In 2023, the second edition of the Finnish National Policy and Executive Plan for Open Access to Research Data was published. This policy was co-authored by a broad coalition of stakeholders from the higher education and research community. Within the policy, research data is defined as follows:

This definition positions research data as a broad and inclusive concept, encompassing not only datasets in the traditional sense but also research methods, software, research infrastructures, and related materials. Crucially, it is the context in which the data is used that determines its classification as research data. In other words, data becomes research data when it is employed within a research process to generate or validate scholarly or artistic knowledge.

Data publishing refers to the formal process of making research data publicly available in a structured, accessible, and citable form, typically through a trusted data repository or a data journal. It involves the application of persistent identifiers (such as DOIs), standardised metadata, and documentation to ensure that the dataset can be discovered, accessed, reused, and cited by others.

Unlike informal data sharing, data publishing treats datasets as legitimate research outputs, often subject to quality control or peer review. It supports transparency, reproducibility, and the broader goals of open science by enabling long-term preservation and scholarly recognition of research data.

The National Library of Medicine (NLM) defines data publication:

According to NLM, a data publication can be either an article discussing research data or a journal that publishes articles about research data. Their definition also includes the requirement of peer review process.

It is important to note that open data is not synonymous with data publication. While both concepts relate to the accessibility of research data, data publication entails a more formalised process that aligns with established scholarly practices. For instance, data publication involves elements such as persistent identifiers and mechanisms for citation—features that are not necessarily present in all instances of open data. As outlined in Data citation: A guide to best practice (Publications Office of the European Union, 2022), the citability of data is a key criterion distinguishing data publication from more general forms of data sharing.

The concept of data publication has been discussed in scholarly literature since at least 2006. Klump et al. (2006) proposed two fundamental prerequisites for data to be considered formally published:

These criteria underscore the notion that data publication is not merely about making data available, but about ensuring that it meets standards of discoverability, reliability, and scholarly utility. In this sense, data publication positions research data as a legitimate and citable research output, akin to traditional publications.

The concept of data publication has been examined more closely in an article published by Parsons & Fox (2013): “Is data publication the right metaphor?” The article comprehensively explores the complexity of the concept and considers alternative terms that may better describe the phenomenon. Terms mentioned as closely related or at least partially overlapping include data citation, data quality, data preservation, data archives and research infrastructures.

Data journals publish articles about existing datasets, databases or other kinds of data collections. Some journals also serve as a platform for publishing data, while some have articles that refer to data located elsewhere (e.g., in a domain-specific data repository).

Data articles are a distinct form of scholarly publication dedicated to the detailed description of research datasets. They are also referred to as data papers, data descriptors (as used in Scientific Data)³, dataset articles, or metadata articles. While terminology may vary across journals, the underlying concept remains consistent.

As Callaghan et al. (2012) describe, a data paper “would describe the dataset, providing information on the what, where, why, how and who of the data. The data paper would contain a link back (a DOI) to the dataset in its repository, and the journal publishers would not actually host the data. This means that even in situations where the data paper might be restricted access, the dataset could still be open.”

Similarly, Data in Brief⁴ (Elsevier) defines a data article as “a short description of research data that have been made publicly available through a repository that makes it easier to comprehend and reuse. It does not offer conclusions or interpretive insights. Data articles give scientists the opportunity to describe and share their raw data and hence participate in open science and satisfy funder requirements.”

Regardless of the term used, the purpose is the same. Unlike traditional research articles, data articles do not present hypotheses, results, or theoretical interpretations. Instead, they focus on providing comprehensive metadata, contextual information, and technical validation to support the reuse, reproducibility, and proper citation of the dataset.⁵

Typically, data articles include:

Data articles are peer-reviewed and citable, offering researchers formal recognition for their data contributions.

It is generally understood that research data requires a peer review process if it is to be considered comparable to traditional scholarly publications (Callaghan et al., 2012). However, peer review of research data differs in important ways from the peer review of conventional journal articles. In the context of data, the emphasis shifts towards aspects such as curation, archiving, and data quality (Parsons & Fox, 2013).

