Open Data Manual › Practical guides › The FAIR Data Principles Explained

The FAIR Data Principles Explained

By Open Data Manual editors · Published 4 November 2026 · Editorial standard

The FAIR principles are a short set of guidelines for making data Findable, Accessible, Interoperable and Reusable. They started in research data management but are now cited by funders, publishers and governments far beyond science. This guide goes through all fifteen principles in plain language and explains the point that causes the most confusion: FAIR is not the same thing as open.

Where FAIR comes from

The principles were published in March 2016 as "The FAIR Guiding Principles for scientific data management and stewardship" by Mark Wilkinson, Michel Dumontier and colleagues in Scientific Data (volume 3, article 160018, doi:10.1038/sdata.2016.18). The authors came from academia, industry, funders and publishers. A distinctive feature of FAIR is its emphasis on machines: data should be findable and usable by software acting with little human help, because the volume of data is now far beyond what people can sort by hand.

The full list is maintained by the GO FAIR Foundation. Quotations below follow that list.

Findable

• F1. "(Meta)data are assigned a globally unique and persistent identifier." In practice this usually means a DOI or another persistent URL that will not break when a website is reorganised.
• F2. "Data are described with rich metadata." Enough description for someone to judge whether the data is relevant without downloading it.
• F3. "Metadata clearly and explicitly include the identifier of the data it describes." The record and the dataset must be linked unambiguously.
• F4. "(Meta)data are registered or indexed in a searchable resource." A dataset no catalogue knows about is effectively invisible; the manual's chapter Make data discoverable makes the same argument.

Accessible

• A1. "(Meta)data are retrievable by their identifier using a standardized communications protocol." HTTPS is the obvious example.
• A1.1. "The protocol is open, free, and universally implementable."
• A1.2. "The protocol allows for an authentication and authorization procedure, where necessary." This is where FAIR explicitly allows restricted access.
• A2. "Metadata are accessible, even when the data are no longer available." If a dataset is withdrawn, its description should remain so that citations still resolve to something meaningful.

Interoperable

• I1. "(Meta)data use a formal, accessible, shared, and broadly applicable language for knowledge representation." RDF and well-documented JSON or CSV schemas are common choices.
• I2. "(Meta)data use vocabularies that follow FAIR principles." Shared code lists and controlled vocabularies, themselves published with identifiers.
• I3. "(Meta)data include qualified references to other (meta)data." Links that say how two resources relate, not just that they do.

Reusable

• R1. "(Meta)data are richly described with a plurality of accurate and relevant attributes."
• R1.1. "(Meta)data are released with a clear and accessible data usage license." See our guides to CC0 and CC BY 4.0.
• R1.2. "(Meta)data are associated with detailed provenance." Who produced the data, how and when, and what was done to it.
• R1.3. "(Meta)data meet domain-relevant community standards." Use the formats and vocabularies your field already uses.

FAIR vs open data

Open data, as defined by the Open Definition and discussed in the manual's chapter What is Open?, is data anyone can freely use, reuse and redistribute, subject at most to attribution and share-alike. FAIR asks a different question: can the data be found and processed reliably? The two overlap but neither implies the other.

	Open data	FAIR data
Main concern	Legal and practical freedom to reuse	Findability and machine-actionability
Access control allowed	No	Yes, where necessary (A1.2)
Licence	Must be an open licence	Must be clear (R1.1), not necessarily open
Metadata and identifiers	Encouraged	Central requirement

Clinical trial data is the classic example of data that can be FAIR without being open: it can carry persistent identifiers, rich metadata and a clear access procedure, while access itself is restricted to protect participants. Conversely, a spreadsheet posted on a website under an open licence but with no metadata or stable address is open but scores poorly on FAIR.

The European Commission's research programmes summed up the relationship with the phrase "as open as possible, as closed as necessary": make data FAIR in all cases, and open wherever there is no good reason to restrict it.

Applying FAIR in practice

1. Give each dataset a persistent identifier. Repositories such as Zenodo or institutional repositories issue DOIs; government portals can commit to stable URLs.
2. Write metadata to a recognised standard. DCAT for catalogues, schema.org Dataset for search engines, and domain standards where they exist. See metadata standards for open data.
3. Register the dataset in at least one catalogue. A national portal, a domain repository, or both.
4. Choose open, well-documented formats. CSV with a column description or JSON with a schema; see CSV vs JSON.
5. Attach a licence and provenance. State who created the data, the method, the period covered and the processing applied.
6. Keep the metadata after retiring the data. Replace the download with a notice explaining what happened, rather than deleting the page.

Common misunderstandings

• "FAIR is a certification." It is not. There are assessment tools and maturity indicators, but the principles themselves are guidance.
• "FAIR means using RDF." The principles are technology-neutral; they require a shared, formal language, which can be a documented schema in any format.
• "Putting files in a repository makes them FAIR." A repository helps with identifiers and indexing, but interoperability and reusability depend on the metadata, vocabularies and licence the publisher supplies.

Frequently asked questions

What does FAIR stand for?

Findable, Accessible, Interoperable and Reusable. The principles were published in 2016 by Wilkinson and colleagues in the journal Scientific Data.

Is FAIR data the same as open data?

No. FAIR is about how well data and metadata can be found and processed, by people and machines. FAIR data can sit behind authentication and access control (principle A1.2), while open data must be free for anyone to use and redistribute.

How many FAIR principles are there?

Fifteen: F1 to F4, A1 (with sub-principles A1.1 and A1.2) and A2, I1 to I3, and R1 (with sub-principles R1.1 to R1.3).

Does FAIR require a particular file format or repository?

No. The principles do not prescribe technologies. They ask for persistent identifiers, rich metadata, standard protocols, shared vocabularies, a clear licence and provenance, and leave the choice of tools to each community.

More practical guides

• CSV vs JSON
• Metadata standards (DCAT)
• The CC0 license
• CC BY 4.0 for data
• Open Database License (ODbL)
• Database rights