Create dataset record versions

In this guide

Overview
Create a new dataset version
Version constraints and relationships
Best practices for versioning

Overview

Create new versions of dataset records to maintain a historical record of how published datasets evolve over time. Versioning creates a new draft record that is linked to the original through a version chain, allowing you to track the lineage and evolution of your data while the old version is automatically deprecated.

Create a new version when:

• You add new data to the dataset: For example, you add temperature measurements to a temperature monitoring dataset on a daily, weekly, or monthly basis. Each new version contains all previous data plus the new measurements.

• You add new variables to existing data: For example, some lab results take longer to process than others. Version 1 of your dataset contains the first 10 results for each sample; Version 2 adds an additional 5 results. All previous data remains available in the new version.

• You correct or adjust data in the old version: For example, you correct data to compensate for a calibration error or convert variables to different units (feet to metres, pounds to kilograms). The overall shape and statistics should remain similar—same number of variables and records. If the data looks materially different (for example, you removed a substantial number of records or variables because they were incorrect and unfixable), it can still constitute a new version, but clearly document the changes in the dataset record.

How versioning works:

• New versions can only be created from published dataset records that don't already have newer versions.
• The new version maintains the same identifier as the previous version to ensure continuity.
• The new version starts as a draft with a link to its previous version.
• The original dataset is automatically deprecated when you publish the new version.
• The new version inherits all metadata from the previous version, including authentic source labels

Other ways to connect dataset records

Versioning allows you to track the evolution of datasets over time. If you need to group independent datasets into an ordered collection, create a series. If you want to express specific relationship types between datasets (such as "depends on," "is part of"), use dataset relations.

Create a new dataset version

To create a new version of the dataset record, you must be a Catalogue Manager or have the permission to publish datasets. Check your permissions.

1. Go to the Published tab in the datasets panel.

2. Select the options icon ( ⋮ ) on the dataset record you want to version.

3. Select Create a Version from the menu. The system creates a new draft version and opens the editor.

Restrictions

You can only create a new version from a published dataset record that does not already have a newer version. If the dataset record already has a newer version, the Create a Version button is disabled.

4. Update the new version with your changes:

   • The title and other metadata are copied from the previous version
   • All distributions and data dictionary entries are included
   • Dataset relations and qualified attributions are preserved
   • Authentic source assignments remain intact
   • The status is set to Draft

5. Make your updates, then select Preview & Save to review your changes.

6. Select Save as Draft to save the new version. From here, you can continue editing, submit it for approval, and publish it. See: Dataset record lifecycle.

   To view the previous version of the dataset record, select the previous dataset record link in the dataset card.

What's next?

After publishing the new version, you can view the version history to see how the dataset has evolved over time.

Version constraints and relationships

Versioning follows specific rules to maintain clarity and consistency in dataset lineage:

• Same identifier: New versions retain the same identifier as the previous version to ensure continuity
• One active version: Only one version in a chain can be published without being deprecated
• Linear chain: Each dataset can have only one previous version and one next version
• No version of versions: You cannot create a version of a dataset record that is already a newer version of another published dataset (wait until you publish the new version first)

What's included in the new version:

• All metadata fields: Title, description, temporal coverage, rights statement, etc.
• Distributions: Access URLs, formats, and all distribution metadata
• Data dictionary: Entries with authentic source assignments preserved
• Dataset relations: Links to other datasets or external resources
• Qualified attributions: Role assignments for organisations

What's not included in the new version:

• Activity logs (the new version starts with a fresh activity log)
• Comments (comments from the previous version are not copied)
• Status (the new version always starts as Draft)
• Approval and publication dates (these are set when the new version goes through the workflow)

Best practices for versioning

• Use clear version identifiers: Update the Version field to make versions easily distinguishable (for example, 1.0, 1.1, 2.0). For guidance on semantic versioning for data, see the ANDS data versioning guide^↗.
• Avoid year numbers in version identifiers: Do not use year numbers as version identifiers to prevent confusion with annual snapshots, which should be separate dataset records linked via a dataset series.
• Add version notes: Use the Version_notes field to document what changed in this version.
• Update temporal coverage: Adjust time periods to reflect the data covered by the new version.
• Review all metadata: Ensure all metadata fields are current and accurate for the new version.
• Verify data dictionaries: Check that field definitions remain relevant or update them as needed.
• Test distributions: Ensure all distribution links and access methods work correctly.
• Add comments during workflow: Use comments when submitting, approving, and publishing to document the versioning rationale.