june13_b.indd C&RL News June 2013 284 In his classic work, Weeding Library Col-lections, Stanley Slote identifies a number of obstacles to weeding, including the time it takes, public displeasure, emotional or intellectual barriers that librarians face when it comes to removing books from their col- lections, the conflicting and subjective criteria that librarians often rely upon, and the ex- pense and effort it can take to compile more objective criteria.1 More than a decade after Slote’s work was last published, Rick Lugg and Ruth Fischer point out that many tools and data exist today to make weeding easier, less risky for librar- ies, and more accurate.2 They argue for the use of a data-driven, rules-based approach to weeding that removes the subjective factors included in weeding.3 In this approach, local data such as holdings, circulation history, and acquisi- tion dates are combined with external data such as other libraries’ holdings, core lists like Resources for College Libraries (RCL), and content in online archives like Ha- thiTrust. Libraries then establish criteria or rules based on these data to identify titles that might be withdrawn. Background Grand Valley State University (GVSU) Librar- ies consist of a main library, three branch libraries, and an off-site storage facility. In 2013 the university will be opening a new library. At that time the library’s off-site stor- age facility will be closed and all materials at this location will be moved into the new library’s automated storage and retrieval system (ASRS). The university has an existing ASRS at one of its branch libraries and has documented the issues with weeding out of such a system.4 Looking to avoid the difficulties associated with weeding out of an ASRS, and the time and expense of moving low-use monographs that may no longer meet the needs of the uni- versity, the library knew it needed to engage in a weeding project. A traditional title-by-title weeding project of the storage facility in 2007 had been a very time-consuming project that had resulted in approximately 8,000 volumes being withdrawn. The library’s desire was to develop a plan that in the short-term would allow for a second, more efficient weeding of the offsite storage facility that would produce a greater yield than the previous weeding project. In addition, the library wanted a long-term plan that would make weeding part of liaison librarians’ workflow in a way that would not add significantly to their workload. In 2009 GVSU began working with Lugg and Fischer’s company, Sustainable Collections Services (SCS), to pilot and implement a data-driven deselection project. This article outlines that project and the process that GVSU and SCS went through to Doug Way and Julie Garrison Developing and implementing a disapproval plan One university library’s experience Doug Way is head of collec tions and scholarly communications, e-mail: wayd@gvsu.edu, Julie Garrison is associate dean of research and instructional services at Grand Valley State University Libraries, e-mail: garrisoj@gvsu.edu © 2013 Doug Way and Julie Garrison June 2013 285 C&RL News implement what Lugg and Fischer described in one article as “The Disapproval Plan.”5 First steps After GVSU and SCS agreed to collaborate on this project, one of the first steps was to communicate the concept to the liaison librar- ians and to gather their input. The liaisons embraced the idea of rules-based deselection and identified a list of criteria they used in past weeding projects. These included past use and circulation, the book’s age, reviews it had received, whether it was on standard lists, the number of other libraries that owned the item, the author, the publisher or series, the type of book, and citations to the book. The library discussed with SCS what data were actually possible to use for analysis, and, in the end, SCS ran the library’s holdings data and circulation data against HathiTrust, RCL, WorldCat, and CHOICE: Current Reviews for Academic Libraries. Reviewing the data After the analysis was performed by SCS, the libraries received a summary of its collection, breakdowns by both liaison area and Library of Congress Classification, and detailed lists of potential candidates for withdrawal based upon established criteria. Initial withdrawal candidates were books published prior to 2000, held by more than 100 libraries in the United States, not currently in RCL, never reviewed by Choice, and having no circula- tions since 1998. The withdrawal candidate criteria had been agreed upon by the liaison librarians, and were dictated by a number of factors. For example, because this project was focused on the library’s storage facility, which had no books published after 2000, that date was chosen by default. Likewise, the library only had circulation data dating back to 1998, so this was the earliest year that could be used as a cut-off. The liaisons then had the opportunity to revise their lists from the original criteria in any way that the data allowed. In most instances liaisons revised the criteria for holdings and circulation data. For example, one liaison decided she was not comfortable discarding any books held by fewer than ten libraries in Michigan. While another liaison decided that she would be willing to discard books that had circulated three or fewer times since 1998, as long as it had not circulated since 2008. It had been assumed that the liaisons would be more conservative than the original criteria, decreasing the number of withdrawal candidates for consideration. In reality, enough liaisons, especially those in the health sciences, sciences, and technology disciplines, were willing to be more aggres- sive with their criteria and the total number of potential withdrawal candidates increased by nearly 9,000 items. The review of candidates Three possible methods for reviewing the withdrawal candidates were identified by liaisons: 1) reviewing materials from the detailed list of candidates to determine which books should be retained, 2) staging materials by pulling withdrawal candidates off the shelves so liaisons could quickly re- view the candidates with the book in hand, and 3) flagging the withdrawal candidates in place in the stacks so liaison librarians could see them in context with materials in the collection. We felt each method had its advantages and would work best