# Data Sharing vs. Data Reuse: What's the Difference? Data sharing is the act of making research data available to other people, usually by depositing it in a repository with documentation. Data reuse is the act of taking data that someone else (or you, in a earlier project) already collected and using it to answer a new question, verify a published result, or combine with other datasets. Sharing is the supply side of the research data economy, and reuse is the demand side. They are two halves of one cycle, and each one depends on the other working well. Confusing the two terms causes real problems for students and early-career researchers. Funder policies talk about both, but they require different things from you. Sharing asks you to prepare, document, and deposit data. Reuse asks you to search, evaluate, harmonize, and analyze data that someone else produced. This article explains the data sharing vs data reuse difference in plain language, shows where each one fits in the research lifecycle, and gives you practical steps for doing both well. ## Data Sharing vs. Data Reuse: The Core Difference ![Comparison table of data sharing and data reuse across who acts, timing, main activity, key skills, success measure and obstacles.](/article-images/illustrations/data-sharing-vs-data-reuse-difference-sharing-vs-reuse.webp) The simplest way to hold the two concepts apart is to ask who is doing what, and when. **Data sharing** happens at or near the end of a project. The team that collected the data organizes it, writes documentation, chooses a repository or access route, and releases it under a license or access agreement. The goal is for the data to be findable, understandable, and usable by people who were not part of the original project. Modern guidance frames this around the FAIR principles, which stand for Findability, Accessibility, Interoperability, and Reusability, and around supporting practices such as documentation standards, [data management](/blog/guides/data-management-basics-principles-processes-and-best-practices) plans, and sharing the code and workflows that go with the data [1]. **Data reuse** happens later, often years later, and usually by someone with no connection to the original study. The reuser finds an existing dataset, judges whether it fits their question, combines it with other sources if needed, and analyzes it. Surveys across 20 research disciplines show that combining multiple existing datasets to answer new research questions is the most common form of reuse in practice [2]. Here is a side-by-side summary: | Dimension | Data sharing | Data reuse | |--|--|--| | Who acts | The original data producer | A new researcher or team | | When in the lifecycle | End of a project, at or after publication | Start of a new project or analysis | | Main activity | Curating, documenting, depositing, licensing | Searching, evaluating, harmonizing, reanalyzing | | Key skills | Data management plans, metadata, repository use | Database searching, [data wrangling](/blog/data-analysis/what-is-data-wrangling), critical appraisal | | Success measure | Data is findable and usable by strangers | New findings, verified results, or saved resources | | Typical obstacles | Time cost, privacy rules, unclear ownership | Datasets are hard to find, poor documentation, format mismatches | One more distinction helps: primary data is data collected fresh for a specific study, and secondary data is data originally collected for one purpose and used again for another. The reuse of secondary data is sometimes called secondary use, a term that appears constantly in health research policy [1]. ## Why the Distinction Matters Sharing and reuse fail in different ways, so they need different fixes. If sharing fails, the cause is usually on the producer's side. Data stays on a lab hard drive, documentation is thin, or the file formats are so idiosyncratic that no outsider can parse them. If reuse fails, the cause is usually on the consumer's side. Researchers cannot find relevant datasets, cannot tell whether a dataset is trustworthy, or cannot integrate it with their own data. Evidence shows the two behaviors feed each other. A survey of 476 biological scientists in the United States, analyzed with structural equation modeling, found that scientists with prior data-reuse experience rated the community and career benefits of sharing higher, saw sharing as less risky and less effortful, and felt stronger social pressure to share. Those beliefs in turn predicted their intention to share their own data [3]. In other words, people who have benefited from someone else's shared data become more willing to share their own. A large international survey of authors across 20 Scopus disciplines found the same pattern, with researchers who had reused data before being more likely to share their own data, and sharing and reuse both increasing over time, though neither is yet ubiquitous in any field [2]. This interconnection has a practical consequence for research leaders and funders. If you want more sharing, one lever is to make reuse easier and more rewarding, because reusers turn into sharers [3]. ## Data Sharing: What It Actually Involves Sharing is more than uploading a spreadsheet. Done well, it is a small project of its own. ### Documentation and metadata Metadata are descriptions of a dataset: what was measured, with what instruments, under what conditions, in what units, and with what known limitations. Good metadata is what turns a pile of files into a