allison.p65 56 College & Research Libraries January 2000 Database Selection: One Size Does Not Fit All DeeAnn Allison, Beth McNeil, and Signe Swanson With increasing costs and decreasing budgets, it is important for librar­ ians to make sound purchasing decisions for electronic resources. What factors are important to consider in making a decision? How can librar­ ians balance these factors, which may appear to be contradictory, and also meet the expectations of administrators, staff, and users? This ar­ ticle describes a strategy for making delivery decisions that address local conditions, pricing, feature options, hardware costs, and network availability. Finally, a model is presented for evaluating the decision based on dollar requirements and local issues. ibraries share the common goals of enhancing learning and ensuring access to infor­ mation. They also assist people in obtaining high-quality information that librarians have reviewed for accuracy. However, with the advent of electronic resources, these goals have become in­ creasingly difficult for libraries to fulfill. Libraries now must be able to successfully determine the most cost-effective way to deliver information, while competing with the Internet and other nonlibrary information providers. Over the past five years, electronic re­ sources have undergone significant tech­ nological changes. Graphical user inter­ faces and the evolution of the Internet have created new options and complexi­ ties, as well as technologically savvy us­ ers who demand improved access. Deter­ mining the optimal delivery method for these electronic resources can be difficult. Libraries must use cost-measurement techniques to evaluate alternatives or risk losing budget and staff support because libraries are judged as not meeting user needs. Local conditions, pricing, feature options, hardware costs, and network availability should be considered in the selection process. At the University of Nebraska-Lincoln (UNL) Libraries a project was undertaken to replace the popular, but technologically obsolete, DOS-based database system with newer technology based on World Wide Web access. Although the available databases did not change significantly, the switch to the new system resulted in heavier use. For example, one of the da­ tabases, PsycLIT, dramatically increased from an annual access of 23,000 connec­ tions to more than 74,000 in just three and a half months. This article discusses the DeeAnn Allison is a Professor and Chair of Automated Systems at the University of Nebraska-Lincoln; e- mail: deeanna@unllib.unl.edu. Beth McNeil is an Assistant Professor and Executive Assistant to the Dean of Libraries at the University of Nebraska-Lincoln; e-mail: bethm@unllib.unl.edu. Signe Swanson is an Assistant Professor and Reference Librarian at the University of Nebraska-Lincoln; e-mail: signes@unllib.unl.edu. 56 mailto:signes@unllib.unl.edu mailto:bethm@unllib.unl.edu mailto:deeanna@unllib.unl.edu Database Selection: One Size Does Not Fit All 57 necessary questions and criteria to con­ sider when selecting a delivery method for electronic resources, with examples from the UNL libraries’ experience. It also presents a model for assessing that deci­ sion, keeping in mind that no one answer fits all situations. Selection Criteria Database pricing is based on a variety of formulas determined by the vendor and/ or copyright holder. Many factors com­ plicate the pricing of electronic resources, the most basic being the delivery method. Vendors’ and libraries’ definitions of de­ livery methods vary. For instance, connec­ tions can be limited to specific buildings or geographic areas, and some databases may or may not be Internet accessible. It is important that library and vendor have a mutual understanding of these defini­ tions when entering into licensing/con­ tract negotiations. For the purpose of this article, the fol­ lowing delivery method definitions are used throughout. A non-networked com­ puter is one that is not connected to a net­ work. It has either CD-ROM or hard-drive accessible database(s), and only one per­ son may access any of the databases at a time. A local area network (LAN) is a sys­ tem that includes a local server capable of handling multi-users in a multi-task­ ing mode over a locally or remotely ac­ cessible network, which may include Internet access. Databases are loaded on the local server, and the library is respon­ sible for maintenance and network con­ nectivity. A remote Internet connection pro­ vides access to databases that are housed on a vendor’s remote server and deliv­ ered over the Internet. All of the database maintenance is carried out by the vendor, with the local institution responsible only for the connection to the vendor. In addition to delivery method, pric­ ing factors include the number of concur­ rent users/ports/passwords