Data Management: Between Pain Points and Opportunities
What’s stopping researchers from sharing high-quality data—and how can it change?
Researchers are increasingly encouraged to make high-quality datasets publicly available, yet persistent gaps in training, infrastructure, and systemic support continue to hinder the widespread adoption of the data practices needed to support these efforts.
A lost file, a corrupted hard drive, an experiment impossible to replicate due to unreliable lab notes: most, if not all, researchers will deal with the fallout of poor data management practices at some point in their career.
“These mistakes are common,” said Dr. Kristin Briney, biology and biological engineering librarian at the California Institute of Technology. Such mishaps can lead to stress, resource wastage, and situations which, “every once in a while, […] can be very bad.”
“Data has always been seen as incredibly valuable,” added Gareth Knight, research data manager at the London School of Hygiene and Tropical Medicine; yet for a long time, publications have remained the only required academic output. With datasets remaining unpublished within labs, good data management—if it happened at all—often relied on “a researcher sacrificing their personal time to do the extra work,” said Dr. Alexia Cardona, co-director at Reproducible Research Cambridge and course director at Professional and Continuing Education, University of Cambridge.
This status quo has now been challenged by funders, journals, and institutions increasingly demanding that data be made more openly available. “The recognition that [data] will be audited by external bodies, and [that] you will be expected to make these resources available in a form that can be understood creates that incentive to make sure the data is better organized, that it's easier to find, and it's documented in an appropriate form,” noted Knight.
On the ground, however, adoption of better data management practices remains patchy. “It's about culture change,” acknowledged Cardona. “And it's very hard, as you know, to change culture.”
A patchwork of sharing mandates
Over the past two decades, high-profile academic misconduct cases paired with a growing reproducibility crisis have cast in sharp relief the need to strengthen data integrity—the accuracy, consistency, and trustworthiness of data throughout its lifecycle. Open science, with its emphasis on transparency and the right to access publicly-funded research, has offered a compelling framework to improve the self-correction of the scientific process.
When, in 2014, the publisher PLoS first required researchers to make all data publicly available upon publication, “people melted down over it,” recalled Briney. “And now, it’s standard.”
On the funding side, by 2017, required submission of data management plans (DMPs) with each bid, outlining how data would be handled throughout a project, and how it would be made public. While exact implementation details still confusingly differ across and within funders, journals, or institutions, they broadly align with the same vision: data should be FAIR (Findable, Accessible, Interoperable, and Reusable).
In turn, such requirements rely on researchers adopting a range of good data management practices not just at the point of sharing, but throughout the research data cycle.
Figure 1: The Data Lifecycle. Funders, governments, and institutions are increasingly striving for data to be FAIR: Findable, thanks to persistent identifiers, metadata, and being indexed; Accessible, through clearly defined processes; Interoperable—that is, structured and obtained in such a way that it can be combined with other datasets; and Reusable, meaning that the appropriate license and documentation should be provided. This requires researchers to improve how they handle data from the planning stage (e.g., DMPs, awareness of regulations), through to the collection, processing, and analysis steps (e.g., version control systems, third-party code review, open software), and to the preservation, sharing and reuse stages (e.g., regular backups, selection of appropriate repositories, and licensing, providing thorough documentation and metadata). Credit: RDMkit by ELIXIR (CC BY 4.0).
In addition, “the expectations of what type of resources should be shared has increased over time,” noted Knight. Code quickly followed in the footsteps of data, for example, and software is on the horizon. That trend is strengthening, fuelled by the realization that data alone is not enough to guarantee that findings can be replicated. “Targeting really the reproducibility aspect,” explained Cardona, “is still a big task that we need to resolve.”
A range of mandates have underpinned these changes, including the 2023 NIH Data Management and Sharing Policy. Such directives sit, sometimes uneasily, alongside legal requirements that emphasize privacy rights, such as the UK GDPR and Data Protection Act, as well as ethical guidelines balancing open practices with the interests of historically exploited populations (e.g., CARE principles).
Building Reproducible qPCR Workflows From Experimental Design to Data Analysis
This guide walks through a practical six-step workflow covering experimental design, sample preparation, primer validation, reference gene selection, and data analysis to help generate reproducible qPCR data.View How To Guide
While external requirements have been necessary to drive change, on the ground, data sharing requirements have yielded mixed results. Compliance still lags behind, though funders are moving towards stricter enforcement—for example, by introducing machine-actionable DMPs that facilitate monitoring.
Perhaps more importantly, poor data management practices have also resulted in so-called “data dumping grounds”: publicly shared datasets that cannot be reused due to incorrect documentation, missing metadata, or licensing issues.
Overcoming knowledge and culture barriers
“The issue is training, but also reward and recognition,” said Cardona. Indeed, nearly 70% of researchers feel their data sharing efforts go unrecognized. Knight echoed this sentiment, adding that high-quality datasets should count towards hiring and promotion, and that data generators need opportunities for awards and professional visibility.