dataset a stranger can understand. Guidance on data sharing emphasizes that metadata is what allows data to be genuinely reused, not just downloaded [1]. For population health data specifically, researchers have proposed frameworks that make metadata machine-readable, using standardized formats such as the Data Documentation Initiative (DDI) Codebook and common data models such as OMOP, so that both computers and humans can discover, access, and integrate the data across platforms [4]. Documentation matters even when the data itself is imperfect. The Public Health Alliance for [Genomic Epidemiology](/knowledge/bioinformatics/genomic-epidemiology-integrating-pathogen-genomics-into-outbreak-investigations) (PHA4GE) developed standardized contextual data tags so that public health laboratories can share sequence datasets with known quality issues, clearly marked, rather than withholding them. Sub-optimal datasets can still be vital for validation, training, and pipeline testing, but only if the quality problems are annotated so that downstream users do not mistake them for high-quality reference data [5]. ### Repositories A data repository is a curated, searchable archive that stores datasets and makes them citable. The landscape includes generalist repositories, discipline-specific archives, institutional repositories, and journal-supported repositories. Survey evidence from across disciplines shows that the use of institutional and journal-supported repositories is increasing, but personal websites are still frequently used, which makes data harder to find and preserve [2]. Discipline-specific archives illustrate what good sharing infrastructure looks like. OpenNeuro, a brain imaging data archive built around the FAIR principles and the Brain Imaging Data Structure (BIDS) standard, hosts more than 600 datasets covering more than 20,000 participants across multiple species and measurement modalities. The shared data has already fueled more than 150 published reuses, and the BIDS standard is credited with making curation, sharing, and reuse effective [6]. In ecology, major archives such as Dryad and GBIF hold large volumes of deposited datasets that researchers draw on for synthesis work [7]. ### Licensing, consent, and access controls ![Flowchart showing clinical trial data access tiers from reduced datasets at publication to individual participant data under governance review.](/article-images/illustrations/data-sharing-vs-data-reuse-difference-tiered-access.webp) Sharing human-subject data adds legal layers. Under the General Data Protection Regulation (GDPR), the European data protection law, consent for secondary use of personal data in research has produced confusion, because research purposes often cannot be fully specified in advance. This has led to idiosyncratic consent models across [genome research](/blog/guides/genome-research) and clinical trials, and to interest in mechanisms such as data altruism, defined in the European Data Governance Act, as a way to close the gaps that consent-based approaches leave behind [8]. Access does not have to mean fully open. A tiered framework proposed for clinical trial data matches the level of access to the scientific purpose. Tier 1 provides reduced datasets at publication time that support endpoint-level numerical reproducibility. Tier 2 provides reproducibility packages with fuller data and code. Higher tiers provide individual participant data under governance review for deeper reuse. The point is that different scientific claims need different data objects, and transparency should be planned accordingly [9]. ## Data Reuse: What It Actually Involves Reuse looks different depending on the goal. Three patterns cover most cases. ### Verification and reproducibility Reusing data to check whether published results hold is a cornerstone of scientific self-correction, yet it remains rare in practice. One documented exception is a multi-team datathon in which five independent teams used R or Python to reproduce the key disease-free survival findings of the CAO/ARO/AIO-04 rectal cancer clinical trial from shared trial data. Most teams reproduced the main conclusions, which strengthened confidence in the trial, and the exercise also surfaced minor reporting issues in the original publication. The teams noted that supplementary material from the original study was not always easy to find, which is a sharing problem showing up as a reuse problem [10]. ### New questions from old data Pooling datasets can answer questions no single study could. In pediatric pharmacokinetics, the study of how drugs move through the body in children, pooling data from different studies enables a single robust analysis with covariate testing that no individual pediatric trial could support, while also saving costs and allowing new interpretations of old data [11]. In pharmacometrics more broadly, open initiatives that share data, models, and workflows aim to address new questions or give more precise answers to old ones by merging information across organizations, with disease-specific databases now operating in areas such as Alzheimer's disease, malaria, oncology, Parkinson's disease, and tuberculosis [12]. Population health research shows the scale reuse can reach. The HELICON project in Belgium links health surveillance records with administrative data sources to build a case-cohort covering 1.2 million randomly selected Belgians and 4.5 million Belgians with confirmed COVID-19, including 108,211 hospitalized patients, in