for certain subject areas. Because the library saw this project as a pilot that would hopefully lead to the imple- mentation of an ongoing systematic weeding program, we used all three methods. Liaison librarians decided on the method that worked best for their subject areas and their comfort with the process. Traditionally during weeding projects librarians decide if there is a reason to with- draw a book. With this project, the assump- tion was that if a book was a withdrawal candidate then it should be withdrawn un- less there was a reason to keep the book. The library had used this method in the past on smaller weeding projects and found it increased the yield and seemed to reduce librarian anxiety. C&RL News June 2013 286 As the liaisons reviewed the books on their withdrawal candidate lists, they were required to provide a rationale for every book that was retained. This compelled the liaison to consider the precise reason why he or she was retaining a book, especially in light of all the factors weighing against its retention. It also made retaining a book a little more onerous than simply allowing it to be with- drawn and gave the library data that could be used now and in the future. The results Through the process developed with SCS, 87,751 titles located in the storage facility were analyzed. Of these titles, 38,662 titles were identified as withdrawal candidates and 33,353 (just over 86% of the candidates) were ultimately withdrawn from the collection. Librarians elected to save more titles in the humanities than in any other area. However, this was also the largest area targeted for withdrawal. Of the almost 9,000 withdrawal candidates identified in the areas of science, technology, engineering, and medicine, all but 4% were withdrawn from the collection. Ten percent of social sciences withdrawal candidate were retained (see Table 1 for details). Staff and librarians completed this project over the period of a few summer months. Between June and August 2011, a few student employees working with one full-time staff member staged and flagged books for those librarians who requested a physical review of materials. Approximately 19,000 humanities titles were flagged and 10,000 social sciences titles were staged, taking between 160 and 240 work hours to complete. During this time, students also performed other work in the storage facility, including the removal of the nearly 9,000 titles in ar- eas, such as science and engineering, where librarians did not require an on site review of materials before making retention decisions. Once materials were flagged or staged for review, librarians received notification to visit the storage facility. Titles flagged within the overall collection took longer to review than titles that were staged separately. Li- brarians took approximately one week to review 10,000 flagged titles, whereas it took only a few hours for librarians to review staged titles. Conclusion The library considered its data-driven dese- lection pilot project to be a huge success. Librarians were astonished by how quickly they were able to review the selected can- didates, how well the criteria identified candidates, and how many of those can- didates they ended up withdrawing. The next year the library conducted a smaller weeding project in its main library with equal success. Moving forward, the library plans on implementing a “disapproval plan” that will generate smaller lists of books for librarians to review each year based on the criteria they identified. This will allow weeding to be integrated into their workload and allow the library to avoid large-scale weeding projects in the future. Subject Area All Filtered Items Withdrawal Candidates Saved Withdrawn Humanities 53,020 24,391 4,412 19,979 (82%) Social Sciences 13,500 5,388 523 4,865 (90.3%) STEM 16,638 7,644 194 7,450 (97.5%) Medicine 4,593 1,239 180 1,059 (85.5%) Totals 87,751 38,662 5,309 33,353 (86%) Table 1. Weeding statistics by subject area. June 2013 287 C&RL News Since this initial project two things have impacted the criteria that is being used in subsequent weeding projects. Bowker will no longer license RCL to SCS, so we no longer use that in our criteria. More importantly, GVSU is partnering with other libraries in the state on a shared print management project. GVSU worked with six other universities to establish the Michigan Shared Print Initiative (MI-SPI), a distributed shared print mono- graph project for the management of legacy monographs.6 Two libraries in MI-SPI com- mit to retain a share of widely held, low-use monographs in their collection, allowing other libraries in the group to withdraw those items. The MI-SPI project has identified more than a half-million monographs that can be safely discarded, while still maintaining access in the state. Weeding may never be a popular task, but through the use of data and collaborative partnerships, libraries can eliminate the bur- den, anxiety, and fear this activity has created in the past. Notes 1. Stanley J. Slote, Weeding Library Collec- tions: Library Weeding Methods (Englewood: Libraries Unlimited, 1997). 2. Rick Lugg and Ruth Fischer, “Future Tense—Weeding: The Time is Now,” Against the Grain 20, no. 4 (2008): 87–88. 3. Rick Lugg and Ruth Fischer, “Future Tense—The Disapproval Plan: Rules-based Weeding & Storage Decisions,” Against the Grain 20, no. 6 (2009): 74–76. Rick Lugg, “Data-driven Deselection for Monographs: A Rules-based approach to Weeding, Storage, and Shared Print Decisions,” Insights 25, no. 2 (2012): 198–204. 4. Patricia Bravender and Valeria Long, “Weeding an Outdated Collection in an Auto- mated Retrieval System,” Collection Manage- ment 36, no. 4 (2011): 237-45. 5. Lugg and Fischer, “The Disapproval Plan.” 6. Visit www.mcls.org/mi-spi for more on the Michigan Shared Print Initiative. http://muse.jhu.eduThe Trusted Voice In The Scholarly Community. Project MUSE is an indispensable resource for research and discovery for scholars, students, and faculty, providing 24/7 digital access to current, full-text scholar- ship in the humanities and social sciences. With current scholarship and rich archival content packaged into affordable and flexible collections, Project MUSE is an essential scholarly resource. Get to the core of digital research. Academic To The Core. PRO3473 Launch_Academic_CRL_Layout 1 10/10/12 9:55 AM Page 1