A research support librarian sent me a screenshot last week that summed up the problem better than any policy memo could. Her university had just received a funder query about the outputs from a three-year grant. The journal article was easy to find because it carried a Digital Object Identifier, or DOI. The preprint, two datasets, and the software release were harder to trace. They existed across different repositories. They used different identifiers. Nothing in the public record said they belonged to the same award.
Crossref and DataCite, the two main DOI registration agencies for scholarly content, have now published metadata guidance aimed at that exact failure. The recommendations describe how publishers, data repositories, preprint servers, and research institutions should record the relationships among grants, preprints, and research data. The approach is less about creating new standards than about using existing persistent identifiers in a more disciplined way.
Metadata is machine-readable information about an object: its title, creators, publication date, funding source, and links to related items. Without consistent metadata, a funder or institution cannot assemble an accurate picture of what a grant produced. The Crossref and DataCite guidance argues that the pieces already exist; what has been missing is the habit of depositing them together.
The Missing Links Between Grants, Preprints, and Data
Funding agencies increasingly require evidence of public access and research outputs. A preprint may be the first public version of an article, yet many preprint records do not point to the eventual journal version, or vice versa. A dataset may be cited in the text of a paper but never linked by a formal related identifier. A grant number may appear in a PDF acknowledgment but not in the metadata submitted to Crossref or DataCite.
The consequences are practical rather than theoretical. Literature reviews miss datasets because citation indexes do not know the dataset is connected to a paper. Funder progress reports require manual assembly of outputs. Institutions trying to track research data compliance end up running separate searches across disconnected systems. The guidance treats metadata completeness as the cheapest available fix.
Crossref registers metadata for journal articles, conference proceedings, books, and posted content including preprints. DataCite registers metadata for datasets, software, and other research outputs. Both have long supported funding fields and related identifiers. The new guidance frames those fields as a connected workflow, not an optional afterthought.
What the Guidance Asks Depositors to Record
The core recommendation is to record four things every time a research object is deposited: a persistent identifier for the object itself, a persistent identifier for each creator, a funder identifier with the award number, and relationship metadata linking the object to its related preprints, datasets, articles, or grants.
For people, the identifier is an ORCID iD. For research organizations and funders, ROR IDs have become the default. The scholarly metadata community now encourages ROR identifiers for research organizations and funders; the Research Organization Registry maintains the open identifier list. For the objects themselves, the DOI remains the anchor.
Crossref's schema has included funding metadata for years. Publishers can deposit funder names, funder identifiers, and award numbers along with an article. DataCite's schema has a parallel funding reference block for datasets and other outputs. The guidance asks depositors to fill these elements even for preprints and early-stage records, because retrofitting them after peer review is harder and less reliable.
Relationship metadata does the linking work. A preprint record can carry a preprint-to-article relation, and a data repository can record that a dataset is supplemented by an article or references a grant. Crossref and DataCite already define these relation types in their metadata schemas. The guidance pushes for wider and more consistent use. The Crossref schema documentation and the DataCite Metadata Schema describe the fields in detail.
Preprints, Data, and the Missing Connections
Preprint servers have grown quickly, particularly in the life sciences and social sciences. Crossref treats preprints as posted content and has supported preprint-to-article relationships for years. DataCite allows a preprint resource type and relationship metadata. The gap has not been the technical ability; it has been uptake.
Many preprint records still arrive without a matching journal article relationship. That leaves a reader who finds the preprint through Google Scholar or Crossref unable to see whether a later version was published. The guidance encourages preprint servers, publishers, and authors to deposit relationship updates when a preprint becomes a journal article. Crossref's earlier work on retraction metadata across preprint servers, covered by Academic Jobs, shows how relationship metadata can connect scattered records after deposit.
Data linkage raises similar issues. A dataset may be cited in a reference list, but a human-readable citation does not always create a machine-readable link. Journal policies on data citation are shifting. Recent coverage on data citation mandates addressed how publishers are tightening requirements. The new Crossref and DataCite guidance complements those mandates by specifying the metadata fields that make compliance visible to machines.
What Libraries and Research Offices Can Do
Research administrators sit close to the workflow. When a grant is awarded, the award number and funder identifier should be recorded in a machine-readable form, not only in a PDF. When a preprint is deposited, the depositor can include the grant link. When a dataset is published, the repository metadata can include the related article DOI once it exists.
Several practical steps follow from the guidance:
- Record funder names and identifiers in the same metadata block as the award number.
- Add ORCID iDs for all creators at deposit, including preprint co-authors.
- Link preprints to their published articles with relationship metadata once the article DOI exists.
- Update data repository records when a related article is published and when a funder report is due.
The harder part is timing. A preprint is often deposited before the journal article exists. The guidance suggests treating metadata as a living record: update the preprint record when the article DOI is assigned, add the dataset link when the data are published, and attach the grant relationship at the first opportunity rather than waiting until the final version.
What the Guidance Does Not Fix
Metadata guidance is not a compliance mechanism. Crossref and DataCite do not require publishers or repositories to deposit every relationship; they publish recommendations and make the fields available. The result is uneven coverage. A large publisher may deposit funding metadata thoroughly for journal articles while a small society journal does not. A preprint server may expose relationship fields but leave them optional. A data repository may record a related article only when the author provides the DOI at deposit time.
This unevenness means researchers and funders still need to verify records. A missing link is not always evidence that no connection exists; it may mean the depositor skipped an optional field. The guidance is most useful when institutions treat the fields as defaults rather than extras.
One Small Step for This Month
Most researchers will not deposit metadata themselves; a publisher or repository does that. But the person who controls the broadest slice of the record is the researcher. Check your ORCID record this month. Add the grant number if it is missing, link any dataset you have deposited, and confirm that your preprint record points to the published version if one exists.
That single pass through an ORCID profile will reveal how often the links are missing. It takes less than an hour and gives a research office a cleaner machine-readable trail for the next funder query. The metadata guidance from Crossref and DataCite will mean little if depositors do not fill the fields. The fields are now the easier part; the workflow is the next hurdle.
Photo by Ato Aikins on Unsplash