Still, lack of data management knowledge remains one of the most stubborn barriers: in a recent survey, 46% of biologists cited being unaware of how to organize data in a presentable and usable way as their main reason for not sharing it.
Training is still rarely included in formal education, and, unlike industry, very few labs have the capacity to develop structured data workflows. As a result, “a lot of graduate students will come in, […] and they're making up organisation systems as they go,” explained Briney. This leads to practices that are, at worst, faulty, and at best, disparate.
Alongside books, publications, and resources, many initiatives have emerged to address this need, spearheaded by organizations such as UKRN, ELIXIR, EOSC, or the Carpentries.
Advancing AAV Characterization Through Mass Photometry
This infographic shows how single-particle mass photometry delivers AAV purity analytics in less than five minutes while supporting development and GMP manufacturing workflows.View Infographic
Such courses, stressed Cardona, do not detract from the need to have “that go-to person” in the lab who provides advice tailored to students’ exact data needs. Yet, “a lot of the well-established researchers who are responsible for teaching a younger generation didn't necessarily learn data management skills,” said Briney, having been trained before the rise of digital data.
Communities of practice can help bridge this gap, including local data champion groups. “Find other people,” urged Cardona. “Together, you can have a bigger voice,” and keep on top of a rapidly evolving field. As for time-poor PIs, change can start with simply opening frank conversations with trainees: “'What do your lab notebooks look like? Can we talk about this as a group?' I don't see that discussion happening a lot,” Briney said.
Small changes, big rewards
Trying to improve practices can quickly feel overwhelming, so labs should not try to tackle everything all at once. “Pick something that you think you can do and make better, and [then] make it routine and make it practice,” Briney recommended.
This can mean simple things, such as ensuring each column represents a single variable and being consistent when building a dataset, explained Cardona. As for Briney, she “always come[s] back to file naming”: a clear, consistent convention applied from the start “can pay off hugely in the end” when locating files and checking for data gaps.
As for data analysis, habits that greatly improve reproducibility include getting code checked by a third party before it is shared, as well as the adoption of version control systems that allow collaborators to see and contribute to coding, and track input from all researchers.
Above all, documenting every decision in a living data management plan—from how files are organized to naming conventions, storage systems, backup expectations, and software used—serves not just collaborators and one’s future self, but “helps you make sure that you're following the practices that you set out,” explained Briney.
Finally, “talk to your library,” she stressed. “I really think that's the best place to start.” Research data stewards can be particularly helpful when navigating the minutiae of funder requirements, especially in health research, where various mandates may “have different aims—protecting participants on one hand and expecting resources to be shared from a different perspective,“ noted Knight.
Beyond individual responsibility
Despite their willingness, researchers keen to improve their practices often face infrastructure barriers: “In the sciences, the data repository landscape is not consistent yet,” explained Briney, and while certain data types have specialist repositories—e.g., GenBank for genetics—33% of researchers indicate not knowing where to deposit their files.
How Agentic AI Is Transforming the Modern Lab
This whitepaper examines how agentic AI is reshaping laboratory operations and what this shift means for the future of scientific discovery.View Whitepaper
Similarly, safely storing increasingly large and complex datasets continues to be a challenge, let alone, as Briney noted, “put[ting] a terabyte of data up in the cloud and expect[ing] people to download it easily.” Cloud providers are slowly filling that gap, though not without concerns as to their reliability and how they may use entrusted data.
Who should pay for the time and resources needed for data management also remains a sticking point. Researchers are still “very concerned about whether they'd make themselves look uncompetitive by including these extra costs” in their funding bids, said Knight. “There are still a large number of research funders that don't allow the costs, or are very vague.” Changes in that regard may be afoot, driven by organizations such as UKRI.
Finally, standardization offers one of the most promising systemic levers. “People focus specifically on the requirements of their project without necessarily taking into account the broader standards that exist in the wider community,” said Knight, citing established format guidelines like OMOP and DDI Lifecycle.
Providing researchers with established, well-designed servers or data collecting tools such as REDCap and Open Data Kit means “they don’t have to reinvent the wheel themselves.” Building such infrastructure “takes time, and it requires a combined effort,” acknowledged Cardona, but it allows datasets from disparate projects to later be brought together, and, crucially, reduces the burden placed on researchers.
Scientists have long been acutely aware that they are, ultimately, the ones to benefit most from good data management—from the productivity, smooth workings, and above all, peace of mind that it fosters.
“I think that perspective gets lost a little bit when you're talking about it like a check mark for a grant application,” said Briney. Still, “things are getting better,” Cardona noted. “I see changes in my students, even in the researchers themselves.” It is now for funders and institutions not just to demand this shift in practices, but to help provide the resources required for it to unfold.
Add Technology Networks as a preferred Google source to see more of our trusted coverage.