Notably, the terminology and practices surrounding data peer review remain unsettled. Despite growing recognition of the distinct nature of data review, even recent formal definitions—such as those in Standard Terminology for Peer Review (National Information Standards Organization, 2023)—continue to focus primarily on traditional article-based peer review. In practice, data peer review varies considerably depending on the context, whether it occurs in relation to traditional journal articles, data papers, or open-access data repositories.

Mayernik et al. (2015) identify three key challenges in the development of effective data peer review: variability in data accessibility, the need for specialised expertise, and the balance between pre- and post-publication review. Addressing these challenges is essential for enhancing scientific integrity and ensuring that data publication processes scale effectively with increasing data volumes.

Peer review of research data may occur as part of the review process for traditional research articles, where publishers may ask reviewers to assess the underlying data (e.g. Springer Nature’s research data policy). However, this practice remains relatively uncommon and, as noted by Pop and Salzberg (2015), supplementary data are rarely reviewed—even when explicitly included in reviewer guidelines.

Dedicated data journals, such as Scientific Data (Springer Nature), offer a more structured approach. In these venues, peer review not only assesses the clarity and completeness of the data description but also considers other aspects such as usability, accessibility, and compliance with metadata standards. As Carpenter⁶ notes, further guidance is emerging to support consistent and transparent peer review of data, though practices remain diverse.

Mayernik et al. (2015) further distinguish between four models of data publication—traditional journals, data journals, data repositories, and hybrid approaches—and examine how each contributes to data quality assurance. Despite differences in review mechanisms, all models share core requirements: data accessibility, adequate documentation, and clear reviewer guidance. Data journals, often in collaboration with repositories, provide the most formalised peer review structures. In contrast, traditional journals and repositories frequently lack consistent standards. Tools such as metadata validation, visualisation platforms, and statistical checks can support reviewers, though challenges persist, particularly with large or complex datasets.

To strengthen data peer review, Mayernik et al. recommend that scientific societies update author and reviewer guidelines, promote collaboration with repositories, and endorse established data citation principles. Broader adoption of practices such as integrated submission pipelines and rating systems could further enhance transparency and data quality. Looking ahead, key priorities include ensuring accessibility, engaging qualified reviewers, and balancing pre- and post-publication review in ways that scale with the growing volume and diversity of research data.

Data curation is a multifaceted concept that can be approached from at least two complementary, yet distinct, perspectives:

Together, these perspectives illustrate that data curation is both a data management process and a strategic activity. It involves not only managing data over time but also making informed decisions about what data should be preserved, how it should be enhanced, and under what conditions it should be shared or published (Johnston et al., 2024; Lee & Stvilia, 2017).

Curated repositories for research data refer to carefully managed and quality-controlled collections, often organised around a specific discipline, theme, or data type. Unlike raw or minimally processed data repositories, curated archives involve active human oversight to ensure that datasets are accurate, well-documented, consistently formatted, and suitable for reuse. Curation typically includes the verification of metadata quality, the usability of file formats, the clarity of data structures, and the presence of persistent identifiers, licensing information, and citation guidance.

However, there is no single curated repository that accommodates all types of research data. Each service defines its own standards for curation, appraisal, and selection. Since the introduction of the FAIR principles (Wilkinson et al., 2016), many repositories have adopted these as benchmarks, particularly regarding machine readability and interoperability.

An illustrative example of a curated data archive is the Finnish Social Science Data Archive (FSD), which applies rigorous appraisal criteria focused on technical quality, documentation, and legal rights (see FSD, Operational guidelines)⁷. Another widely recognised example is RefSeq (Reference Sequence Database), which hosts curated DNA, RNA, and protein sequences. RefSeq employs a two-stage curation process: an initial phase of automated validation followed by expert-led quality assurance. This hybrid model is supported by a broad international network, enabling the rapid release of data while ensuring continuous improvement as new information becomes available. The curation status of each dataset is clearly communicated to users through standardised indicators, with explicit labels such as “Reviewed”, “Provisional”, “Model”, or “Predicted”. These labels help users assess the reliability and level of curation applied to each record.

In the context of research data, the terms data repository and data archive are closely related and often used interchangeably. Formal definitions of these terms are absent from the scholarly literature, and discussions concerning their potential distinctions are found only in blog posts.¹¹,¹² The terms may carry slightly different connotations depending on the context and institutional practices.

