Using an automated tool to calculate return on investment and cost benefit figures for resources: the Health Sciences and Human Services Library experience

In difficult economic times, libraries experience greater pressure to justify budget expenditures to their fiscal managers. Over the last decade, library administrators have become increasingly interested in incorporating return on investment (ROI) and cost benefit analysis (CBA) techniques into their reporting structures to provide quantitative evidence of the worth of services and resources. 
 
The Health Sciences and Human Services Library (HS/HSL) of the University of Maryland appointed a small task force in 2009 to experiment with an automated instrument for calculating ROI and cost benefit ratios (CBRs) for its collection of books and journals. The tool was developed by the National Network of Libraries of Medicine (NN/LM), MidContinental Region [1], and promoted by them in a series of presentations [2, 3]. This article describes the process used to compile the statistics for the calculator, first using data from the 2008 fiscal year and then with data from the 2011 fiscal year. 
 
The HS/HSL supports the graduate school and the schools of dentistry, medicine, nursing, pharmacy, and social work, which have a combined faculty and student full-time equivalent (FTE) of 6,797. Additionally, some of the library's resources are available to the University of Maryland Medical Center community, a 705-bed teaching hospital with approximately 6,500 employees as well as 1,000 attending physicians. The collection includes approximately 8,500 print and online journal subscriptions, 174,000 print and online books, and 96 databases. The HS/HSL is also home to the NN/LM Southeastern/Atlantic Regional Medical Library. 
 
 
ROI is expressed as a percentage and represents the “rate of return” (loss or gain; in other words, benefit) versus the initial “investment” (total cost). The mathematical formula for calculating ROI is: 
 
 
 
 
 
CBA results in a CBR that compares a dollar spent versus dollars reaped in benefit, which is determined by dividing total benefit by total cost [4].


INTRODUCTION
In difficult economic times, libraries experience greater pressure to justify budget expenditures to their fiscal managers. Over the last decade, library administrators have become increasingly interested in incorporating return on investment (ROI) and cost benefit analysis (CBA) techniques into their reporting structures to provide quantitative evidence of the worth of services and resources.
The Health Sciences and Human Services Library (HS/HSL) of the University of Maryland appointed a small task force in 2009 to experiment with an automated instrument for calculating ROI and cost benefit ratios (CBRs) for its collection of books and journals. The tool was developed by the National Network of Libraries of Medicine (NN/LM), Mid-Continental Region [1], and promoted by them in a series of presentations [2,3]. This article describes the process used to compile the statistics for the calculator, first using data from the 2008 fiscal year and then with data from the 2011 fiscal year.
The HS/HSL supports the graduate school and the schools of dentistry, medicine, nursing, pharmacy, and social work, which have a combined faculty and student full-time equivalent (FTE) of 6,797. Additionally, some of the library's resources are available to the University of Maryland Medical Center community, a 705-bed teaching hospital with approximately 6,500 employees as well as 1,000 attending physicians. The collection includes approximately 8,500 print and online journal subscriptions, 174,000 print and online books, and 96 databases. The HS/HSL is also home to the NN/LM Southeastern/Atlantic Regional Medical Library.
ROI is expressed as a percentage and represents the ''rate of return'' (loss or gain; in other words, benefit) versus the initial ''investment'' (total cost). The mathematical formula for calculating ROI is: CBA results in a CBR that compares a dollar spent versus dollars reaped in benefit, which is determined by dividing total benefit by total cost [4].
To complete the NN/LM calculator for ROI and CBR, the library must supply thirteen values ( Figure 1, online only), across three sections, one for salary data and two with cost-and benefit-related fields for each type of resource. The specific elements are listed in Table 1.
Once all required metrics are entered, the tool automatically calculates benefit and cost figures and the resulting ROI and CBR values. The formula used to calculate the benefit for books or journals is: usage|price per article or book ð Þ z user salary=hours worked per year ð Þ ½ | usage|time saved per article or book ð Þ The formula used to calculate the cost of books or journals is: book or journal budgetz library salary budget|percent of staff time ð spent managing book or journal collectionÞ

METHODS
The task force began its work in 2009, using library metrics for the fiscal year 2008 (July 2007-June 2008), since the standard reports regularly issued by the HS/HSL had been completed and were available for that fiscal year. Subsequently, an error was discovered in the 2009 calculations, and the exercise was then repeated with 2011 data. This study reports the results for both years; differences in calculations between the years are noted below.

