In February 2010, PUL convened a Library Discovery Assessment Group to assess the landscape of these new discovery systems and recommend whether the library should adopt one. Ultimately, the group recommended adopting Primo, primarily because it felt that Primo offered the best option for updating our aging catalog interface. Since PUL uses Ex Libris’s Voyager ILS, Primo held out the promise of a smoother transition and better integration of circulation functions and other features. However, there was some hesitation. Primo Central had only just been launched in beta testing in January 2010, and there were no live installations that the group could examine.² Also, there were concerns about the amount of content in Primo’s unified index, especially compared to Summon, which at the time had broader article coverage and boasted a significant body of news content. In the end, PUL opted for a hybrid approach: using Primo (without its centralized index) as the new discovery interface to Princeton’s local content—cataloging data and other repositories, like senior theses and finding aids—while also subscribing to Summon to leverage its unified index of article and news content.

In fall 2010, PUL created the LDIG to implement this hybrid solution. Over the past three years, LDIG’s work has gone through three main phases.

The first phase of implementation involved becoming familiar with and configuring the new discovery systems. This primarily meant Primo, as there were numerous configuration choices necessary to prepare and normalize the Library’s local data for ingestion. PUL contracted with Ex Libris to integrate the Summon content into the Primo interface via Summon’s API, so Summon required little set up beyond turning on our full text packages in the ProQuest knowledgebase.³ The new system was hosted by Ex Libris on a custom domain, and the Library branded this new system SearchIt@PUL.

LDIG initially planned for a public beta launch in early 2011, but that timetable proved too ambitious. LDIG introduced the new system to staff in April 2011, while the public beta launch was pushed back to the beginning of the fall 2011 semester. Various issues contributed to the delay. For one, it took time to learn how PUL’s MARC records mapped to Primo’s new PNX records, and figuring out which fields were being displayed and indexed for search often required investigation and experimentation. We also ran into significant challenges trying to tailor algorithms for search, deduplication of records, and FRBRization in Primo to work well with our large and diverse collections. For example, there were numerous issues related to the indexing and retrieval of records in foreign characters, especially Chinese, Japanese, and Korean, which was of particular concern because PUL has a highly regarded East Asian library.

Working through these and other issues was time-intensive for LDIG members. During this phase, LDIG often worked with Ex Libris’s support staff, who were very responsive in answering questions and assisting the group’s work. Ultimately, though, Ex Libris was unable to solve some of these problems. Perhaps the most critical of these was the integration of Summon into Primo. After many attempts, LDIG decided that trying to incorporate results from Summon into the Primo interface was counterproductive. Results didn’t blend, the different levels of detail were confusing, and filtering options for the two systems were too dissimilar. While LDIG was still committed to using both systems for discovery, the group decided to take a different approach in preparation for launching the system to users.

LDIG’s new approach was to give up the idea of a single search box and interface for catalog and article discovery and instead present users with a tabbed search box on PUL’s homepage for searching Primo and Summon separately (see figure 1). This search box had three tabs: Catalog+, Articles+, and Course Reserves.⁴ We placed the Catalog+ tab first because we assumed most users would search in the default tab, and our local catalog results would be the data users were expecting to see. When users searched from the Catalog+ tab, they were taken to results in our hosted Primo interface (see figure 2); when they searched on the Articles+ tab, they were taken to results in our hosted Summon interface (see figure 3). LDIG added the new tabbed search box to our library’s homepage in fall 2011.

As users put the new system through its paces, LDIG discovered some problems with our implementation. We soon realized that the Primo interface offered limited flexibility for adding or extending features to improve our users’ experience. Library systems staff began coding workarounds to deal with some of these issues, including better display of records, paging of results, and a more robust item request form.⁵ We also experienced problems in the connection between Primo and Voyager, such as slow response times, the lack of current holdings information for serials, and the inability to handle serials with very large holdings records (e.g., Science or Nature). In Summon, staff and users expressed confusion over the various options for filtering results, and they had difficulty figuring out what content was indexed in the system. Additionally, because we use Ex Libris’s SFX as our e-resources knowledgebase, we had difficulty ensuring that the right content and journal packages were turned on in the ProQuest knowledgebase, and we had to alter our existing workflow to ensure new content was turned on in both places.