A data repository typically refers to a platform or service where research data can be stored, managed, shared, and accessed. Repositories emphasise active use and discoverability, and they often support metadata standards, persistent identifiers (such as DOIs), and access control mechanisms. Examples of widely used repositories include Zenodo, Figshare, Dryad, and Harvard Dataverse.

In contrast, a data archive generally implies a stronger focus on long-term preservation, with an emphasis on ensuring the durability, integrity, and future accessibility of data. Archives may involve more rigorous curation processes and preservation strategies. Data archives provide a variety of benefits including the preservation, discovery, control, reuse and repurposing of research data (Rieger, 2007). Notable examples include the UK Data Archive, ICPSR, and the CESSDA network of archives.

While the distinction between data repositories and archives is not always clear-cut, repositories are generally associated with enabling immediate access and reuse of research data, whereas archives are more closely aligned with its long-term stewardship and preservation. In our survey, however, the terms data archive and data repository were frequently used interchangeably to refer to repositories—that is, platforms designed for data sharing and citation—rather than to long-term preservation services, which will be discussed in the following section.

A long-term data preservation service refers to an infrastructure or system designed to ensure the sustained accessibility, usability, and integrity of selected research data over extended periods—long beyond the duration of the original research project. These services typically implement recognised preservation standards, conduct regular integrity checks, apply format migration strategies, and maintain secure storage environments to protect data from technological obsolescence and degradation (Albani et al., 2020).

Importantly, not all data can or should be preserved indefinitely. Due to resource constraints and the need for curation, data selected for long-term preservation is typically assessed based on its scientific value, legal or ethical obligations, potential for reuse, and relevance to institutional or disciplinary priorities. This selection process is often guided by appraisal criteria and may involve collaboration between researchers, data stewards, and archivists. Such services are frequently provided by institutional or national archives and may be certified under frameworks such as the CoreTrustSeal, ensuring adherence to best practices in digital preservation (CoreTrustSeal Standards & Certification Board, 2022).

A data repository and a long-term preservation service serve distinct roles in research data management. A data repository focuses on the dissemination, discoverability, and reuse of data over the short to medium term. It supports metadata standards and often allows versioning to facilitate open science and data citation. In contrast, a long-term preservation service ensures the integrity, authenticity, and accessibility of data over decades or even centuries. It employs preservation strategies such as format migration and checksums, and typically adheres to frameworks like the Open Archival Information System (OAIS)¹³. While repositories promote visibility and accessibility of research outputs, long-term preservation services safeguard the longevity and usability of data for future generations.

Data catalogues are structured, searchable indexes that provide standardised metadata descriptions of research datasets, thereby facilitating their discovery, citation, and potential reuse. Rather than storing the data itself, a data catalogue functions as a centralised metadata repository, aggregating information from multiple sources such as institutional repositories, disciplinary archives, and national data services. Typical metadata elements include dataset content, provenance, format, access conditions, and responsible parties (Labadie et al., 2020).

By consolidating metadata from diverse origins, general-purpose data catalogues enhance the visibility and governance of research data assets across institutional and disciplinary boundaries. They play a pivotal role in supporting the FAIR principles—particularly Findability and Accessibility—and are often maintained by national infrastructures, research consortia, or domain-specific services committed to advancing open science and responsible data stewardship.

In contrast, institutional data catalogues are developed and maintained by individual research organisations, such as universities or research institutes, to showcase and manage the datasets produced by their own researchers. These catalogues contribute to institutional research profiling by highlighting ongoing research activities and outputs. They may also reveal interdisciplinary connections by identifying thematic overlaps across departments or faculties. Improved awareness of existing datasets within the institution can promote internal data reuse and foster collaboration.

Institutional data catalogues may be implemented either by integrating dataset metadata into existing Current Research Information Systems (CRIS) or by developing dedicated platforms specifically designed for managing research data. However, building a data catalogue is not necessarily straightforward, and initial versions may fall short of expectations. As Rumsey and Jefferies (2013) observed in their account of developing a data catalogue for the University of Oxford, the process can be complex and iterative.

