld/freq/daily"},
  "dcat:contactPoint": {"@id": "urn:corvid:team:sales-data"},
  "dcat:distribution": {
    "@type": "dcat:Distribution",
    "dcat:accessURL": {"@id": "https://catalog.example.com/datasets/sales-orders"},
    "dcterms:format": "Apache Iceberg table"
  }
}

The example uses a reserved example.com host. For file distributions, state dcat:mediaType with a media type registered at IANA, and check the registration before relying on it in exports.

10. Lifecycle, review and deprecation

11. Compliance, exceptions and metrics

11.1 Metrics

11.2 Exceptions

Exceptions expire after at most 6 months and are listed in the council pack.

12. Glossary and references

12.1 Glossary

12.2 References

Standard and licence note

Based on the W3C Data Catalog Vocabulary (DCAT) Version 3, a W3C Recommendation of 22 August 2024, which is published under the W3C Document License; class and property names are used as identifiers and no specification text is reproduced. DCAT-AP is named only. The metadata set, badges, scoring formula, roles and examples are original, and Corvid Commerce is fictional. This template is original text written for this collection. Check the current official versions of DCAT and of any application profile before use. Template licence: MIT.