Users' average annual salary
According to a recently completed survey conducted by an outside consultant, the HS/HSL user population consists of the following categories and percentages: undergraduates, 12.5%; graduates, 56.5%; faculty and postdoctoral fellows, 18%; University of Maryland Medical Center staff, 9%; and unaffiliated, 4%. For practical purposes, usage of the library was considered to be equivalent across all population categories.
Financial aid and stipend figures were substituted for salary totals for undergraduate or graduate students with data from the university's student financial assistance and education department. The faculty and staff salary average was obtained from the university's human resource services office. For the unaffiliated group, Baltimore salary and unemployment numbers were compiled from the US Department of Labor, Bureau of Labor Statistics [5], and the Maryland Department of Labor, Licensing and Regulation [6], with the average salary figure adjusted down to reflect the unemployment rate. The totals for each category were weighted according to the Supplemental Figure 1 is available with the online version of this journal.
percentage represented for a combined average to be used in the specified element.

Users' hours worked per year
The full-time total of 2,080 hours per year was used with the understanding that it is higher than the actual number because no adjustment was made to account for those who work part time. However, it was considered to be the most ''justifiable,'' even for the student population, given the considerable hours spent on academic activities.

Library salary budget
The sum of all regular personnel salaries including fringe benefits was used for this figure. This element required considerable manual calculation because the library currently has no single report that provides the data for both salaries and fringe benefits. Salaries for the NN/LM staff were excluded from the total.

Book and journal budgets
Expenditures for both print and electronic formats were compiled for each collection, with bindery costs added to the journal total.
Portion of all staff time devoted to the book or journal collection Two fields are associated with the percentage of library staff time spent processing and handling books and journals. Only those individuals whose responsibilities were directly related to acquiring, organizing, and maintaining the collections were included in the figures. Therefore, the totals reflected staff primarily involved with acquisitions, cataloging, physical processing, e-resources management, shelving, and reserves. Several personnel split time between collectionrelated tasks and other activities. In those cases, only the hours devoted to the former were included in the calculations. Department heads reported the hours per week spent by staff on each type of resource. Total hours for the year were calculated, converted to a percentage of total staff hours, and entered in decimal form. However, these were estimates as no time studies have been conducted.

User time saved for each book borrowed
The user time saved for each book borrowed figure, and the equivalent element for journals, were taken directly from the examples provided in the NN/LM presentation [2]. Expressed in 100ths of an hour, 0.25 (15 minutes) and 0.33 (20 minutes) were used for books and journals, respectively.

Average retail cost of a book
The Matthews Medical Books Average Price Report was used to calculate the average retail cost of a book, with the same value applied to both print and electronic formats.

Per-article price from a vendor
Article prices were taken directly from the examples in the NN/LM presentation. The 2 values used in these examples were $30 and $35; the former was chosen as the more conservative figure.

Number of books borrowed or used
Usage statistics for print books were compiled from circulation data. In-house use of print monographs (left on tables and in group study rooms) was not included in the book total since it has never been tracked.
Fiscal year 2011 usage statistics for electronic books were calculated from those vendors who were Counting Online Usage of Networked Electronic Resources (COUNTER) compliant, using either ''successful title requests'' or ''successful section requests.'' Of the two non-COUNTER compliant providers, one reported the ''number of downloads of a title,'' which was acceptable. However, existing data from the second vendor were not usable because of the high probability of multiple counts for a single access. Upon request, the provider was able to furnish the number of single ''visits'' per month for each book.
Usage statistics for fiscal year 2008 were also calculated from vendor reports. However, the Brief communications: Bodycomb and Del Baglivo e-monograph figure included ''page view'' totals for one of the packages, which led to an inadvertent inflation of the usage figure for that year. At the same time, the first six months of usage for three emonograph packages were unavailable for the 2008 fiscal year due to vendor error, leading to an undercount of usage for these packages. As a result, after some discussion, the decision was made to repeat the process, with fiscal year 2011 data.