If research data is to be recognised as a valuable scholarly output, it must be made more visible within both institutional and broader research ecosystems. Establishing and maintaining a data catalogue—whether general or institutional—is a critical step towards achieving this goal. Moreover, once metadata is systematically collected, tools such as altmetrics can be employed to monitor and demonstrate the broader impact and reach of research data.

As this type of service is still largely in the conceptual stage in most libraries, there is currently little to no literature available on the subject. Nevertheless, we chose to include it in our study in order to explore the extent to which libraries are beginning to engage with data journal related services.

In principle, libraries can support researchers in publishing data articles in dedicated data journals. This support may include helping researchers identify trusted repositories, ensuring appropriate metadata and persistent identifiers, which can be linked to the data article. Libraries can also advise researchers on selecting suitable data journals and crafting data articles that meet both scholarly and technical standards. Additionally, they may offer guidance on licensing, ethical considerations, and compliance with open science policies.

Beyond their advisory role, many academic libraries are also active participants in scholarly publishing. Some operate as university presses or collaborate with institutional publishing platforms, and in some cases, they may even host or publish data journals themselves.

Libraries are increasingly recognised as key actors in supporting the use, development, and sustainability of domain-specific data repositories. Their contributions may include guiding researchers in selecting appropriate repositories, assisting with metadata creation and compliance with disciplinary standards, and providing training on data deposition practices. Libraries may also collaborate with repository infrastructures to ensure interoperability with institutional systems, support compliance with funder and institutional policies, and contribute to data curation and quality assurance.

In addition to these operational roles, libraries can advocate for the value of domain-specific repositories and help sustain them through institutional partnerships or funding mechanisms. By engaging in these activities, libraries enhance the visibility, accessibility, and long-term value of disciplinary research data, thereby contributing to the broader goals of open science and responsible data stewardship.

Experiences from the EOSC-Nordic project have shown that such repositories do exist within institutions and that libraries have supported efforts to improve their FAIRness and to pursue CoreTrustSeal certification (Alaterä et al., 2022; CoreTrustSeal Standards & Certification Board, 2022; Meerman et al., 2021). Libraries are well positioned to promote the use of persistent identifiers such as DOIs, ensure that repositories are registered in international directories such as re3data and FAIRsharing, and raise awareness of certification schemes. Where needed, libraries can also facilitate access to expertise to support certification applications.

Moreover, libraries can act as ambassadors for the FAIR principles, offering practical guidance on their implementation. Tools such as the FAIR Evaluation Service (F-UJI)¹⁴, tested in the EOSC-Nordic project, allow users to assess the FAIRness of datasets by entering a URL or persistent identifier (Alaterä et al., 2022). The tool generates a FAIRness score and provides targeted feedback, highlighting both strengths and areas for improvement. This feedback can serve as a valuable starting point for discussions on data quality and machine-actionability—an aspect often overlooked when assessments are made solely from a human perspective.

The repositioning of research data from a supplementary resource to a primary scholarly output has been underscored. This shift necessitates a redefinition of institutional responsibilities and a broader understanding of academic publishing. If data were more widely regarded as publishable content—subject to peer review, persistent identification, and quality control—its role in research evaluation and scholarly communication could be significantly enhanced. Equally important is the citability of research data, which ensures that datasets receive appropriate academic recognition and can be reliably traced and reused. Enhancing data citation practices not only supports transparency and reproducibility but also incentivises data sharing by formally acknowledging the intellectual contribution of data creators.

The influence of national and international data policies on the discourse of service providers, as compared to researchers’ own conceptualisations, warrants further investigation. When seeking to position research data alongside other scholarly outputs, the terminology employed becomes particularly significant. If data policies were to explicitly refer to the publication of data, this could help raise awareness that research data is not solely to be opened and shared—it can also be formally published.

In this context, scholarly publishers are uniquely positioned to advance the recognition of research data as a legitimate academic output. By mandating data availability statements (DAS) and requiring authors to cite underlying datasets, publishers can help embed data sharing into the fabric of scholarly communication. This, however, presupposes that data are made available in a citable form, with persistent identifiers and adequate metadata. Even in cases where access to the data must be restricted, the publication of metadata remains essential to ensure discoverability and transparency. Furthermore, peer reviewers should be encouraged—and guided—to consider the quality, accessibility, and relevance of the supporting data as part of the review process. Such measures would not only reinforce the scholarly value of data but also foster a more robust and accountable research ecosystem.

