Coming in June: a simpler, more consistent ID format

[danabauer]

GERS IDs will become UUIDs in June

In next month's release, Overture is standardizing the id property across all themes by adopting UUIDs, stored as strings with dashes: xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx

Currently the structure of our IDs varies slightly across datasets, with some themes embedding metadata (e.g. H3) in the id property. The new UUIDs will be randomly-generated identifiers with no encoded information. Moving to UUIDs enables consistent ID generation and registration across our data pipelines, which is necessary for the successful implementation of GERS.

This is a one-time breaking change for our ID system. We plan to support our users by providing a file that maps the old IDs to the new UUIDs. Please reach out with your comments and questions.

Tip

Check out the 2025-05-21.0 release notes to see what's new in May.

[ghost]

I can see there being advantages to embedding metadata in the id property, as you had done before with H3 ids. Could you share some more insight into why you made the switch to UUID?

[danabauer]

Thanks for your question, @rami-dv. We're making this switch for several reasons:

1. As we centralize data pipelines from member companies to Overture infrastructure, a unified ID format simplifies ID generation and registration across themes.
2. Deterministic IDs -- with embedded metadata based on H3 cell or theme or any other property -- require careful design to prevent collisions across the entire Overture corpus, which includes all IDs for all time. Randomly-generated UUIDs help us avoid this issue and allow for ID stability even if an entity moves to an adjacent H3 cell or is transferred from one Overture theme to another.
3. Not all of the data Overture uses in the conflation process for each theme is open data; some data sources from our contributors are proprietary. Deterministic IDs based on data properties could inadvertently expose underlying data information, creating privacy and licensing issues.