Data Collection Guidelines: Overview and examples of application

June 5th, 2025

In order to maintain relevant and viable collections for their patrons, academic libraries have long created and maintained collection development policies and guidelines (Tokarz, 2024). These policies and guidelines grow and change with the research and technology landscapes surrounding libraries. This means that policies and guidelines have changed to incorporate non-traditional resources such as software and datasets (Foster, et al., 2019). At the University of Tennessee, Knoxville (UTK) Libraries, there has been an increase in requests for datasets as collections. In response to this, librarians developed a set of collections guidelines to help navigate this different type of resource acquisition while continuing to foster a structured, transparent, and accessible collection for UTK patrons. This editorial will outline the current status of UTK Libraries’ dataset collection guidelines, give examples of how they have been applied, and discuss next steps when considering datasets as collections.

Why dataset collection guidelines and what are UTK Libraries’ guidelines?

If datasets are not the only non-traditional resources academic libraries acquire, then why are guidelines specific to datasets necessary? These guidelines can tackle various needs, such as reducing wait times for data access, addressing funding challenges, preventing duplicate purchases, and improving overall data access across campus. UTK Libraries have found that guidelines specific to acquiring, storing, and accessing data can ensure librarians and patrons are able to overcome the challenges of treating datasets as collections such as funding, storage, equitable access, and licensing for perpetual use and access. Having these specific guidelines has helped UTK Libraries provide the funding for  dataset purchases (verses the users having to pay for data access), find and/or create long term storage plans for acquired datasets, ensure acquired datasets are accessible and available beyond the requestor, and allow acquired datasets to be licensed for usage beyond the original request.

Currently, before a dataset is considered for purchase at UTK, the potential value of the dataset is evaluated to ensure it is a viable asset for the entire community. If a dataset could benefit more than just the requesting individual or their project, information relating to the contents of the dataset, such as methods, formatting, metadata, and technical documentation, is requested from the vendor by the collections and acquisitions teams.  When necessary, the Libraries collaborates with key stakeholders (such as the IT arm of the University) to host the datasets, or they use the Libraries’ available digital storage when possible. Overall, the current dataset collection guidelines fall into these four general categories:

• Purpose: Ensure datasets benefit multiple users/projects and uphold Open Science principles. As the Libraries’ serve the whole campus, any purchase must benefit more than one patron or one group and can be used widely across campus.
• Documentation: Vendors must provide detailed technical documentation about the dataset’s source, acquisition, curation, format, metadata, and added value. By requiring detailed documentation, the Libraries’ can better plan for storage and access needs for the dataset moving forward, and can know if the dataset is in a viable format before purchase.
• Hosting: UTK Libraries may host datasets for researcher access, contingent on digital team consultation (inside and outside the Libraries) and available space. Digital storage space (in general) is not unlimited and the size and whether the Libraries’ are able to host it (or work with other digital teams on campus to use it) can be a deciding factor if the Libraries’ will acquire a dataset.
• Transparency: Clear electronic resource use policy will be available on the website. The Libraries need clear licensing language to understand how the data can be used and reused and protect the Libraries from liability as the dataset is used and reused by patrons.

These four areas are still developing and are not comprehensive, but they help UTK Libraries navigate the challenges of funding, storage, access, and licensing datasets as collections. The goal of these guidelines is not to limit the number of datasets that UTK Libraries purchases, but rather to ensure that the collection continues to remain valuable as it grows and evolves with the research and technology needs of campus.

Below, Figure 1 provides three examples where the application of these guidelines can be seen in action.