Research libraries have been identified as key actors in facilitating the publication of research data through services such as institutional repositories, long-term preservation infrastructures, data catalogues, and support for data journals and domain-specific repositories. Their role is both strategic and operational, requiring coordination of services and alignment with institutional research goals.

However, notable gaps remain in areas such as data appraisal, quality assurance, and the development of sustainable service models. Collaborative service models that align with disciplinary practices and institutional capacities are essential. Libraries must support researchers in navigating data publication pathways, promote the use of data journals, and ensure that metadata standards are upheld to maximise discoverability and reuse.

The integration of data publishing into research assessment frameworks remains limited. Without robust metrics and cataloguing systems, the scholarly impact and reuse of datasets cannot be effectively measured or recognised. Libraries must therefore advocate for the inclusion of data outputs in evaluation systems and support the development of appropriate indicators.

While libraries already play a pivotal role in research data management—through guidance, training, and coordination of multiprofessional support—an exclusive focus on data management risks overlooking the broader benefits of FAIR data and the importance of quality assurance.

A data catalogue serves to record the metadata of research data produced within an institution. In the absence of such a catalogue, the establishment of effective assessment practices becomes significantly more challenging. Beyond supporting the development of metrics, catalogues are essential for identifying datasets containing personal information (in accordance with GDPR requirements) and those requiring long-term preservation.

At present, many CRIS systems do not support the construction of high-quality data catalogues, as they often lack compatibility with established metadata standards. It is the responsibility of libraries to ensure that metadata adhere to international standards and schemas, thereby enhancing interoperability and supporting broader discoverability and reuse.

The emergence of research data publishing as a formal scholarly activity presents new challenges to existing skill sets, service models, and business frameworks within academic libraries. While libraries already possess expertise in knowledge management, long-term accessibility, and open access publishing, further development is required in areas such as data quality, documentation, and metadata. These competencies are critical for effective data appraisal and selection and extend beyond traditional librarianship.

To support the development of these new competencies, sustained and deep collaboration is essential. Initiatives such as the Advancing RDM Careers: A Framework for Expert Education in Finland (2024) provide valuable frameworks for guiding this evolution.

The implementation of services such as long-term data preservation necessitates close cooperation with institutional stakeholders, including IT services (for technical infrastructure and data security), university leadership (for policy alignment), data protection officers (for GDPR compliance), and legal experts (for data rights management). Within this collaborative environment, libraries are expected to act as active expert partners.

Proactive engagement requires that library professionals are afforded opportunities to familiarise themselves with academic practices from the perspective of research data. The ability to specialise as data management experts enables the delivery of high-quality, targeted services that directly benefit the research community. Investment in such specialisation is not only beneficial but essential, as it facilitates the acquisition of the necessary expertise to meet evolving institutional and disciplinary needs.

While the sample of the survey is neither large nor representative of all university libraries in the region, the responses appear to come predominantly from institutions that have been early adopters in the provision of research data management (RDM) services. Given that the survey was mainly distributed through mailing lists specifically aimed at RDM professionals—such as the DataCite list in Estonia, the Dataverse network in Norway, and the DataSupport chat in Finland—it is likely that the survey primarily reached libraries actively engaged in developing such services.

Had the survey been disseminated more broadly via general university library mailing lists, it might have reached institutions where RDM services are less developed or still emerging. It is also plausible that libraries with dedicated RDM personnel were more inclined to respond to a survey on this topic. This may help explain the absence of responses from libraries where such services are currently limited—institutions which, in our view, do exist within the region.

The primary aim of this study was to explore the nature and scope of data publishing services being offered, rather than to assess their overall distribution or uptake. The data nonetheless offer valuable insights into the types of services currently being provided in support of research data publishing, although the number of responses is modest. However, the limited sample size precludes any definitive conclusions regarding the prevalence of these services across the broader library landscape.