The FAIR Principles (Findable, Accessible, Interoperable, Reusable) describe what a well-managed dataset should look like. They do not, on their own, tell you whether a specific dataset actually meets that bar. F-UJI is the tool most commonly used to close that gap: it takes a dataset’s persistent identifier, automatically tests the metadata and access behaviour behind it, and returns a machine-generated FAIRness score with a detailed report of what passed and what didn’t. This guide covers F-UJI specifically — what it tests, how the scoring works, who built and maintains it, and how a data manager or repository operator uses it in practice. For the underlying principles themselves, see the FAIR Data Principles glossary entry and the step-by-step checklist for making a dataset FAIR. For how F-UJI compares to the other assessment options on the market, see the CASRAI overview of FAIR data assessment frameworks.
What F-UJI is
F-UJI (the FAIRsFAIR Research Data Object Assessment Service — “F” for FAIR, “UJI” meaning “test” in Malay) is an open-source, automated tool that assesses the FAIRness of a single research dataset from its persistent identifier alone. Give it a DOI (or another resolvable PID) and F-UJI retrieves the associated metadata, follows the access and typing information it can discover, and runs a battery of automated tests against a defined set of FAIR metrics. It returns a numeric score, broken down by FAIR principle, along with a log explaining exactly which checks passed and which failed and why.
The tool was developed by Anusuriya Devaraju and Robert Huber at PANGAEA (the data publisher hosted at the University of Bremen and Alfred Wegener Institute) as part of the FAIRsFAIR project (“Fostering FAIR Data Practices in Europe”), an EU Horizon 2020-funded initiative (grant agreement 831558, 2019–2022) tasked with turning the abstract FAIR principles into practical, adoptable guidance and tooling for European research infrastructure.
How F-UJI’s assessment works
F-UJI does not ask a human to self-report FAIRness — that is what tools like DANS’s FAIR-Aware are for. It tests machine-discoverable evidence directly:
- Persistent identifier resolution and persistence — whether the PID resolves, uses a recognised, globally unique scheme (DOI, Handle), and is registered with a resolution service.
- Metadata retrieval and richness — whether machine-readable metadata can be retrieved (via content negotiation, embedded structured markup, or a metadata API), and how complete that metadata is against core descriptive elements.
- Use of structured, standardised metadata schemas and controlled vocabularies — e.g. whether the record uses a recognised schema (such as DataCite’s metadata schema or schema.org’s
Datasettype) rather than free-text description only. - Licence and access-condition clarity — whether a machine-readable licence is declared, and whether access protocols and conditions are explicit.
- Data typing, format, and provenance information — whether the file format is identified and documented, and whether the record links to related resources (source code, related publications, other datasets) in a machine-traceable way.
These checks are organised against the FAIR sub-principles (Findable F1–F4, Accessible A1–A2, Interoperable I1–I3, Reusable R1.1–R1.3) as operationalised in the FAIRsFAIR Data Object Assessment Metrics — a set of roughly 16–17 metrics that F-UJI implements as automated tests. Each metric contributes to that principle’s sub-score, and the sub-scores roll up into the overall FAIR score shown in the report. Because everything F-UJI checks has to be discoverable by a machine, a dataset can score well on F-UJI while still being poorly documented for a human reader — the tool measures machine-actionable FAIRness, not documentation quality or scientific validity, and CASRAI’s guidance is to treat it as one input alongside curator review, not a replacement for it.
How the FAIRsFAIR metrics relate to the RDA FAIR Data Maturity Model
F-UJI’s metrics were built to align with the RDA FAIR Data Maturity Model, the Research Data Alliance’s 41-indicator rubric for FAIR assessment finalised in 2020. The Maturity Model defines what “FAIR” means operationally, at four maturity levels per indicator, but deliberately does not prescribe how to test for it. F-UJI is, in effect, one automated implementation of that rubric at the level of a single dataset object — which is why its results are broadly comparable to assessments produced by other tools built against the same indicator set, even though the specific checks and scoring weights differ.
Using F-UJI in practice
F-UJI is available in three ways, which matter for different roles in a research-data management workflow:
- Hosted web service (f-uji.net) — paste in a single DOI or PID and get an interactive report. This is the right entry point for a researcher or data steward checking one dataset before or after deposit, or for spot-checking a specific record flagged in a compliance review.