Dataset A	Dataset B	Dataset C
Patron requests dataset from a large, well-established data archive	Patron requests a large, un-hosted dataset from known data producer	Patron requests one-time dataset download from known data producer
Factors from dataset collection guidelines:	Factors from dataset collection guidelines:	Factors from dataset collection guidelines:
Data would benefit multiple users/projects Data archive provides datasets across multiple domains and topics	Data would benefit multiple users/projects Dataset could be used by multiple individuals and across domains and topics.	Data would NOT benefit multiple users/projects Dataset is very narrow in topic and unlikely to be used by other individuals or across domains and topics.
Data would not need to be stored by UT Libraries Data archive provides own hosting and platform for discoverability	Data would need to be hosted and made discoverable by UT Libraries Dataset is in a known file format and size does not exceed 1 TB, data could be locally stored and given access by link on the library website.	Data could not be hosted and made discoverable by UT Libraries Data is either in a unique file format and/or vendor does not allow sharing beyond individual
Cost and licensing of data use is clear Data archive requires yearly subscription fee and licensing agreement for access to all data within archive	Cost and licensing of data use is clear, vendor is reliable Data producer is known and provides amenable licensing terms for reuse, and cost is one-time.	Cost and licensing of data use is unclear It is unclear how the data can be reused, and cost seems to be by individual and not one-time
Decision:  Data was acquired	Decision: Data was acquired	Decision: Data was NOT acquired

Next steps and considerations for treating datasets as collections

UTK Libraries will continue to acquire datasets as long as there is a continued need, but from the above examples and planning for the future, these are the next steps and considerations when treating datasets as collections.

Usage Statistics and Storage

Already the impact of dataset reuse is difficult to track (Hemphill, et al., 2024). Therefore it is important to find ways to meaningfully track how purchased datasets are being used, especially if there are continued licensing and storage costs. This can be done in house (if the data is hosted by the Libraries) or by working with the data vendor (if the data is hosted by the vendor). Ongoing assessment of the continued value of purchased datasets is just as necessary as it is for more traditional digital collection items like electronic books and journals. If ongoing storage and cost compared to usage is an issue, access to datasets may need to be discontinued.

Working with non-traditional vendors

As datasets are a newer and non-traditional collections item, the vendors of the datasets may be new to working with academic libraries as purchasing customers. UTK libraries have found that having dataset collection guidelines helps when working with these new vendors and can shorten the wait for access for the patrons. These guidelines serve as a helpful communication tool with vendors and the patrons so that both parties can understand why an acquisition of a dataset is taking a long time or why an acquisition was successful or not. Additionally, they help vendors new to working with academic libraries understand important priorities libraries as customers may have when it comes to critical elements of collections such as accessibility and perpetual licensing and availability of acquired datasets. 

Licensing and liability for data reuse

Since datasets can be ingested, reworked, and made into innumerable derivatives, it is important that dataset licensing protects the library and the patron from liability for the reuse of datasets. UTK’s guidelines cover the need for transparency when it comes to access and reuse. Especially with the increase of things like machine learning in research, it is important patrons are able to use purchased data as freely as possible in order to maintain that transparency and freedom.

Why guidelines and not policies?

These dataset purchasing guidelines are currently named guidelines, rather than strict policies, to allow UTK Libraries to grow and evolve as dataset requests increase and evolve. Flexibility is paramount when it comes to being able to meet unique needs of patrons, but as datasets as collections continue, these may develop into more stringent policies especially around issues of storage and licensing. But for now, these guidelines ensure that UTK Libraries can meet patrons’ needs while working towards supporting the greater cause of transparency and access of Open Science. 

References

Foster, A. K., Rinehart, A. K., & Springs, G. R. (2019). Piloting the Purchase of Research Data Sets as Collections: Navigating the Unknowns. Portal: Libraries & the Academy, 19(2), 315–328. https://doi.org/10.1353/pla.2019.0018

Hemphill, L., Thomer, A., Lafia, S., Fan, L., Bleckley, D., & Moss, E. (2024). A dataset for measuring the impact of research data and their curation. Scientific Data, 11, 442. https://doi.org/10.1038/s41597-024-03303-2

Tokarz, R. E. (2024). Exploring Collection Development Policies at R1 Research University Libraries. Collection Management, 49(1–2), 46–66. https://doi.org/10.1080/01462679.2024.2323428