Keywords

Demand-driven acquisitions; research; ebooks; cap prices; short-term loans; cost per use; expenditure per volume; circulation; planning; budgets; graduate students; faculty; undergraduate students; interlibrary loan

3.1 Cap Prices, Short-Term Loans, and Triggers

The first cost consideration for DDA programs should come in setting the parameters of the pilot. Choices made in the setup phase, like the inclusion of short-term loans and the price cap on books in the discovery profile will have an impact on the expenditures of the program and the cost per loan in the assessment. The price cap determines the cost of each individual book included in the discovery profile. Librarians can set the cap so that more expensive titles are excluded from DDA programs altogether or require mediation by a librarian before they are purchased.

This cap price may also influence the subject area of the books included since many STEM titles may be excluded from a price cap that is too low (Waller, 2013). Purdue University experienced this issue when they transitioned to an interlibrary loan-to-purchase program in 2000. When they started the program, they set the price cap at $50, but expanded it to $100 to include more technical titles (Ward, 2002). Determining a price cap will be different for each library depending on the subject areas, projected expenditures, and goals of the program, but these choices can be adapted based on outcomes to achieve desired results even once the program is underway. The University of Mississippi chose a cap of $150, because 95% of their firm orders in the previous fiscal year were under $150 (Herrera & Greenwood, 2011) and this approach was successful. A later study from Purdue found that even within their STEM title DDA, most of the titles purchased cost around $100 (Bracke, 2010). Their cap price was set at $150 and analysis of DDA triggers revealed that this was an appropriate cap. It seems particularly common in the research for institutions to begin pilots with more restrictions on purchases and lower cap prices and then loosen these restrictions once assessment of the collection reveals improved usage numbers and lower cost than expected (Booth & O’Brien, 2011; Doyle & Tucker, 2011; Gilbertson, McKee, & Salisbury, 2014; Joyner Cramer, 2013; Ward, 2002).

A more inclusive cap price can be intimidating for libraries beginning a DDA program, The University of Kansas Libraries found a compromise by customizing the price cap workflows for their disciplines. Social science titles below $150 were added into the catalog, but titles above $95 required mediation by a librarian to purchase. For STEM titles, the same workflow was set up, except titles under $200 were loaded and titles above $150 had to be mediated (Currie & Graves, 2012). A study from the University of Nebraska–Lincoln suggests that this compromise is ideal for libraries transitioning from liaison selection to DDA because it frees librarians from selection at the lower level and lets them use their subject expertise to choose fewer investment titles that will help ensure deep collection building (Tyler, Falci, Melvin, Epp, & Kreps, 2013). Mediation will make the DDA program more hands on, and could help libraries that are apprehensive about going over budget get more control over their acquisitions during the pilot, while still offering their users a large discovery pool. The cap price should not be too low though or it may exclude titles that would be useful for students and tip the discovery pool towards more arts and humanities titles at the exclusion of STEM titles. Analysis and testing are good methods for determining a solid cap price for DDA collections, but $150 also seems like a research-backed rule of thumb.

Starting a DDA program does not require libraries to choose a single cap price that will apply to all materials. The University of Kansas Libraries had librarians in different disciplines choose different cap prices for their ebook DDA materials. Social sciences librarians chose a $94.99 price cap for unmediated materials, but discovery records up to $150 were available with mediation. Science and technology librarians used a $150.00 unmediated price cap and up to $200 mediated. The humanities were not included in this DDA trial. The average price for materials already in their collections was $113 for social science ebooks while science and technology ebooks had an average price of $125 (Currie & Graves, 2012). This bespoke solution is another good strategy for libraries that are worried about exclusion with a lower price cap. Bumping up the price cap just for STEM titles might help expand the discovery pool while controlling costs.

Many DDA programs also include a short-term loan (STL) program as part of their initial strategy. Short-term loans allow library users unlimited access to the triggered titles for a limited amount of time. Many programs build in a set number of short-term loans before purchase is triggered and in most systems this process is seamless for the user. Patrons can access these titles as they would access any ebook in the system and will often not be notified that they have triggered either a short-term loan or a purchase. The most significant cost-impacting factor for this type of loan is the number of loans before purchase is triggered. It is common in the research to set this high, at three or four STLs before a purchase, but this will result in more funding allocated to temporary collections than building a permanent collection. When beginning and evaluating DDA programs for the number of short-term loans, it is important to establish a goal for the program’s loans. If access is the ultimate goal and collection building is less important, more STLs should be included.