and other options. Pricing usually varies among dif­ ferent vendors providing the same data­ base. Some vendors offer subscription pricing for titles “bundled” together, such as OCLC’s FirstSearch database. One question to ask these vendors is whether the library can determine the titles in the bundled package. Some databases have differentiated pricing based on whether the library also subscribes to the print version of the title. Other resources may have additional charges for outputs, such as printing or downloading. Licensing requirements or consortial agreements also may affect the overall charges to an individual library for a particular re­ source. Outsourcing is a possibility, but it will be important to have good response time for maintenance and troubleshooting. The American Psychological Association’s (APA) electronic databases provide a good example to demonstrate the pricing situ­ ation. APA publishes Psychological Ab­ stracts, the print index for the field of psy­ chology. PsycINFO, PsycLIT, PsycFIRST, and ClinPSYC are electronic variations of Psychological Abstracts. Each version is available from several vendors.1 For ex­ ample, as of June 1999, PsycINFO was available as an online service from CompuServe IQuest, DataStar, DIALOG, DIMDI, HealthGate, and Ovid. Institu­ tional leases for PsycINFO were available directly from APA, as well as from Aries Systems Corporation, Cambridge Scien­ tific Abstracts, EBSCO Publishing, NISC, OCLC, Ovid Technologies, SilverPlatter Information, The Gale Group, and UMI. PsycLIT, a PsycINFO subset, is available only on CD-ROM from Aries Systems Corporation, EBSCO, NISC, Ovid Tech­ nologies, and SilverPlatter Information. The electronic versions of Psychological Abstracts provide a complex pricing model that varies depending on database vendor, delivery method, and content. Any of these may be the best choice for a particular library. Functionality is another important con­ sideration when selecting a database de­ livery method. Electronic databases usu­ ally offer several levels of functionality, 58 College & Research Libraries January 2000 TABLE 1 Database Features: Output Options Features Specific Considerations Displaying What is default display (citation, abstract, full-text)? Can default be changed? Is scrolling through results possible from both citation list and record by record? Are search terms highlighted in results? Printing Can parts/sections of document/page be printed? Can print limit (number of pages, records, etc.) be set? Downloading Can saving be done in various file formats (ASCII, HTML, etc.)? Can results be saved and imported into bibliographic software program? E-mailing Can an entire marked list of articles/citations be e-mailed? Can both ASCII files and attached image files be sent? Customization Can output be sorted (by date, author, publication, etc.)? Can one (of all output formats) select certain fields for output other than default? Can length-of- field names be changed? Is search history included? Document delivery Can requests be made directly from the database (through either ILL or a fee-based service)? ranging from base features that are built into the system to optional value-added enhancements, which may come at addi­ tional cost. Database features and options also differ among vendors, and one ven­ dor may offer different nonnetworked, LAN, and remote Internet access versions of the same database. Before selecting a delivery option, li­ braries should determine what database features are necessary for successful implementation. The inclusion of a par­ ticular database feature can never be as­ sumed. For example, if a library wants to focus on e-mailing search results as a way to cut down on local printing costs, it is important to verify that a database supports e-mail. The following list de­ scribes database/interface features, which should be identified as either re­ quirements or preferences for the local situation. Requirements are “must-have” features needed to satisfy user needs, while preferences are desirable, but less necessary, options. The individual fea­ tures can be grouped into five categories: output options, search interface, search options, content, and local issues (see tables 1–5). As the UNL libraries LAN server neared obsolescence, plans were made for migrat­ ing the DOS-based CD-ROM databases to new delivery methods. Of particular con­ cern was finding a successful replacement for the heavily used PsycLIT database. The current PsycLIT SilverPlatter CD was as­ sessed for both positive and negative fea­ tures, and ideal database features for the new system were identified. Aside from pricing, hardware, and network concerns, feature requirements played an important role in influencing the final selection. Besides licensing considerations and lo­ cal requirements, the library’s local situa­ tion may limit the options for delivery of electronic resources. For example, both LAN and remote access require reliable networks. The network is the wire that connects the computer to the server or the Internet and should be of sufficient speed to support graphical user interface connec­ tions. In addition, there should be mini­ mum network downtime when relying on the network for access to important elec­ tronic databases. Moreover, for a LAN, it is necessary to have or be able to hire com­ petent staff to set up and maintain the net­ work and servers. Outsourcing is a possi­ Database Selection: One Size Does Not Fit All 59 TABLE 2 Database Features: Search Interface Features Specific Considerations Stability Is it a fairly static product that undergoes little change in interface? Differences between Are there some things that one interface does that another versions (nonnetworked, does not? Can users move easily between them? LAN, Web, etc.) Interface customization Can you determine which search options appear on the screen, and where? Can you enable/disable selected options? Can you add local information to the search screen(s)? Can you select the default search type (e.g., keyword vs. subject)? Can you set up different interfaces for different user groups? Z30.50 availability Integration/uniformity How important is it for the new product to look and feel similar to existing products? If the interfaces are not exactly identical, are things like truncation symbols and adjacency operators the same? Help screens Is help context sensitive? Is a searchable help index avail­ able? bility, but it will be important to have good response time for maintenance and troubleshooting. In cases without a reliable network or adequate staffing, the only vi­ able option will be databases on nonnetworked computers. A final delivery method consideration is hardware. When planning for the de­ velopment and implementation of new hardware, libraries should not anticipate technology more than five years ahead. For example, according to a research note published by Gartner Group in 1997, the technology for Web clients is predicted to change from simple browser access to more complicated hybrid clients, so it would seem unrealistic to plan beyond a five-year time frame.2 Nonnetworked computers have an ini­ tial start-up cost that is lower than the LAN cost. However, increased staff sup­ port requirements for maintaining mul­ tiple nonnetworked computers can re­ quire a large investment over a five-year time period. A LAN system where the databases are maintained on a single server can save from 30 to 40 percent over the cost of the similarly configured nonnetworked setup where duplicate databases are running on multiple com­ puters. Although less expensive over time than nonnetworked computers, building a LAN still involves several cost consid­ erations. After it is determined that the network is of sufficient reliability and staff is available for ongoing support, the sys­ tem must be properly scaled to support the databases. Proper scaling will include determining the correct number of pro­ cessors, hard-drive requirements, RAM requirements, and planning for upgrades over the next five years. The chart shown in table 6 addresses issues that are impor­ tant in determining the size and future upgrade path for hardware. One system that is commonly used to provide LAN access to electronic data­ bases is Citrix WinFrame. WinFrame has been described in an article by Anna Turner as a “combination of high perfor­ mance and low cost,” which provides a single point of control for database main­ tenance with universal application access that is platform independent.3 The system has built-in support for Web access using 60 College & Research Libraries January 2000 TABLE 3 Database Features: Searching Options Features Specific Considerations Boolean and proximity Can Boolean commands be typed in on the command line as operators well as chosen from a pull-down menu? Is "AND" assumed between search terms, or is phrase searching done automati- cally? Is nesting allowed? Can proximity operators be used to specify more specific relationships between search terms? Field-specific searching Is searching initiated through pull-down menus and/or from the command line? Thesaurus/subject list Does it include broader, narrower, and related terms? Can you select and execute searches directly from the thesaurus? Can you "explode" terms? Will the thesaurus suggest alternatives if your search matches nothing? Lateral searching Are search terms (subject headings, authors' names, etc.) hot- linked and/or searchable from the results list? Limiting Can limiting be done on the initial search, or do you have to limit search results? Are a variety of limit options available? Saving/combining Can you view your search history and combine and/or modify