Every major persistent identifier (PID) registry used in research administration — DOI, ORCID iD, ROR ID, and RAiD — is operated independently, by a different organization, with its own registration process and metadata schema. Federation is the layer of work, standards, and infrastructure that lets those independently-operated registries be queried and traversed as if they were connected, without merging them into one system. This guide covers what that federation layer actually consists of, what’s currently built, and why it matters operationally. It is deliberately conceptual/infrastructure-level; for the practical task of matching a specific person, organization, or output across identifier schemes, see Identifier crosswalk in the CASRAI Dictionary.
Federation is not one system — it’s cross-referencing between several
No single registry contains everyone and everything. DOIs for research outputs are registered through two main agencies — Crossref (mostly journal articles and books) and DataCite (mostly datasets, software, and other outputs). ORCID identifies people. ROR identifies research organizations. RAiD, standardized as ISO 23527:2022 and operated by Australia’s ARDC in partnership with DataCite, identifies research projects and activities themselves rather than any single output of them.
Federation happens at the metadata level: each registry stores fields that point to identifiers minted by the other registries. A Crossref-registered DOI record can carry the ORCID iDs of its authors and the ROR IDs of their affiliated organizations. An ORCID record can list a Scopus Author ID or a ResearcherID under its external-identifiers block. A ROR record carries known equivalences to GRID, ISNI, Wikidata, and the Crossref Funder ID for the same organization. None of this requires a merged database — it requires each registry to store, and expose via API, correct pointers into the others. CASRAI’s identifier crosswalk entry covers this record-level mapping mechanic in more detail; this guide is about the infrastructure that aggregates those crosswalks into something queryable at scale.
The PID Graph: treating identifiers as nodes in a network
The clearest articulation of PID federation as an explicit goal came out of FREYA, an EU Horizon 2020 project (2017–2020, grant agreement 777523) run jointly by DataCite, Crossref, ORCID, and other partners. FREYA’s central concept, the PID Graph, treats persistent identifiers as nodes in a graph and the metadata relationships between them — creator-of, affiliated-with, funded-by, derived-from — as edges. The point is that a query no single registry could answer alone (e.g., “which datasets, funded by which grants, were produced by researchers at this organization?”) becomes answerable by traversing edges across DOI, ORCID, ROR, and funder-identifier nodes together.
The PID Graph isn’t a hypothetical model; it’s implemented. DataCite Commons, launched in pre-release form in 2020, is DataCite’s public interface onto it, joining DataCite DOIs, a subset of Crossref DOIs, ORCID records, and ROR records into a single browsable and queryable resource. OpenAIRE Graph takes a broader, EU-funded approach — ingesting from Crossref, DataCite, ORCID, and any repository that follows OpenAIRE’s harvesting guidelines, then applying deduplication and link-inference to build a unified scholarly knowledge graph that also distinguishes itself from (and interoperates with) institutional CRIS systems. OpenAlex, run by the nonprofit OurResearch, is a further example of the same underlying pattern applied at large scale: aggregating works, authors, institutions, and funders from multiple upstream PID sources into one queryable index.
Access to the underlying PID Graph data has historically gone through DataCite’s GraphQL API. That API is now itself an example of how federation infrastructure keeps evolving rather than staying fixed: DataCite has confirmed it will deprecate the GraphQL API on 1 July 2027, having already migrated DataCite Commons itself onto the DataCite REST API, which now exposes the same metadata and relationships. Anyone building against the PID Graph today should be pointed at the REST API, not the GraphQL endpoint, going forward.
Federation is also happening at the registry level, not just the query level
A second, less-visible kind of federation is registries actively converging on each other rather than just cross-referencing. The clearest current example: Crossref announced in 2023 a long-term plan to retire its Open Funder Registry (the source of the legacy Crossref Funder ID) and transition funder identification to ROR instead. Crossref members can already submit ROR IDs anywhere they previously submitted Funder IDs, and legacy Funder IDs continue to be mapped automatically to their ROR equivalents. Crossref has been explicit that this is a long-term transition, not an overnight cutover, and legacy Funder IDs remain valid in the interim — but the direction is a single organization-identifier layer (ROR) rather than parallel, semi-overlapping ones. That’s federation working in the direction of consolidation, distinct from the query-federation the PID Graph performs across registries that stay separate.
Where the standards and coordination work is happening now
Federation across registries that are run by different organizations, in different countries, under different funding models, doesn’t happen by accident — it requires ongoing standards coordination. The Research Data Alliance’s Persistent Identifier Interest Group is the longest-running forum for this: an RDA-recognized interest group that coordinates PID-related working groups and use cases across the research, data, and infrastructure communities. More recently, RDA-US and the National Information Standards Organization (NISO) announced in February 2025 that they are jointly developing an ANSI/NISO standard for a national PID strategy in the United States, building on a 2024 Open Research Funders Group report on the same topic. NISO is administering the working group; RDA-US is contributing the PID-specific expertise. That a formal national standards body is now writing PID federation into an ANSI standard is itself evidence of where this space is heading: from informal registry-to-registry cross-referencing toward something closer to a governed, national-level policy layer.