Cost allocation to different models of access should be examined as part of any assessment effort. The number and duration of STLs is easily customizable and should be adapted based on assessment from the pilot to maximize cost savings. If there are worries about triggering enough titles to maintain the rate of collection building, DDA programs may want to adopt a system with fewer STLs than the ones piloted at the University of Texas and Open Polytechnic. A study of the NetLibrary usage across five academic institutions found that a very low percentage of titles were used more than four times (Christianson, 2005). A trigger set at or below four uses seems common, but the number of triggers between two and four STLs varies greatly and will depend on institutional factors.

Access versus ownership is an issue that was initially raised by interlibrary loan-to-purchase DDA programs, which theorized that purchasing materials that aligned with collection goals was likely to save the institution money in the long run over borrowing the item. It is still unclear that digital books provided by a vendor are actually owned by an institution, though there are efforts to more clearly define this relationship (Hamaker, 2016). Though many ebooks are sold on perpetual licenses, it is difficult to predict whether and how the materials will remain accessible as technologies change (Kieft, 2010). They open up a bigger pool of resources: California State University, Fullerton, created a DDA pool based on their existing approval plan, but outside of this, librarians were also sent notifications for items which did not meet the criteria for automatic purchase, but that they could select. Fullerton decided to further integrate these longer tail items into the discovery pool if they were available for DDA. They used STLs to drive down costs with four STLs before purchase was triggered. This method allowed the library to cut their approval plan costs by about half, while maintaining access to a similar number of titles as the past fiscal year and using more ebooks, which was a goal. They recommend a hybrid plan with approval/selection and DDA (Roll, 2016).

There is also some evidence that more STLs may save money. Grand Valley State University investigated the ideal number of STLs in great detail. For their trial they used three STLs before purchase and the results of their trial showed that they spent $41,891 on STLs and $26,947 on purchases. They examined scenarios with four to seven STLs and found that the more STLs before purchase, the more money the library would have saved up to seven STLs since many titles did not see usage after purchase. Grand Valley State University found that a combination of short-term loans and nonlinear lending access saved the library a lot of money. They spent about $70,000 to give their users access to over $4 million worth of content. They subsequently decided to remove price limits from their profile (Way & Garrison, 2011). The University of Iowa began their program without STLs but found that about 21% of the titles had been used only once (trigger purchase) adding one STL would have saved them $108,000 over the 4-year program. They implemented one STL with purchase triggered on the second (Fischer & Diaz, 2014).

Virginia Tech has an e-preferred policy and took up a DDA program in 2011 to help save money. They loaded 12,800 discovery records into the system, of which 1204 were triggered at a cost of roughly $17,000. They also included 532 short-term loans and projected that they saved about $82,000 for the same level of access (Stovall, Lener, & Gilmore, 2014). Wake Forest University’s program had five STLs with trigger on the sixth purchase, an unusually high number but one they considered to be successful. A total of 2224 books were loaned at least once, but only 67 were triggered (Joyner Cramer, 2013).

Kent State University Libraries took this idea in the other direction. Though their own program featured a trigger to purchase strategy with no STLs, they modeled a scenario with three STLs before purchase. In this scenario, 78.2% of their loaned DDA materials in the model were subsequently going to purchase. If they had adopted the STL model instead they would have been dumping money into loaning books that they were eventually going to own (Yin, Downey, Urbano, & Klingler, 2015). The Kent State University model does not include consideration of the circulations triggered after purchase, but it seems likely in light of the Grand Valley State University, University of Iowa, and Wake Forest University studies cited above, along with the usage ranges we tend to see in DDA studies, that extending the number of STLs could be a cost-saving measure as well, though many fewer items would be permanently acquired for the collection.

This cost saving has a few caveats though, STL items are not added to the catalog and most of these studies do not include good data for after the pilot year, when circulation may continue to grow. The use of STLs is institutionally specific and factors such as the level of scrutiny at the discovery level, the program’s budget, and the allocation for permanent collection over short-term use will affect whether to include STLs and how many to incorporate before the item is triggered for purchase. Kent State University’s STL program was plagued by a large amount of inquiry spending, the money spent on STLs to discover that the item actually belongs in the collection. The more inquiry spending the institution is willing to do will have a strong influence on the amount of acquisitions spending that occurs in the DDA process. In the case of STLs, there is a markup for inquiry that should be factored in to the savings price of a large number of STLs.