Number of articles read by all users
Print usage for journal articles was compiled from reshelving activity, with one volume shelved equivalent to one article read. Electronic article use was calculated using ''number of successful full-text article requests'' from COUNTER-compliant vendors.
Of the two providers who were non-COUNTER compliant, one reported ''number of downloads'' and the other ''full text visits,'' both of which were considered suitable. The hypertext markup language (HTML) and portable document format (PDF) figures were added together for those vendors offering users a choice between the two. The HTML count was used in the case of a single journal for which the PDF is accessible only by first going through the HTML version.

RESULTS
After entering the required data ( . This result suggests that the HS/HSL figures are valid even though the comparison group is extremely small. As noted above, the HS/HSL fiscal year 2011 CBR and ROI values for monographs were lower than those found for fiscal year 2008, partially due to the inadvertent inclusion of ''page views'' in the latter. However, the difference in actual usage totals between the 2 years was only 8,408, a drop of 9%. There are several possible explanations for the smallerthan-expected difference. The 2008 total was missing 6 months' worth of statistics for 3 packages. Print book renewals were included for 2011, whereas they were excluded in the earlier calculations. Finally, the emonograph collection has grown since 2008 with a corresponding increase in usage. A contributing factor in the decline of the values for monographs over the 3 years was a $372,991 rise in the library salary budget, 49% of which is attributable to the cost of fringe benefits. Approximately one-third of the total increase is due solely to a change in the State of Maryland accounting procedures for retiree health benefits. Despite the decrease in CBR and ROI values for monographs, the combined CBR and ROI for 2011 was actually higher than in 2008, primarily due to e-journal usage, which rose by 30%. The HS/HSL migrated the majority of the journal collection to online over the intervening years with the result that print usage dropped by 46% (17,055 for fiscal year 2008; 9,157 in fiscal year 2011). The increase in the salary budget noted above had far less of an impact on journals than monographs because the number of articles read is so much greater than the number of books used.
The NN/LM's automated tool incorporates only those expenditures that are entirely attributable to each collection. Indirect costs such as those associated with physical plant operations, and software and hardware purchases are not included. However, apportioning these expenditures to books and journals would be extremely difficult. Therefore, the ROI and CBR values should be evaluated in the context of the calculator's limitations.

CONCLUSIONS
Compiling the figures for the calculator for 2008 was a time-consuming process. Some elements required * The J. Otto Lottes Library's values are a ''quick look'' and do not reflect a complete analysis because their data are often combined with that from other campus libraries (via personal communication with Barbara Jones, 11 Oct 2011). significant manual computation (e.g., average user annual salary and library salary budget), while data obtained from departments often required repeated clarification to ensure that there were no misunderstandings regarding what was or was not included. Despite all of the efforts to ensure accuracy, the original monograph usage total was flawed. Therefore, when the process was repeated using data for the 2011 fiscal year, considerable time was spent reviewing and discussing the different types of usage statistics reported by vendors, especially from those vendors who were not COUNTER compliant.
Having learned from the initial experience, the compilation process for 2011 was less time consuming, although manual computation was still necessary for the same elements as had been required when analyzing data for 2008. However, the effort was worthwhile because the results provided yet one more justification for the annual expenditures for journals and books. Detailed documentation for the compilation process has been maintained for both years. Suggested improvements for the future include conducting a time study for staff involved in managing resources and instituting a method for collecting usage statistics for print monographs that have circulated ''in house.'' The HS/HSL is considering repeating the project on a triennial basis. Not only are the final results informative, but the entire process serves to clarify and define all of the data associated with the two collections.