Consolidating a CRIS in twelve weeks.

The starting picture

The faculty had grown organically over fifteen years. Each department had developed its own way of tracking publications, grants and citations. By 2025 there were three live spreadsheets — one for the Faculty of Science, one for Humanities, one for the cross-faculty research office — plus two SharePoint sites with conflicting versions of the same data and one Access database that no one had owner access to anymore.

When the Vice-Rector for Research was asked to file a national accreditation return, the office spent six weeks reconciling the four sources by hand. The return was filed on time. The Vice-Rector resolved that this would be the last time.

What we changed first

Before any data migration, we did three things that took a week and broke nothing:

• One canonical author identifier per researcher. Where ORCIDs existed, we used them. Where they did not, we created internal IDs and committed to back-filling ORCID later.
• One canonical organisational unit tree. The Faculty had three different department names for the same physical office. We picked one and the others became aliases.
• One canonical output type taxonomy. Journal article, conference proceedings, book chapter, dataset. Where the spreadsheets distinguished between “conference paper” and “conference proceedings”, we collapsed them.

The twelve weeks

The full timeline ran in three phases. None of them involved heroic effort. All of them involved repetition.

• Weeks 1–3. Data audit and canonical schema. We loaded all four sources read-only into a staging area and ran reconciliation reports. Researchers were given individual reports of their outputs and asked to flag duplicates and miss-attributions. Most did.
• Weeks 4–8. Migration. We brought entities into the new CRIS in dependency order: people, then organisational units, then funding sources, then outputs, then grants. Each phase ended with a sign-off from the research office before the next began.
• Weeks 9–12. Workflow connection. We wired the CRIS to the institution’s Open Access repository so deposits flow automatically; to the HR system so author affiliations stay in sync; and to the funder reporting templates so the next accreditation return is a database query, not a six-week reconciliation.

What didn’t work

The first version of the deposit-at-acceptance workflow had four approval steps. It worked correctly. Authors found it tedious and most found a way around it. We removed three of the four steps in week 11. Acceptance-to-deposit time dropped from a median of twenty-two days to four.

The lesson, predictably, is that compliance designed for the audit fails when it slows down the author. The version that works is the version that the author barely notices.

Where it sits now

Twelve weeks after kickoff, the Faculty’s research record is a single source. The national accreditation return for 2026 was filed in four working days. The research office redirected the reclaimed six weeks of time to grant-writing support, which is what they had wanted to do all along.