A few libraries have found success in preevaluating the usage of their existing ebook collections to right size the number of STLs for the way their users were already accessing content. The University of Texas Libraries examined 10 years of NetLibrary ebook usage data before embarking on their DDA pilot and found that very few of the titles were used more than four times, so they set their trigger model at three STLs and then a purchase on the fourth use so that only titles with demonstrated repeated use would be purchased perpetually. Their program was automatic below $50 rental and mediated above $50, so STL requests were sent to an approval queue and the mediator could send the book to outright purchase if they thought that was a better option. The Library spent $190,043 on STLs and $96,806 on purchases (Macicak & Schell, 2009). Although these data were not used towards a DDA program, a similar ebook analysis conducted on collections from Louisiana State University, Auburn University, the University of North Texas, Texas A&M Commerce, and Nunez Community College reinforced the idea that the majority of titles receive no use, most receive fewer than four uses, and a very small percentage get hyper use. This is a strong case for cost savings in STL.

STLs can be a powerful tool in the DDA arsenal, but the 2014 NISO report notes that some publishers are concerned about the sustainability of STL purchases because they can be so revealing and economical for libraries (NISO DDA Working Group, 2014). It is worth experimenting with the number of STLs because they have a significant impact on the cost of DDA programs, but it is probably wise to be cautious of relying on them as a part of permanent practice. STLs could increase in price, which would spike inquiry spending and shift the balance back towards fewer STLs before purchase.

California State University–Fullerton began a DDA program with 25,596 discovery records, which by the end of the trial had nearly doubled in size due to monthly additions. They excluded disciplines that were covered under their consortial agreements, anything over $250, non-English titles, and anything published before 2008. They set up three STLs with purchase happening on the fourth trigger. Over the 8-month trial they had 1813 transactions and spent $13,688.70 on short-term loans and $5840.43 on purchases. They showed 269 downloads of 111 unique titles by 104 downloaders. The average use time of items was 44 minutes for purchases and loans, with purchased titles earning longer watch times. Overall, they found that each transaction during the program had an average cost of $10.78, which they deemed successful (Breitbach & Lambert, 2011).

Programs that forego STLs and rely on patron usage indicators to trigger purchases should also be mindful that the effort required to trigger can sometimes be negotiated to control budget. Kent State University’s DDA program uses a standard 10-10-1-1-1 trigger model in which 10 page turns, 10 minutes spent reading a book, or one copy, print, or download will cause the auto purchase of the item (Yin et al., 2015). Open Polytechnic in New Zealand wrote thoughtfully about trigger mediation after their DDA trial in which they chose to mediate STL triggers that were above $15. After this experience, they decided that it would have been better for them to choose a lower cap price that would apply to all of their discovery materials rather than trying to slow STLs at the point of purchase (Kelly, 2010).

Once items are triggered, libraries can set up the kind of license type that will be purchased. Common license types include single user, multi user, and nonlinear lending. Grand Valley State University’s DDA program used the nonlinear lending option. With nonlinear lending, titles are competitively priced, but have a limited number of triggering uses per year. Once the uses are expended, the library can buy an additional copy, but each year on the anniversary of purchase the uses renew. Grand Valley also chose to use STLs with their program and to let librarians mediate STLs if they were above $40.00. The researchers looked at a 12-month COUNTER BR1 report and found that the 6239 titles that had been acquired through the DDA program were used 10,514 times. Of these uses, 336 were postpurchase uses. They spent $26,947 on purchases and $41,891 on STLs. The cost of purchasing all of the titles they accessed during the year would have been $550,464. They looked deeply into short-term loans to see if they were using the optimal number of loans before purchase. They found that the majority of purchased titles had no subsequent circulations during the trial period (though this was over only a matter of months) so Grand Valley determined that they had purchased books too early. If they had purchased books on the fourth loan instead of the third, 89 titles would have been triggered and Grand Valley would have saved $14,055.11. Up to seven loans would have saved the library money though this may change as time goes on and more users access the purchased books. The nonlinear lending was seen to be a good thing, none of the books were even close to running out of uses over the 12-months (Way & Garrison, 2011).

This case brings to attention the curve of patron usage. Determining the number of uses that the majority of used titles earn and setting STLs to trigger purchase either before or after this point, at the apex of the patron usage curve, will maximize savings in electronic DDA. The choice between triggering before the max use or after the max use depends on whether the program's goals include purchasing or are simply about access. Programs that emphasize purchasing, like Kent State University Libraries (Yin et al., 2015), should eliminate STLs or set them below the max use, while programs that focus on access and want to save money, like Grand Valley State University (Way & Garrison, 2011), should set their triggers after this max use to eliminate the purchases of all but the most frequently used materials. Libraries can use other ebook acquisitions methods already in place, like package collections or single-title ebook purchases, to determine a good guess for max use on ebook materials.