order to study social inequalities and long-term health impacts. The protocol explicitly notes that interoperability and reuse of different data sources pose technical, administrative, and data security challenges [13]. ### Structured collections built for reuse Some projects are designed from the ground up for reuse efficiency. The Attentional Control Data Collection is a SQL database holding 64 attentional control datasets, accessible through a web app and an R package, that lets researchers pull a large body of data from their research area at once. An initial analysis of the collection showed that reliability of individual differences is generally low and suggested that future studies should use at least 400 trials per person and condition. A resource like this serves goals that ordinary one-off data deposits do not, supporting meta-analysis and study planning directly [14]. Linked open data resources work the same way across whole domains. Wikidata, a multilingual, collaboratively curated knowledge base, lets botanists contribute and query interconnected information about taxa, collectors, collections, publications, and chemical compounds through tools such as the Wikidata Query Service and Bionomia, enabling reuse that no single institutional database could support [15]. ## How Much Reuse Does Sharing Need? The Economics Sharing has real time costs: curation, documentation, formatting, and answering reuser questions. Reuse also has costs: searching, evaluating, harmonizing, and integrating external datasets. Theoretical "break-even" models ask at what level of community reuse the shared investment starts paying off in saved time. An empirical test in ecology combined a survey of 163 practicing ecologists with repository analytics on 320 datasets deposited in archives such as Dryad and GBIF, then used Monte Carlo simulation to estimate break-even reuse thresholds and to separate high-value from low-value data products. The study is notable because key parameters, such as the time needed to prepare data for reuse or to integrate external datasets, are rarely quantified for specific disciplines, and the results give ecology its first discipline-calibrated estimates [7]. The practical lesson for a student is simple. Sharing costs fall when you build documentation into the project from day one, and reuse value rises when datasets are deposited where searchers actually look, with metadata that answers the questions reusers ask [1], [2]. ## Data Repository Comparison: Where Sharing Happens Choosing where to share is one of the first decisions a data producer makes, and the options differ in reach, curation, and audience. Generalist and institutional repositories accept nearly any discipline and are often supported by universities or journals. Discipline-specific archives, such as OpenNeuro for neuroimaging [6] or Dryad and GBIF for ecology and biodiversity data [7], offer field-standard formats and community curation. Personal websites and lab pages remain common in some fields despite making data hard to discover and preserve over time [2]. What makes a repository good for reuse downstream? Three features recur in the literature. First, structured metadata that is machine-readable, so datasets surface in searches and can be integrated across platforms [4]. Second, standards that enforce consistent structure, such as BIDS for brain imaging [6] or ontologies that give standardized, machine-readable definitions, as the Vaccine Ontology does for the more than 670 manually annotated cancer vaccines in the CanVaxKB database [16]. Third, annotation of known limitations, so reusers can interpret data appropriately, as the PHA4GE quality control tags do for public health sequence data [5]. Ontologies and vocabularies matter for reuse of analysis methods too. The Neuroelectrophysiology Analysis Ontology (NEAO) standardizes descriptions of analysis processes so that researchers can find, understand, and reuse analysis results, resolving the ambiguity that arises when different software tools use different names for the same algorithm [17]. ## Governance, Ethics, and the Human Side Both sharing and reuse run on trust, and that trust has several components. Data producers worry about losing credit. Investigators often consider their data personal property, fear that reusers will publish from the data without acknowledging the original team, and worry that data will be used for purposes beyond those for which it was collected. Providing data can also require extra effort to reformat. These concerns are long-standing and documented across fields. Patients and participants have their own views. In a survey of 150 adults with intellectual disability enrolled in precision medicine research, views on the research itself were highly positive, but only 59% were comfortable with data sharing. Support ran high (79%) for reuses aimed at improving health, quality of life, and understanding causes of disability, and lower (64%) for reuses aimed at prevention. Many participants worried specifically about reuses that could harm people like themselves. The authors argue that addressing these concerns requires partnership with the affected communities, not just consent paperwork [18]. Healthcare providers who supply data to research platforms carry their own risk calculations. A [qualitative interview](/blog/guides/conducting-qualitative-interviews-a-guide-for-life-science-researchers) study with 21 German stakeholders in secondary use of health data found