Why many dataset pages underperform

How SwiftSchema helps

How it works

What is Dataset structured data?

Dataset schema describes datasets so they’re discoverable via Google Dataset Search, knowledge panels, and research tools. It captures what the dataset contains, who created it, how it’s licensed, how to access it (distributions), and contextual metadata (temporal/spatial coverage, variables, formats). Quality metadata increases trust and reuse, assisting researchers, journalists, and public APIs.

Essential properties

1. name
   — human-readable title.
2. description
   — summary covering scope, methodology, and coverage; avoid marketing fluff.
3. creator
   /
   publisher
   — Organization or Person responsible for the dataset.
4. license
   — URL to license terms (CC BY, proprietary, etc.).
5. datePublished
   /
   dateModified
   — publication and update timestamps.
6. keywords
   — tags for search (topic, domain, dataset type).
7. distribution
   — array of DataDownload objects with
   contentUrl
   ,
   encodingFormat
   ,
   name
   ,
   contentSize
   where applicable.
8. temporalCoverage
   /
   spatialCoverage
   — coverage period and geography.
9. variableMeasured
   — list of variables/columns.
10. identifier
    /
    sameAs
    — DOIs, catalog IDs, or canonical references.

Content prep checklist

• Write a concise overview addressing what, who, where, when, how (methodology), and why the data exists.
• Gather metadata: creator, publisher, license, release date, update cadence, contact info.
• Compile download URLs (CSV, JSON, Parquet, API endpoints) and note their content type, size, and access method (open, requires login).
• Document key variables/columns, units, and data dictionary references.
• Capture temporal span (e.g.,
  2018-01/2024-12
  ) and geographic coverage (city, country, global).
• Include contact or access instructions for restricted datasets.

Implementation workflow

1. Update the dataset page with the metadata above and ensure downloads are accessible.
2. Generate Dataset JSON‑LD with
   name
   ,
   description
   ,
   creator
   ,
   license
   ,
   datePublished
   ,
   keywords
   .
3. Add distributions for each file type; include
   contentUrl
   ,
   encodingFormat
   , optional
   name
   ,
   contentSize
   ,
   requiresSubscription
   .
4. Include coverage metadata (
   temporalCoverage
   ,
   spatialCoverage
   ,
   variableMeasured
   ) when available.
5. Embed the JSON‑LD on the dataset landing page; avoid duplicating across multiple pages unless they represent distinct datasets.
6. Validate using Google’s Dataset structured data tester and Rich Results Test.

Linking datasets to other resources

• Use
  isPartOf
  to link dataset collections (e.g., “Traffic Data Portal”) or parent publications.
• Reference supporting documentation (methodology PDF, data dictionary) via
  citation
  or
  distribution
  .
• Link to related APIs using
  distribution
  entries with
  contentUrl
  pointing to API endpoints.
• Provide
  sameAs
  links to GitHub repos, DOIs, or government catalog entries so search engines can cross-reference.

Versioning and maintenance

• Include
  version
  or mention versions in the description; update
  dateModified
  whenever new data is added.
• Maintain change logs noting schema updates or variable changes.
• If you retire a dataset, mark it as deprecated in the content and remove distribution links or update them to note archival status.
• Monitor download endpoints for uptime; broken contentUrl links hurt trust.

Troubleshooting checklist

• Missing distributions: ensure each DataDownload entry has
  contentUrl
  and
  encodingFormat
  .
• Vague descriptions: provide detailed summaries; avoid boilerplate marketing copy.
• Absent license: link to a public license or usage terms.
• No coverage metadata: include temporal and spatial coverage when the dataset is time- or location-specific.
• Unlinked identifiers: add DOIs or catalog IDs in
  identifier
  or
  sameAs
  .

Common Errors & Fixes

• Missing distribution details: include
  contentUrl
  ,
  encodingFormat
  , and descriptive names for each file.
• Vague description: summarize content, collection method, and coverage clearly.
• Absent license: link to a public license or usage terms.
• No creator/publisher: specify the owning organization or person for credibility.
• Outdated metadata: update
  dateModified
  and distributions when data changes.