Cap prices, STLs, and trigger management represent ways for libraries to build budgeting into their initial DDA plans or tweak these plans after observing high or low spending in pilots. Norm Medeiros’ article exploring the DDA trend in 2011 noted that while the content delivery aspect of the process was attractive, many libraries were experiencing uncontrolled spending and an inability to plan budgets with this new process (Medeiros, 2011). In contrast, Cal Poly Pomona found an advantage to DDA pledge funds in that they were able to secure those funds ahead of time for library spending and avoid reallocation of unencumbered funds by the University (Vermeer, 2015). These options have a strong influence on spending and the balance of purchases to accesses. Libraries considering DDA programs should outline goals before choosing a strategy for their discovery profiles and triggers and monitor these options closely to customize the rate of expenditure.

3.2 Cost Per Use and Expenditure Per Volume

Are DDA programs cheaper than traditional acquisitions methods? There are several ways to interpret the value and cost of volumes in the collection. There are also significant costs to the library involved in selecting, accessing, and making materials available to patrons. Together these costs represent expenditure per volume, the money and time spent by the library outside of the acquisition of the book to make it available to users. DDA programs are also very diverse, the costs and expenditures for physical book interlibrary loan to purchase programs will be very different from those associated with catalog-integrated ebook DDA. Cost per use is frequently measured for both physical and digital DDA programs and this measure is useful in different ways.

There are also many ways to interpret the value of physical and digital collections. The most significant measures for physical collections are initial year circulation rates and the value of the collection over time as captured by circulation percentages as the collection ages. For electronic subscription collections, collection use percentages are a useful measure, but for selected electronic acquisitions and DDA programs we generally look to both total circulations and more granular uses, prints, and downloads. When we think about the cost per use for materials it is necessary to keep in mind that there is a big difference between uses for physical and digital materials and this can influence the way we measure the impact of DDA programs.

Circulations are the only measure we have for physical volumes and it represents only one particular act, the grand gesture of taking home a particular book. It does not represent how useful the book was to the patron’s research or how much time (if any) they spent using it. This number also leaves out all of the uncountable uses physical books receive as patrons examine their contents in the catalog or visit the book on the shelf without checking out. Ebook usage numbers represent all of these uses, which is part of the reason cost per use of these collections looks so impressive.

Many studies of both interlibrary loan-to-purchase programs and catalog-integrated DDA remove the first circulation of these materials from the data since this was the transaction that triggered the purchase (Tyler et al., 2011). Should this be the case? It depends on the mission you are seeking to serve. If the goal of your assessment is comparing to peer collections, focusing on preservation, or comparing to print or single-title purchase ebook collections, it makes sense to remove this transaction in your data to level the playing field. If you are assessing the DDA program as useful within your own community or as a way to serve the research interests of particular groups (faculty or researchers) then including this circulation might provide important proof that the item was required and delivered by the library. Most libraries of course want to strike a balance between serving immediate needs and building subjectively good collections, but for evaluation it is useful to consider the goals of the research when constructing the parameters. There are a number of studies examining the difference between the costs and values of physical and digital DDA programs across many different types of libraries.

One of the reasons that programs often adopt electronic DDA as a strategy is to reduce the expenditure per volume for adding materials to the catalog, but this may not always lead to reduced expenditures. Auburn University at Montgomery conducted a very nuanced analysis of the expenditures and costs associated with adding and maintaining an ebook or physical book in their library and determined that ebooks, though their expenditures were 30% lower than adding a physical book, were still more expensive due to ebook markup. They decided to continue to purchase ebooks due to space constraints, but to delay adopting a DDA program because budgets were uncertain (Bailey et al., 2015).

The way libraries deliver service can impact whether they decide to adopt ebook DDA programs. Open Polytechnic in New Zealand had been sending materials by mail to their 24,601 students at significant cost when they embarked on the DDA program to bolster access to electronic materials and ebooks, which are cheaper for the institution to maintain and easier for students to access (Kelly, 2010).

There is also evidence that cost per use is low for DDA programs. Cost per use is used in many institutions to measure the value of resources by dividing the entire cost of a collection or purchase by the number of patron uses it inspired. The University of Arizona had a good print circulation rate for its traditionally acquired monographs (60%), but DDA bumped this up significantly. Patron-driven acquisition (PDA) books circulated an average of 2.6 times and all titles circulated. As for electronic books, though the University spent over $700,000 on ebooks, the cost per use was very low: $0.63 for language and literature, $0.47 for the sciences, $0.63 for business, and $0.75 for engineering (Dewland & See, 2015). Kent State University also found a reduction in cost per use for both print and digital books when they began a cross-format DDA program in 2012. They loaded 22,018 discovery records into the system, and of that 456 records were triggered. Print DDA books circulated 2.6 times on average at a cost per use of $18.24, while the cost per use for ebooks ranged from $0.11 to $5.28 (Downey, 2014).

