Operationalizing CRediT across a 40,000-author institutional repository

Profile

Composite institution profile

The challenge

What problem were they trying to solve?

The institution had a long-standing institutional repository with strong open-access compliance, but contributorship metadata was effectively lost on deposit. Publications arrived with CRediT statements embedded in the publisher PDF — sometimes structured in JATS, sometimes only as a narrative paragraph at the back of the manuscript — and there was no machine-readable bridge from the published article back into Symplectic Elements, the institutional CRIS. That gap created three downstream problems: NSPM-33 disclosure reviewers could not see who actually did the funded work on a given output, the tenure-and-promotion office could not surface contribution patterns when constructing dossiers, and the library could not give faculty an accurate "what have you actually been doing" report when narrative-CV-style applications arrived. The institution decided the fix had to be operational, not just policy: contributor-role metadata had to land in the CRIS the same week the article was published, not the next annual review cycle.

The approach

How they implemented it

The project ran as a partnership between the library, the office of research, and central IT. The first six weeks were spent on what the team called "the listening tour" — interviewing PIs in five colleges to understand how they actually filled in author and contributor metadata at submission time, and what they expected to get back. The headline finding was simple: senior PIs delegated CRediT assignment to first authors or lab managers, but no one was confirming the assignment after publication, so errors propagated. The technical work, led by the library Symplectic administrators, focused on three integration points. First, the Crossref XML harvester was extended to read the JATS <role vocab="credit"> elements that Cell Press, eLife, and the PLOS portfolio emit; those publishers cover roughly 28% of the institutional output, so they were the fastest path to coverage. Second, for publications arriving without structured CRediT, the deposit form added an optional CRediT capture step modelled on the /credit/for-authors guidance, with the 14 roles displayed as the canonical NISO Z39.104-2022 labels and definitions. Third, the VIVO researcher profile layer pulled the role assignments through and surfaced them on each researcher's public-facing page, grouped by the four CRediT functional categories (conceptual, technical, writing, project management). The team explicitly chose to follow the CASRAI Dictionary domain mapping rather than invent a local schema, so that downstream federation with ORCID and Crossref would not require translation.

Timeline

Rollout phases

1. Months 1–2

   Listening tour + scoping

   Twenty-two PI interviews across five colleges, two library focus groups, and a research-office workshop. Produced a written requirements document signed off by the AVP for Research. Library committed staffing.

2. Months 3–4

   Symplectic harvest extension

   Library Symplectic administrators extended the Crossref XML harvester to parse JATS <role> elements. Tested against a 500-publication sample from the previous 18 months; coverage matched Crossref deposit rates from native CRediT publishers.

3. Months 5–6

   Deposit-form capture for un-tagged outputs

   Optional CRediT capture step added to the institutional repository deposit form. Help text linked to the CASRAI /credit/roles index and the NIH funder-credit guidance.

4. Months 7–9

   VIVO researcher profile surfacing

   VIVO researcher pages updated to display contribution roles grouped by the four NISO functional categories. Pilot launched with three faculty volunteers in biomedical engineering.

5. Months 10–12

   Institution-wide rollout + training

   Library ran twelve CRediT literacy workshops, train-the-trainer kit derived from the CASRAI /for-institutions/training materials. Help desk staffed for two semesters. Tenure-and-promotion office adopted CRediT-aware dossier templates.

Outcomes

Illustrative outcomes

Every metric below is illustrative — synthesised from observed patterns across multiple adoption journeys, not attributed to a single real institution.

Lessons learned

What they would tell the next institution

• 01The "listening tour" was the single highest-leverage investment. Building the harvester first and then asking PIs to use it would have produced a different — and worse — system. Several requirements only surfaced when senior PIs were asked to describe their actual submission workflows.
• 02Publisher coverage skews the build order. Starting with the publishers that already emit structured CRediT (Cell Press, eLife, PLOS, AAAS) bought roughly 28% coverage in the first sprint, which made the un-tagged capture step feel like a complement rather than the whole system.
• 03Authors will not re-enter CRediT data they already entered at the publisher. The deposit form had to detect existing CRediT and skip the capture step on those publications; otherwise help-desk tickets spiked.
• 04Tenure-and-promotion adoption needs to be sequenced after the data exists. Pushing CRediT-aware dossier templates before VIVO surfaced the data created a credibility problem that took a full year to recover from in the second cohort.
• 05Following the CASRAI Dictionary domain mapping (rather than inventing local terms) meant the ORCID integration was a configuration exercise, not a translation project.

What's next

Planned next steps

Year 2 focuses on retrospective CRediT enrichment for the previous five years of outputs — a known-hard problem, since contributor memory degrades fast — and on extending the Symplectic harvester to handle the eight additional publishers that have committed to structured CRediT capture in their 2026 product roadmaps. The library is also piloting a CRediT-aware narrative-CV pipeline modelled on the UKRI R4RI structure for faculty applying to the Howard Hughes Medical Institute and Gates Foundation, both of which now accept narrative-CV components.

Q&A with the composite project lead

Composite project-lead Q&A

The questions and answers below are composite — synthesised from interview patterns across multiple real project leads. They are not attributed to a specific real person.

Why start with the publisher harvester rather than the deposit form?

Speed of coverage. Roughly 28% of our annual output came through publishers that already emit structured CRediT in their Crossref deposit. Reading that metadata gave us a four-month head start on demonstrating value, which mattered for the internal budget cycle. We could not have funded the deposit-form work without that early demonstration.

Did you mandate CRediT capture or keep it optional?

Optional, with strong defaults. The provost's office considered a mandate but the listening tour surfaced significant variation across colleges — humanities faculty in particular had concerns about how CRediT maps onto sole-authored work. The default became "the deposit form pre-fills CRediT roles you can confirm or correct," which lifted capture rates faster than a top-down policy would have.

How did the tenure-and-promotion office react?

Cautiously at first. The committee chair had previously seen contribution-aware dossier templates fail elsewhere because the data was incomplete or self-reported without verification. We delayed T&P rollout until VIVO was surfacing CRediT for at least three semesters of recent publications, and we made the committee's training a hands-on session with real (anonymised) dossiers. After that the adoption flipped from skepticism to enthusiasm.

What was the biggest single mistake?

Underestimating the post-launch help-desk load. We staffed two semesters but should have staffed three. The questions were not technical; they were about how to handle multi-PI submissions where the lead PI disagreed with the first author about role assignments. We are now training the library helpdesk to mediate those conversations rather than answer them.

Would you do anything differently?

Yes — we would have built the VIVO profile surfacing before the Symplectic harvest, not after. Authors needed to see what the system was going to do with their data before they would trust the capture step. The ordering we used (harvest first, surface second) cost us roughly six months of skeptical PI engagement we did not need to have.

Cited CASRAI resources

Internal CASRAI resources referenced

• ›Dictionary: CRIS (Current Research Information System)
• ›Dictionary: ROR (Research Organization Registry)
• ›Dictionary: NSPM-33 disclosure requirements
• ›Dictionary: Narrative CV
• ›CRediT role: Conceptualization
• ›CRediT role: Writing — original draft
• ›CRediT for institutions
• ›Implement: Crossref deposit + CRediT
• ›NIH CRediT statement guidance
• ›CRIS integration notes (incl. Symplectic Elements)
• ›Implementation checklist: research-info systems
• ›CRediT adoption — publisher tracker