that providers see clear potential in reuse for quality of care research and healthcare planning, but their perceived risks and resulting reluctance to share, along with the absence of risk-management strategies, could significantly impede research with health data [19]. Finally, the benefits of sharing and reuse are not distributed evenly. A systematic review of 1,084 biomedical research articles from 2018 to 2022 found that 92.2% of anonymized datasets originated from high-income countries and were predominantly used by researchers in that same group, with cross-border flows running almost exclusively toward high-income countries. The authors argue that shared standards for anonymization and data sharing should involve nations from all income groups, because otherwise reuse concentrates value where data infrastructure already exists [20]. For a student planning a career, this is a live issue: the ethics of reuse includes asking whose data you are using and whether the source community benefits. ## Practical Steps for Students and Researchers If you are producing data, build the sharing plan before you collect anything: 1. Write a data management plan that names the repository, the metadata standard, and the access level for your project [1]. 2. Choose a repository early. Prefer a discipline-specific archive that uses community standards [6], [7], or a generalist or institutional repository if no archive fits [2]. 3. Document as you go, not at the end. Record instruments, units, versions, and known quality problems, [5]. 4. Check consent and legal terms for human data. Confirm what participants agreed to and what the law requires before any release. If you are reusing data, do the reverse: 1. Search repositories and structured collections first, before planning new data collection, because existing data may already answer or sharpen your question [2], [14]. 2. Read the metadata and quality annotations before trusting a dataset. Known issues, if well documented, can still make a dataset useful for your purpose [5]. 3. Cite the dataset formally. Researchers across disciplines have identified formal data citations as a key improvement needed to make reuse rewarding for sharers [2]. 4. Expect harmonization work. Combining datasets from different sources is the most common reuse pattern, and it takes real time, which the ecological break-even analysis quantified [7], [2]. The evidence also suggests a career strategy. Because prior reuse experience raises perceived benefits and lowers perceived risk and effort of sharing [3], getting hands-on experience reusing open datasets early in your training builds both a skill set and the disposition to share your own data later. ## Common Mistakes and Limitations Students and new researchers make a predictable set of errors around sharing and reuse. **Treating sharing as an afterthought.** Data thrown onto a personal website at publication time, without metadata, is technically shared but effectively unusable. Surveys show personal websites remain common even as repository use grows, and researchers report struggling to find datasets to reuse [2]. **Confusing "available" with "reusable."** A dataset can be downloadable and still impossible to interpret without documentation describing what each column means, how subjects were recruited, and what the known quality issues are [1], [5]. **Assuming reuse is free.** Searching for suitable datasets takes time, and researchers in most disciplines report difficulty finding datasets to reuse [2]. Integration and harmonization carry measurable time costs [7]. Budget for them in project planning. **Assuming consent transfers automatically.** Data collected under one consent framework may not be legally reusable for an unrelated purpose, particularly for health and genomic data under the GDPR, where consent requirements for secondary use have produced documented confusion [8]. **Overlooking who the data came from.** Participants may support some reuses and not others, and provider organizations weigh their own risks. Reuse decisions are not purely technical, [19]. **Ignoring global inequities.** Most shared anonymized health data flows from and within high-income countries [20]. Reusing data from lower-income regions carries an obligation to think about benefit sharing and fair collaboration. **Assuming reproduction is trivial.** Even with shared data and well-documented trials, multi-team reproduction efforts encounter missing supplementary materials and minor reporting issues, and teams using different software or statistical models can obtain slightly different results [10]. A final limitation applies to this article itself. The evidence base comes from surveys, case studies, and modeling in specific fields, mostly biology, ecology, and health research. Exact time costs, reuse thresholds, and attitudes vary by discipline and by dataset, so treat the numbers here as field-specific measurements rather than universal constants. ## Frequently Asked Questions ### What is the difference between data sharing and data reuse? Data sharing is making your research data available to others, typically through a repository with documentation and a license. Data reuse is using existing data, usually collected by someone else, for a new analysis, verification, or combined investigation. Sharing is the producer's job and reuse is the consumer's job, and the two form a cycle. ### Can data be shared without being reused? Yes. A dataset can be deposited, findable, and never