On the CRIS side — where a research office’s own systems have to consume all of this rather than just observe it — the equivalent coordination mechanism is CERIF, the data model that gives CRIS systems a shared way to represent people, projects, organizations, and outputs so that PID-based joins are possible at the institutional level too. See CASRAI’s CRIS interoperability entry for how CERIF, OAI-PMH, and shared identifiers combine to let an institutional CRIS talk to funders, repositories, and aggregators like OpenAIRE.
Why federation matters, not just how it works
The alternative to federation is what research administration lived with for most of PIDs’ history: identifiers that are internally consistent but externally silent. A DOI that reliably resolves to a paper but says nothing machine-readable about who wrote it, what organization they were at, or what grant paid for it. An ORCID record that’s a trustworthy record of one person’s outputs but isn’t automatically linked to the funding or the institutional project that produced them. Each identifier does its own job well in isolation; none of them, alone, supports the kind of question a research office, funder, or publisher actually needs answered — which requires joining across all of them.
Federation is what makes those joins machine-actionable instead of manual. Concretely, for research administration this means:
- Automated compliance and reporting. A funder can query, without manual matching, which outputs (DOIs) resulted from which grants (RAiDs), produced by which researchers (ORCID iDs) at which institutions (ROR IDs) — instead of requiring a research office to compile that mapping by hand for every report.
- Reduced re-entry across systems. A CRIS system that ingests CERIF records carrying ORCID/ROR/DOI identifiers can auto-populate a researcher profile or project record instead of a human re-typing affiliation and output data that already exists correctly somewhere else in the federation.
- Consistency that survives organizational change. Because ROR (not a free-text institution name) is the join key, a researcher moving institutions, an organization renaming, or a department restructuring doesn’t silently break the linkage between past outputs and their correct affiliation history.
- Discovery that spans the whole research lifecycle. Tools built on the PID Graph or OpenAIRE Graph can answer lifecycle-spanning questions — from a grant, to the project it funded, to the people on that project, to everything they produced — that no single siloed registry could answer on its own.
Limits of federation as it exists today
Federation is real infrastructure, not a finished one, and it’s worth being precise about its current limits rather than overstating them:
- Coverage is uneven. DataCite Commons, by its own description, includes all DataCite DOIs and ROR/ORCID data, but only a subset of Crossref DOIs and Crossref metadata — federation depends on each registry’s willingness and capacity to expose data, which is not uniform.
- Not every identifier participates equally. RAiD, the newest of the four core identifiers covered here, has materially less federated tooling built around it than DOI, ORCID, or ROR simply because it’s younger and its consortium of registration agencies is still expanding.
- Infrastructure changes underneath you. The DataCite GraphQL API deprecation above is a concrete example: a system built against the PID Graph’s original access layer needs active maintenance to keep working as the underlying infrastructure moves to REST.
- Federation at the query layer (PID Graph, OpenAIRE Graph) is different from federation at the registry layer (Crossref Funder Registry retiring into ROR) — the first cross-references separate registries, the second consolidates them. Both are commonly described as “PID federation,” which can cause confusion about what’s actually being claimed in a given context.
Frequently asked questions
Is PID federation the same thing as an identifier crosswalk?
No, though they’re closely related. An identifier crosswalk is a specific mapping between two identifier schemes for the same entity (e.g., a ROR ID’s listed equivalence to a GRID ID). Federation is the broader infrastructure — standards work, APIs, graph databases, coordinating bodies — that makes crosswalks like that discoverable and queryable at scale across many registries at once, rather than a one-off lookup table.
Is the PID Graph a single, centrally-maintained database?
No. “PID Graph” describes a data model and a concept popularized by the FREYA project, not one specific piece of software. DataCite Commons, OpenAIRE Graph, and OpenAlex are separate, independently-run implementations of the same underlying idea, each with different coverage and scope.
Do DOI, ORCID, ROR, and RAiD federate with each other automatically once an identifier is registered?
Not automatically in the sense of zero effort. Federation depends on whoever registers a record actually populating the cross-reference fields (an author’s ORCID iD on a DOI deposit, a ROR ID on an ORCID affiliation entry, and so on). The infrastructure to query across registries exists, but the underlying data quality still depends on correct metadata at the point of registration — which is exactly the practical, record-level work covered in CASRAI’s identifier crosswalk entry.
Who is responsible for PID federation standards going forward?
No single body owns it. The Research Data Alliance’s PID Interest Group is the longest-running cross-community coordination forum; RDA-US and NISO are jointly developing an ANSI/NISO US national PID strategy (announced February 2025); and each individual registry operator (Crossref, DataCite, ORCID, ARDC for RAiD) continues to make its own implementation decisions about how much of its data to expose and how.
How does this connect to CRIS systems specifically?
A CRIS is typically where federation gets consumed rather than produced: an institutional CRIS built on CERIF can ingest ORCID, ROR, and DOI-linked data through harvesting protocols like OAI-PMH and the OpenAIRE Guidelines, populating researcher and project records without manual entry. See CRIS interoperability for the institutional side of this.
Related CASRAI resources
- Persistent Identifiers & Research Information Systems — the cluster hub for this topic
- Identifier crosswalk — the practical, record-level mapping mechanic behind federation
- PID graph
- CRIS interoperability
- CERIF
- OpenAIRE Graph
- How to Find, Look Up, and Resolve a DOI
- How to Register, Find, and Search for ORCID iDs