searches previous searches from the current search session or earlier? Multiple search modes Which type of search is the default? Is it difficult to (basic, advanced, etc.) distinguish or move between the two? Multiple database searching Can you select and search multiple databases simultaneously? Multiple year searching Can you search the current years and the backfile simulta- neously? Do you need to select a range of years before starting the database? Can you easily switch between current and backfile? System-assisted Is weighted or relevance searching used? Is natural language searching querying available? Is SDI (selective dissemination of information) available? Will the system automatically look for variant spellings, plurals/singulars? the ICA–client plug-in for Web browsers. In addition to Internet access, the system has the advantage of supporting Win­ dows-based terminals and low-end net- worked computers. This can reduce the overhead for libraries that provide in­ house access to electronic databases. Using WinFrame version 1.8, a single processor (Pentium-based server) should support up to twenty concurrent users. The server requires the following RAM: 16 Mg minimum for the operating system, 4–8 Mg per user, with an additional 1 Mg for each configured WinStation. The operating sys­ tem also requires 300 Mb of free disk space. The UNL University Libraries selected and implemented the WinFrame solution for providing access to databases as a re­ placement for the DOS-based network. The system consists of a server farm of five multiprocessor Pentium servers, one of which is dedicated to providing Internet access to electronic databases. These databases are loaded from CD­ ROMS onto hard drives on this server. A Web browser uses the ICA–client plug-in to launch a connection to the server that controls the databases. The end user is able to use the native vendor interface to access the databases. Keyserver software Database Selection: One Size Does Not Fit All 61 TABLE 4 Database Features: Content Features Specific Considerations Backfile availability How far back does the product go? Full-text What is the percentage of full-text coverage? Is full- text inclusion an additional subscription? If not available, is it being considered? Can links be made from citations to full-text electronic journal subscrip- tions? PDF (portable document file) Is PDF a viewing/output requirement or option in this format product? Graphics/images Are graphics included? Are they part of a full-text document (as in a PDF file), or must they be handled separately? Currency/updating Does the updating frequency differ among the delivery methods (nonnetwork, LAN, etc.)? is used to monitor licensing and gather statistics. The other four servers are used for in-house access to the databases and other applications. Database Selection Model Libraries are faced with the problem of balancing dollar requirements against lo­ cal issues when selecting a delivery method. The following model conceptu­ alizes the outcome of this decision-mak­ ing process (see table 7). Dollar require­ ments include the vendor’s pricing model (simultaneous users, passwords, ports, etc.), access restrictions (no Internet ac­ cess available, not networked, etc.), and additional charges, including add-on fea­ tures. Also included as a cost is the one­ time installation fee for local hardware. Local issues are those requirements deemed necessary for a successful local implementation. TABLE 5 Database Features: Local Issues Features Specific Considerations Links to local holdings Can the database be linked to the OPAC? Can local holdings information simply be included in the database? Is this an additional cost? ILL restrictions Are there restrictions on lending copies of electronic full-text articles (that are not subscribed to in print)? Authentication (remote access only) Who is responsible for authenticating, the vendor or the local library? What is the method for authentica­ tion (proxy, passwords, etc.)? Plug­inslhelper applications Are extra pieces of software required for accessing the database? Will remote users require different software than local users? Supportlhelp desk hours Is online or phone support available all hours the library is open? What is response turnaround time? Are both technical and search support