The Details That Travel
See which useful details are attached to research records around the world.
Try the project ↓How to use this project
Help
How to explore
A research record can carry a summary, references, reuse terms, and IDs that connect people and organizations. Crossref counted which of those details its members supplied with registered work.
Choose a place
Start with the United States or choose another location where Crossref members registered at least 100,000 records.
Choose a detail
Select references, an abstract, license information, an author ID, an organization ID, or funding details on the record card.
Follow the same detail
Read its timeline, compare member locations, and see how the pattern changes across work types.
Check the limits
Finish with the location and work-type notes before treating a missing field as a judgment about research quality.
- DOI record
- A registered description of a work tied to a Digital Object Identifier, which provides a lasting link to that work.
- Metadata coverage
- The share of registered records that include a particular field, such as an abstract or license link.
- Member location
- The country of the Crossref member that registered a record, not necessarily the author's country or the place studied.
- ORCID iD
- A persistent identifier that distinguishes one researcher from other people with similar names.
- ROR ID
- A persistent identifier for a research organization, such as a university or institute.
Research records · December 2018 to April 2026The details
The details
that travel
See which useful details are attached to research records around the world.
A research record can carry an abstract, references, reuse terms, and IDs that connect people and organizations. Crossref counted what its member organizations supplied with 181.7 million registered records.
Choose a member location
Start with the United States, or choose another place. Then select one detail on the record card. That choice follows you through every view.
Metadata coverageApril 30, 2026
How to read this: 43.6% of records registered by members in United States include references. Select a different row to change the timeline, country comparison, and work-type view. These are separate fields, not parts of one score.
Follow references through the record
The solid line follows United States. The dotted line combines every member location in the source. Each point is the share of all records counted by that date.
United States
United States
All 170 member locations
United StatesWorldwide
How to read this: higher means a larger share of the cumulative record includes references. The dates are uneven before 2025, then monthly. A change can come from new records or updates to old ones.
Compare references across member locations
The list includes the 54 locations with at least 100,000 registered records in April 2026. Filter by region, then choose any row to update the full experience.
All regions20.7%
middle country share for references
Each member location counts once here. The worldwide line above instead gives each DOI record equal weight.How to read this: bars show the share of records with references, not the amount or quality of research. The location belongs to the Crossref member that registered the record; it may differ from the authors' countries or the place studied.
The kind of record changes the pattern
Compare references among eight common Crossref work types registered by members in United States. The share and record count sit together so a tiny group cannot look as important as a large one.
35.8 million records
relative record count
7.9 million records
relative record count
4 million records
relative record count
1.5 million records
relative record count
1.4 million records
relative record count
737.4 thousand records
relative record count
526.3 thousand records
relative record count
184.7 thousand records
relative record count
How to read this: the colored bar is the share with references. The thin line below shows each work type's record count relative to the largest type in United States. Some fields are useful for one kind of work and irrelevant to another.
A missing field is a gap, not a verdict
These counts describe metadata supplied to Crossref. They do not rate the research, publisher, author, or country.
A record is assigned to the country of the Crossref member that registered it. Large international publishers can make one location look much larger than the research produced there.
A dataset may not cite other work. A paper may have no outside funding. Crossref also introduced some identifiers later than others.
Each snapshot includes records submitted or updated by that date. The lines mix new deposits with repairs to older metadata.
Source, reduction, and checks
Crossref published this country-level snapshot on Zenodo under CC BY 4.0. It contains 53,820 rows for 170 member locations from December 2018 through April 2026.
The build sums each work type into country, region, and worldwide totals. Direct comparison keeps locations with at least 100,000 records in the latest snapshot. The worldwide line keeps every location.
The saved source was 7.1 million bytes and matched Zenodo's MD5 plus SHA-256 0b6363a8bd58…. The reduced page data is checked in and the browser makes no source request.
Why this exists
The question
How do the shares of Crossref DOI records with references, abstracts, license information, ORCID iDs, ROR IDs, and funding metadata differ over time, by registering-member location, and by work type?
What to try
Choose a Crossref member location and one metadata field to update its record card, cumulative timeline, country comparison, and work-type breakdown.
Method and limits
About the data
I found the State of Crossref metadata by member country dataset through TidyTuesday's May 19, 2026 catalog entry, then opened Alexandre Bédard-Vallée's original Zenodo record, its CC BY 4.0 license metadata, and Crossref's official explanations of metadata and Participation Reports. I retrieved the immutable Zenodo coverage file on October 3, 2026. The 7,072,519-byte CSV has SHA-256 0b6363a8bd58e6311ffcdd850daeea600e7e8d8d0b7cf8b76833c7c35f76af38 and matches Zenodo's published MD5 f7759d614ee3bfccb91382a4ec6d83bf. It contains 53,820 country-date-work-type rows across 170 member locations and 38 snapshots from December 31, 2018 through April 30, 2026.
The reproducible Python reducer verifies the source byte count, both hashes, required columns, row count, date range, country count, region codes, nonnegative values, and that no selected metadata count exceeds its DOI-record denominator. One Rwanda conference-paper row dated January 31, 2026 leaves the acknowledges_funding count blank; the reducer validates that exact cell as the only blank among the selected measures and treats it as zero when aggregating.
For each snapshot, the reducer sums document-type rows into member-location, region, and worldwide totals. Coverage divides records with at least one selected field by all DOI records in the same group. The six fields are references, abstracts, license information, any ORCID iD, any ROR ID, and any funding metadata. Work-type views merge posted-content subtypes and retain eight named types. Country comparison includes the 54 member locations with at least 100,000 registered records in the April 2026 snapshot; worldwide values retain all 170 locations. The middle-country value gives each displayed location equal weight, while the worldwide value gives each DOI record equal weight. Repeated reduction produced the same 131,849-byte browser snapshot at SHA-256 84c13b4e24b4e9ea6660c1528307cca8090b207ca27b72dde3425a82da9610b1. The page imports that file and makes no runtime network request.
The snapshots are cumulative: each one covers records submitted or updated through its date. A change can reflect new registrations, corrections to older records, shifts in work types, or changes in which members deposited records. The dates are irregular before January 2025 and monthly afterward. The source assigns a country from the registering member, so it does not locate authors, readers, institutions, research activity, or the subject of a work. Large international publishers can concentrate records in one member location. Metadata fields also differ in relevance by work type and changed over time. Missing metadata does not show that a work, author, member, or country is low quality, and these descriptive comparisons cannot explain why coverage differs.
Skills used
- Data analysis
- Data engineering
- Statistical reasoning
- Systems design
- Visualization
- Information design
- Interaction design
- Visual design
- Creative coding
- Product thinking
- Editorial storytelling
- Accessibility
- Performance
What the data shows
- The April 2026 snapshot contains 181,735,699 DOI records across 170 Crossref member locations. Worldwide, 43.7% include references, 44.7% include license information, 22.9% include an abstract, 11.2% include an ORCID iD, 8.0% include funding details, and 0.5% include a ROR ID.
- Among 60,546,701 records registered by members in the United States, reference coverage rises from 31.7% in the December 2018 snapshot to 43.6% in April 2026. Because the snapshots are cumulative, the change mixes new records with updates to older ones.
- The April 2026 abstract share is 53.2% for records registered by members in Brazil and 0.6% for records registered by members in Japan. This difference describes deposited fields, not research quality or where the research happened.
- Work type changes the pattern inside one member location. In the United States, 86.6% of posted-content records include an abstract, compared with 37.2% of journal articles and 7.5% of book chapters.