Brigham Young University looked more deeply into the differences between physical book DDA and traditional acquisitions. They purchased DDA print books through unmediated faculty requests, student suggestions, from their list of items with multiple holds, and through interlibrary loan-to-purchase. The items from the multiple holds list, which already had demonstrated interest, had the lowest cost per use at $0.84. Interlibrary loan had the next lowest at $4.18, followed by student requests at $4.88 and faculty requests at $8.35. Traditional acquisitions had a cost per use of $17.47 (Schroeder, 2012).

There are also many costs associated with obtaining and storing physical books which increase the expenditure per volume, not only at the acquisitions stage but throughout the lifecycle of the book. Paul Courant and Matthew “Buzzy” Nielsen wrote an article for the Council on Library and Information Resources report The Idea of Order: Transforming Research Collections for 21st Century Scholarship that systematically breaks down all of the costs of storing a physical book from keeping the lights on to keeping staff around that can locate materials (Courant & Nielsen, 2010). While keeping and storing physical volumes remains costly, the value of storing materials declines as content becomes freely available online. Purchasing materials in digital format is attractive for many libraries because expenditure per volume is so low.

Though cost per use and expenditure per volume can both be low for DDA programs, controlling the budget spending can still be a challenge and budget challenges are not unique to DDA programs. The University of Nebraska-Lincoln analyzed the costs for their print DDA books alongside their traditional acquisitions strategies. Books in their interlibrary loan-to-purchase program seemed to cost slightly more than approval plans on average, but this difference was modest and only 13 of the 555 books over $200 that they had purchased during the study were requested through the interlibrary loan-to-purchase program.

The University of Nebraska-Lincoln also calculated cost per use for these materials, using both the classic model of dividing the cost of the item by the number of circulations, but including another method because this method favors items that have been on the shelf longer. To avoid this, they also divided the total circulations per year of the books by the total cost of the items to calculate the price of a single circulation for each group of materials they studied. Instead of using Library of Congress subject classes, they separated materials out into topical groups. The first group was a catchall of subjects that did not neatly fall into the arts and humanities subgroup: general literature, biography, and library and information science (LIS). It is interesting that, particularly in LIS, the approval plan, librarian firm orders, and interlibrary loan-to-purchase all performed well, with low spending per circulation and many circulations. They also found that the percentage of funds spent on LIS interlibrary loan-to-purchase contributed significantly to the overall annual circulations of the books. Arts and humanities purchases accounted for 45% of all books purchased. In this collection they found a big advantage in the interlibrary loan-to-purchase method. Books purchased through other methods in this area were almost three times as costly per circulation. Social sciences had a high average turnover for all acquisitions, but interlibrary loan-to-purchase materials were only half as costly per circulation as books purchased through other methods. The University got more circulations on the dollar for interlibrary loan-to-purchase strategies in 19 out of 23 subject groups (Tyler et al., 2011). This is good evidence for the value of a DDA program in one particular university, even though it was a mediated interlibrary loan-to-purchase program.

Because the purchasing activity is controlled by patrons, spending is often difficult to predict. Programs should aim to set a budget but also have plans in place in case the budget is spent too quickly or too slowly (Dahl, 2012). Many vendors offer pledges or prepay amounts to help DDA programs into existing fiscal year workflows. Other ways to control spending in the program include restricting discovery records to particular dates or particular dollar amounts or mediating above cap prices. It is also possible to shut off untriggered discovery records in the system to curtail spending at the same time as the firm order deadline for the library (Cleary, 2015) and this pause in service might provide a good opportunity to clean up the discovery record pool for items that have not been triggered. The University of Iowa ran into budget troubles when they undertook a 1-year PDA program in 2009. It was incredibly successful in terms of both usage and circulation. In the first 2 months of the trial, users spent $28,000 on ebooks, it was so popular that the University had to withdraw many of the discovery records they had loaded into the catalog to fund it sustainably. Eighty percent of the ordered items were used between two and 10 times during the pilot. The institution also used session cookies to ensure that these were unique users (Fischer et al., 2012). Assessing the rate at which the program is spending money is extremely important for DDA. Program spending should be assessed much more frequently than other acquisitions strategies and many vendors deliver weekly reports of triggers.

3.3 Questions for Assessing Collections Based on Cost