SYSTEM Cited by 1 source
Netflix PDM (Primary Data Management)¶
PDM — Primary Data Management — is Netflix's platform for managing authoritative reference data and taxonomies, and is the first named production consumer of UDA (Source: sources/2025-06-14-netflix-model-once-represent-everywhere-uda).
What PDM does¶
From the Netflix post: "PDM is our platform for managing authoritative reference data and taxonomies. PDM turns domain models into flat or hierarchical taxonomies that drive a generated UI for business users. These taxonomy models are projected into Avro and GraphQL schemas, automatically provisioning data products in the Warehouse and GraphQL APIs in the Enterprise Gateway."
Concretely:
- A Netflix team authors a domain model in Upper (systems/netflix-upper) — e.g. a taxonomy of genres, a catalog of territories, a classification of asset types.
- PDM turns that model into a flat or hierarchical taxonomy with a generated UI for business users to manage entries.
- The same model is projected — via UDA's transpiler family (patterns/schema-transpilation-from-domain-model) — into:
- Avro schemas, auto-provisioning data products in the Netflix Warehouse.
- GraphQL schemas, auto-provisioning APIs on the Enterprise GraphQL Gateway.
One authored model → one UI + one warehouse data product + one Enterprise-Gateway API. This is the model-once-represent-everywhere thesis applied to reference data.
Why PDM fits the UDA thesis¶
Reference data + taxonomies are the archetype of "same concept, modelled differently everywhere" — every team historically rebuilt their own genre list, territory list, rating-system table, each in whatever schema best fit their stack. PDM uses UDA's model-as-data substrate to collapse all of that into:
- One source of truth for the taxonomy entries (as data in the knowledge graph).
- One source of truth for the taxonomy shape (as a domain model in Upper).
- Many generated surfaces — authoring UI + Avro + GraphQL + anything else UDA can transpile to later.
Canonical master-data-management instance¶
PDM is the wiki's canonical instance of UDA-driven master data management — taxonomies and reference data as first-class UDA domain models with auto-provisioned storage and API surfaces.
Caveats¶
- Internal-to-Netflix. No fleet / QPS / taxonomy-count / entry-count numbers disclosed.
- Authoring UI details undisclosed. The generation of the UI from the domain model is mentioned, not specified.
- Avro / GraphQL provisioning pipeline opaque. End-to-end pipeline mentioned, not walked through.
- Relationship to prior Netflix master-data-management systems (if any) not discussed.
Seen in¶
- sources/2025-06-14-netflix-model-once-represent-everywhere-uda — named as UDA's first production consumer.
Related¶
- systems/netflix-uda — PDM is an in-production UDA use case.
- systems/netflix-upper — authoring language.
- systems/netflix-enterprise-graphql-gateway — where PDM's generated GraphQL APIs land.
- concepts/domain-model — PDM's input.
- patterns/model-once-represent-everywhere · patterns/schema-transpilation-from-domain-model — the two patterns PDM canonicalises for reference data.