- REST API — the same assessment engine exposed programmatically, which is what makes F-UJI usable at scale. A repository operator or institutional data office can script assessments across an entire collection of DOIs — for example, auditing every dataset deposited under a given grant, or benchmarking a repository’s overall FAIR posture as part of a CoreTrustSeal certification self-assessment.
- Self-deployable source / container — the code is open source (maintained on GitHub, originally under the PANGAEA data-publisher organisation and continued through the follow-on FAIR-IMPACT project), so an institution that wants to run assessments on infrastructure it controls, or extend the metric set for a domain-specific need, can deploy its own instance rather than depend on the hosted service.
In practice, research-data managers use F-UJI at a few distinct points:
- Pre-deposit or post-deposit QA — running a new dataset’s DOI through F-UJI shortly after minting it catches concrete, fixable problems (a missing machine-readable licence, an undeclared metadata schema, a dead landing-page link) before they become a permanent record.
- Funder and DMP compliance checks — where a data management plan commits to FAIR-compliant sharing, F-UJI gives an institution a defensible, repeatable way to check whether that commitment was actually met at project closeout, rather than relying on self-report.
- Repository-level benchmarking — running F-UJI across a sample (or all) of a repository’s holdings surfaces systemic metadata gaps — e.g. a repository whose records consistently score poorly on interoperability because it doesn’t expose a standard schema — that are worth fixing once, at the repository configuration level, rather than dataset by dataset.
Limitations to keep in mind
F-UJI assesses what is machine-discoverable at the moment of the test; it cannot judge whether documentation is actually clear to a human reader, whether the data itself is scientifically sound, or whether a dataset is FAIR in a discipline-specific sense that its generic metrics don’t capture. A perfect F-UJI score does not guarantee a dataset is easy to reuse, and an imperfect score does not necessarily mean a dataset is unusable — some legitimate access restrictions (e.g. controlled-access human-subjects data, appropriately reflected in machine-readable access conditions) will not score full marks on Accessibility by design. Treat F-UJI as a diagnostic that catches concrete, fixable metadata problems at scale, not as a certification of research quality.
Frequently asked questions
Is F-UJI free to use?
The hosted service at f-uji.net is free to use for individual dataset lookups, and the underlying source is open source, so an institution can self-host it without licensing cost.
Does F-UJI work on any dataset, or only ones in specific repositories?
F-UJI works on any dataset with a resolvable, globally unique persistent identifier (typically a DOI) and publicly retrievable metadata — it is not restricted to a specific repository, though how well a dataset scores depends heavily on how much machine-readable metadata that repository exposes.
How is F-UJI different from FAIR-Aware?
FAIR-Aware, developed by DANS, is a structured self-assessment questionnaire aimed at researchers before they deposit a dataset — it is pedagogical, not evaluative. F-UJI instead fetches and automatically tests the actual metadata and access behaviour of an already-identified dataset, producing an objective, repeatable score rather than guided self-reflection.
Who maintains F-UJI today?
F-UJI was originally developed under the FAIRsFAIR project, which concluded in 2022. Development and maintenance have continued through PANGAEA’s public GitHub repository and the follow-on FAIR-IMPACT project, which has also extended the tool’s metrics (for example, work on assessing research software rather than only datasets).
Does a high F-UJI score mean a dataset satisfies a funder’s FAIR data requirements?
Not automatically. F-UJI measures machine-actionable FAIRness against a defined metric set, which is useful evidence for a compliance review, but funder policies (for example under Horizon Europe or NIH data-sharing requirements) may include expectations F-UJI doesn’t test for, such as the substance of a data management plan’s justification for its choices. Use F-UJI as supporting evidence, not a substitute for reading the actual funder policy.
Related CASRAI resources
- FAIR Data Principles — the underlying framework F-UJI tests against.
- How to Make Your Dataset FAIR: A Step-by-Step Checklist — practical steps to improve a dataset’s FAIRness before or after assessment.
- FAIR data assessment frameworks: a buyer’s guide for institutions — how F-UJI compares to the RDA Maturity Model, FAIR-Aware, and other assessment options.
- FAIR Digital Object (FDO) — a related, emerging machine-actionable framework for FAIR data and metadata.
- CoreTrustSeal Certification for Research Data Repositories — repository-level trustworthiness certification that complements dataset-level FAIR assessment.
- DMP assessment — how data management plan commitments are checked against actual outcomes.
- FAIR vs CARE — data principles compared.