By fall of 2012, LDIG’s active implementation work had slowed. A major cause of this slowdown was a new project to redesign our Library’s website, which involved our programmers as well as me and some other LDIG members. Building off LDIG’s experience with our hybrid discovery service, the central feature of the new website was to be a tabbed search box on the homepage integrating both Primo and Summon (see figure 4). In contrast to our past efforts, though, the new website would reinstate an All Search across multiple content sources—Primo, Summon, our website, LibGuides, our finding aids, and more—and to integrate those disparate search results into a single results screen. The results would be displayed in separate silos on the screen in a “bento box” format (see figure 5).⁶ From these brief lists of results, users would be able to click through for full results to the native interfaces of the separate systems (Primo, Summon, LibGuides, Finding Aids, etc.).

The Website Steering Group arranged for multiple rounds of usability testing of the new website, something LDIG hadn’t done in its earlier implementation work. In the initial tests, most users praised the new All Search for its ease of use and also for the “bento box” approach of grouping results by category. They felt that this clarified the results they were seeing and made it easier for them to pursue different avenues of inquiry. Additional testing after the site had been live for several months exposed a divide between less-experienced and more-experienced library users. Lower undergraduate users gravitated to the All Search and were satisfied with their results, while advanced users, like graduate students and faculty, preferred searching separate databases and the old Main Catalog because they provided advanced functionality or required fewer clicks to get to their desired information. Although some of this reaction reflects an aversion to changing habitual workflows, we did see that some tasks are harder to do in the new systems. Finding the right balance between the needs of both less- and more-advanced users has bedeviled LDIG’s implementation work from the beginning.

The new website, which went live in summer 2013, has been a positive step forward in PUL’s efforts at improving resource discovery. However, it hasn’t addressed all of the underlying concerns about either Primo or Summon, its two major components. In the near term, PUL is likely to continue its hybrid approach. But given the rapidly changing discovery services landscape, PUL is keeping its options open and might shift its approach as new developments arise.

Looking back over the past three and a half years, I see several lessons for libraries considering or beginning to implement a discovery service. The first is simply a recognition that “implementation” may be somewhat of a misnomer—our “implementation” is still ongoing after more than three years with no end in sight. Whether it’s one or a combination of services, implementing a discovery service is an iterative process necessitating ongoing testing, tweaking, evaluation, and communication with staff and users. Like other aspects of a library’s web presence, discovery services require a continuing commitment of time and oversight to ensure they’re meeting the needs of users.

Based on my experience with LDIG, I believe that libraries implementing a discovery service should consider hiring a discovery services or user experience librarian to oversee the implementation. Having a librarian with primary responsibility to oversee discovery services or focus on the user’s experience would have helped LDIG in many ways. In the first phase, a discovery services librarian could have helped our group to articulate and clarify our project goals and keep them prominent in our early discussions about design and configuration decisions. A discovery services librarian also could have helped to keep us on schedule in the first few months by helping to distribute the workload and follow up to ensure tasks were completed in a timely fashion. During the second phase, having one librarian in charge of discovery services would have made it easier to solicit and respond to feedback as well as to communicate LDIG’s progress and any known issues. He or she would also have been invaluable in managing LDIG’s growing list of bugs, support requests, and future enhancements by helping to prioritize them, communicating with our vendors about them, and ensuring that the group revisited them as necessary. The discovery services librarian could also have helped LDIG to conduct usability studies to inform our continuing work. Finally, in our most recent phase, a discovery services or user experience librarian would have been helpful in working with the website redesign group to integrate our various discovery services as the design for the new website unfolded.

While Princeton University Library’s particular solution to resource discovery—integrating Primo and Summon—may not be helpful to most libraries, all libraries can learn from LDIG’s experience in implementing this solution. Considering questions of planning, flexibility, communication, and staff time can help libraries ensure that their investment in a discovery service returns the most value for both the library and its users.