downloaded for a new analysis, especially if documentation is poor or the data is hard to interpret. Sharing is a necessary condition for open reuse, not a guarantee of it. ### What does FAIR mean? FAIR stands for Findability, Accessibility, Interoperability, and Reusability. It is a set of principles describing data that is easy to discover, obtain under clear conditions, combine across systems, and use again, and it has become a cornerstone framework for research data sharing [1]. ### What is a data repository? A data repository is a curated, searchable archive that stores datasets, assigns them [persistent identifiers](/knowledge/bioinformatics/persistent-identifiers-in-bioinformatics-selection-criteria-and-implementation-workflow), and makes them citable. Options include generalist repositories, discipline-specific archives such as OpenNeuro for neuroimaging or Dryad and GBIF for ecology, and institutional or journal-supported repositories [2], [6], [7]. ### Does sharing data count as a publication? Depositing a dataset in a repository makes it citable, and formal data citation is one of the improvements researchers have recommended to give sharers credit [2]. Practices vary by field and funder on how dataset authorship is credited relative to journal articles. ### Why do researchers hesitate to share data? Common reasons include fear that others will take credit for their work, reluctance to spend time reformatting data, concern about misuse or uses beyond the original purpose, and unresolved questions about data ownership [11]. Legal and consent constraints, especially for human data, add further barriers [8]. ### What is secondary use of data? Secondary use means using data for a purpose other than the one it was originally collected for, such as reusing clinical records collected for treatment to study quality of care. The term appears throughout health data policy and is central to debates about consent, governance, and risk, [19]. ### How does reusing data benefit a research career? Reusing open datasets builds skills in data wrangling, meta-analysis, and critical appraisal while producing publications at lower cost than new data collection. Evidence from biological scientists shows that reuse experience also increases the perceived benefits of sharing and lowers the perceived risk and effort, shaping more open research behavior overall [3], [14]. ## Sources 1. [Waste Not, Want Not: How to Make Your Data Futureproof Through Good Data Sharing Practices.](https://pubmed.ncbi.nlm.nih.gov/41427491/) 2. [Data sharing and reuse practices: disciplinary differences and improvements needed](https://doi.org/10.1108/oir-08-2021-0423) 3. [The role of data-reuse experience in biological scientists' data sharing: an empirical analysis](https://doi.org/10.1108/el-06-2019-0146) 4. [Making Metadata Machine-Readable as the First Step to Providing Findable, Accessible, Interoperable, and Reusable Population Health Data: Framework Development and Implementation Study.](https://pubmed.ncbi.nlm.nih.gov/39088253/) 5. [PHA4GE quality control contextual data tags: standardized annotations for sharing public health sequence datasets with known quality issues to facilitate testing and training.](https://pubmed.ncbi.nlm.nih.gov/38860884/) 6. [The OpenNeuro resource for sharing of neuroscience data.](https://pubmed.ncbi.nlm.nih.gov/34658334/) 7. [Is Data Sharing Time-Efficient in Ecology? An Empirical Test and Extension of the Break-Even Reuse Model](https://doi.org/10.71448/bcds2454-1) 8. [Data altruism and the "consent" question: a study into the "consent" models used under the GDPR and how the data altruism mechanism can act as a potential solution for the research community in the reuse of health data.](https://pubmed.ncbi.nlm.nih.gov/40070923/) 9. [Clinical trial data transparency should match data access to scientific purpose: a tiered framework for reproducibility and reuse.](https://pubmed.ncbi.nlm.nih.gov/42928221/) 10. [Supporting reanalysis and reuse of clinical trial data: a case study.](https://pubmed.ncbi.nlm.nih.gov/41882735/) 11. [Data sharing for pharmacokinetic studies.](https://pubmed.ncbi.nlm.nih.gov/19558615/) 12. [Open innovation: Towards sharing of data, models and workflows.](https://pubmed.ncbi.nlm.nih.gov/28684136/) 13. [Social inequalities and long-term health impact of COVID-19 in Belgium: protocol of the HELICON population data linkage.](https://pubmed.ncbi.nlm.nih.gov/37202131/) 14. [Attentional control data collection: A resource for efficient data reuse.](https://pubmed.ncbi.nlm.nih.gov/40555897/) 15. [Wikidata for botanists: benefits of collaborating and sharing Linked Open Data.](https://pubmed.ncbi.nlm.nih.gov/40481658/) 16. [Empowering standardization of cancer vaccines through ontology: enhanced modeling and data analysis.](https://pubmed.ncbi.nlm.nih.gov/38890666/) 17. [Improving data sharing and knowledge transfer via the Neuroelectrophysiology Analysis Ontology (NEAO).](https://pubmed.ncbi.nlm.nih.gov/40442093/) 18. ["It is important to me to know where the information is going": Views of adults with intellectual disability on data sharing in precision medicine research.](https://pubmed.ncbi.nlm.nih.gov/42211778/) 19. [Benefits and risks of health data reuse for healthcare providers: stakeholder perspectives from a qualitative interview study.](https://pubmed.ncbi.nlm.nih.gov/40102872/) 20. [Unequal Flows? Economic Dimensions of Sharing Anonymized Data in Biomedical Research.](https://pubmed.ncbi.nlm.nih.gov/42175283/)