provided? �� � �� �� ��� �� ��� ��� ��� �� ��� ��� �� ��� ���� �� ��� ��� ���� ��� ��� ��� ��� ���� ��� ��� ���� ��� ���� � ��� ��� ���� ���� ��� �� ���� � ��� ���� �� ��� ��� ��� ��� ��� �� ��� ���� ��� ��� ��� ��� ��� � ��� ���� ��� ���� ���� ���� ��� ��� �� ��� ���� ��� ��� ���� ��� ��� � �� ��� ��� ��� ���� ��� �� ��� � �� �� ��� ��� ��� �� ��� ���� ��� � ��� ��� ��� ��� ��� ��� ��� ��� � �� ��� ��� ��� ���� ��� �� ��� ���� �� ��� ��� ��� ���� �� ��� ��� ���� ��� �� ��� � �� ��� ��� ��� �� �� �� ��� � �� ���� ��� ���� ��� ��� ��� ��� ��� ��� ��� ��� ��� ��� �� �� ��� �� ���� ��� ���� ��� ��� �� ��� ��� ��� ���� ��� ��� �� ��� ��� ��� ��� ��� ��� ��� � �� ��� ��� ��� ��� ��� �� ��� ���� ���� � ��� ���� ��� � �� ��� �� �� ��� ��� ��� ��� ��� ��� ��� ���� ��� ��� ��� �� ��� ��� ��� ��� ��� ��� ��� ���� � �� ��� ��� ��� ��� ��� � �� ��� ���� �� ��� ��� ��� � ��� ��� ��� ���� ��� ��� ��� � �� ���� ��� ��� �� ��� ���� ��� �� ��� ���� ��� �� �� ��� ��� ��� ��� �� ��� ���� �� ���� ��� ���� ���� � ��� ���� ��� ��� ���� ��� ��� ��� ��� � �� ��� �� �� ��� ��� ��� ��� ��� ��� ��� ���� ��� ��� ��� �� ��� ��� ��� ��� ��� ��� ��� ���� � �� ��� ��� ��� ��� ��� � �� ��� ���� �� ���� ��� ���� ���� � ��� ���� ��� ��� ���� ��� ��� ��� ��� � �� ���� ���� ���� ��� ��� ��� ��� � ���� ��� ��� ��� ���� ���� � ��� ��� �� ���� ��� ���� ���� ��� � ��� ��� ��� ��� ��� ���� ��� � �� ��� ��� ��� ��� ��� ��� ���� ���� ��� ��� �� ��� � 62 College & Research Libraries January 2000 Outcomes on the left side of the model fendable, inconsequential, or question- have low local value and represent risky able. positions that are most likely to be ques­ tioned. As a result, when a decision falls Ideal into this side of the model, it will be nec- In an ideal situation, cost-effective invest­ essary to justify the decision to either ment is made in a database that provides a frontline staff and/or administrators. De- level of access and functionality to meet the cisions on the right side of the model have needs of the local environment. In this situ- high local value and are less likely to be ation, the library has carefully analyzed the questioned by frontline staff because they database features, evaluated staffing sup- satisfy local needs. However, it is possible port levels, and met budgetary constraints. to overinvest in a high-value solution that The library has neither overinvested nor precludes other purchases. These selec- settled for less than is needed. tions may be hard to justify to individu­ als with budgetary oversight. Outcomes Defendable fall into one of four categories: ideal, de- In a defendable situation, the library is Database Selection: One Size Does Not Fit All 63 seem that compromising TABLE 7 on local issues in order to Decision-making Process for Database Selection Database Selection Model High 00 ""31 �1 .11 S1 -"' High Cost Low Value Questionable Low Cost Low Value Inconsequential Low Local Issues providing a high-end, high-cost solution to its electronic access needs. An example would be an expensive, full-text database that serves the library’s needs but also includes features that are never used. In making this kind of selection, the library may be placed in a position to defend the decision because the high cost may have prevented purchase of other resources. In some situations, a decision may be made to purchase a high-cost database that does not meet local needs and is not used. Inconsequential Inconsequential situations result from in­ vestment in a less expensive database that fails to meet local requirements. These de­ cisions do not adversely affect the budget, but they also fail to serve user needs. It may get a lower price is ben­ eficial, but users may not use the database or may use it inefficiently or in­ effectively. Likewise, for locally mounted data-High Cost bases, staff time is High Value wasted in maintainingDefendable underutilized databases. QuestionableLow Cost Questionable situationsHigh Value arise when decisions are Ideal made without regard to local requirements. In High some situations, a deci­ sion may be made to purchase a high-cost da­ tabase that does not meet local needs and is not used. The li­ brary may be put in the position of de­ fending the budgetary decision, particu­ larly in cases where other purchases were sacrificed to purchase the costly database. Conclusion In conclusion, determining an ideal de­ livery method for electronic resources is a complicated process that involves many considerations. This article uses the UNL experience as one example of how this pro­ cess is designed to work. Each library’s situation is unique and one size will not fit all. However, when decisions on purchas­ ing electronic resources are made with due consideration to local issues and dollar re­ quirements, librarians can be assured that the decisions are fiscally responsible and defendable and meet the fundamental goal of providing access to information. Notes 1. “Comparison Chart,” in http://www.apa.org/psycinfo/compare.html (Washington, D.C.: American Psychological Association) [cited May 6, 1999]. 2. Tom Austin, Hybrid Turmoil and Vigor: Web and Traditional Clients (Stamford, Conn.: GartnerGroup Advisory, 1997). 3. Anna Turner, “Thin Client Architecture for Networking CD-ROMs in a Medium-Sized Public Library System,” CD-ROM Librarian 17 (Sept. 1997): 74. http://www.apa.org/psycinfo/compare.html