This paper is the intellectual property of the author(s). It was presented at CAUSE98, an EDUCAUSE conference, and is part of that conference's online proceedings. See http://www.educause.edu/copyright.html for additional copyright information.
Note: This paper was orginally published in CAUSE/EFFECT Volume 21, Number 2, 1998, pp.38-45.
What Do Information Technology Support Services Really Cost?
by Karen Leach and David Smallen
Leaders of IT organizations are troubled by the lack of reliable benchmarks or comparative data on which to base decisions about support services. Presidents are mystified by total cost of ownership (TCO) studies that appear in the popular press and higher education publications. The numbers quoted in these studies don't seem realistic in light of actual expenditures on campus. Institutional trustees, concerned about rising annual costs for information technologies, are asking hard questions about the impact of services on the educational program and the value the institution is getting for its investment. The COSTS project aims to obtain consistent and reliable data, analyze it in meaningful ways, and develop benchmarks for IT services. This article describes the project's origins and work to date and reports initial results in three support services areas.
In November of 1996, Steven Gilbert, moderator of the American Association for Higher Education's technology listserv, was experiencing a personal technology support crisis. His laptop computer was down for repairs, his office server was experiencing problems, and he found himself unable to function in the Internet world on which he depends. In a posting to the listserv, he wrote:
As I've been suffering the effects of my own personal "support service crisis," I've realized that there are a lot of people who want the same thing I do. We want powerful, effective tools that are utterly reliable, available, and easy to use. We want to spend our time figuring out how to use these tools to improve teaching and learning--how to do our work more effectively and efficiently. We don't want to spend much time figuring out how to use these tools and their successors... or how to cope with unexpected limitations or anomalies. We especially don't want to spend time trying to get our machines fixed or dealing with software packages that interfere with each other in mysterious ways. And we want the full costs associated with these capabilities and services to be highly predictable. We're willing to pay a premium for all this, but we'd like that premium to be as low as possible.
Gilbert went on to detail the kind of support services he felt were desirable.1
This listserv posting prompted the authors to realize how few information technology (IT) organizations have a strong understanding of the various costs of delivering IT support services and their components across the industry. Yet we are asked periodically by our senior administrators about benchmarks that our institutions could use to evaluate the services we deliver.
Intrigued with the idea of pursuing a solution to this challenge, at a "birds-of-a-feather" session at the 1996 CAUSE conference a week later in San Francisco, we provided attendees with a proposal to participate in a project to gather some preliminary data to help develop reasonable benchmarks for, and estimates of, the cost of delivering excellent IT support services. Fifty people agreed to participate, and the COSTS project was born. To increase potential participation, the audience was expanded to include the IT directors at the fifty-four schools in the Consortium of Liberal Arts Colleges (CLAC).2 A year later, after preliminary results were presented at the 1997 CAUSE conference more than 100 institutions joined the project.
The project is organized around short- and long-term goals, a conceptual framework, test hypotheses, and templates for data collection.
Clearly defined goals are necessary to explain the project to potential participants and keep our efforts on track. The short-term goals of the project are:
- Identify and understand examples of IT services at institutions of higher education.
- Develop ranges for the unit costs of providing IT services based on institutional characteristics.
- Test simple hypotheses about the unit costs of providing IT services. The long-term goals are:
- Identify a core of IT services that should be common to most institutions of higher education.
- Identify exemplars for each IT service, that is, institutional approaches that deliver exceptionally high levels of service at identified cost levels.
- Develop benchmarks that are useful for comparing the costs of providing IT services among various institutional categories.
- Determine components of the total cost of ownership (TCO) for desktop computer equipment in higher education.
Information technology services in higher education fall into two categories: those that deal with the care and feeding of the infrastructure and those that provide the related support services (see Table 1).
Table 1: Categories of IT services Infrastructure-Related Services Support Services Desktop Hardware/Software
- Desktop Repair
Administrative Information Systems
Research and Development
Infrastructure-related services involve aspects of acquiring, installing, maintaining, and replacing things on an annual basis. Whether the replacement costs are part of the annual operating budget or are to be viewed as deferred maintenance (as many colleges have done in repair of their building infrastructure), they are real costs related to providing the necessary IT services. As such, they must be considered as part of any analysis of the cost of providing these important services. Support services are those areas in which budget components are largely staff driven and relate to provision of support to users of the infrastructure. Ultimately, the COSTS project aims to study all of these services.
Early data analysis indicated it would be useful to have common hypotheses against which to test the data in each service area. One project participant, Richard Parker,3 suggested the notion of grandmother hypotheses; that is, simple hypotheses that even your grandmother would know to be true:
- Economies of scale--The unit cost of a service will decrease as the number of units increase.
- Outsourcing--The unit cost of providing services decreases with the degree to which they are outsourced.
- You get what you pay for--As service levels increase, the unit cost of providing the service increases.
- Complexity--The unit cost of a service increases with the complexity of the environment in which that service is provided.
We focused on one service at a time, developed a template for the data to be collected, and distributed it via e-mail to the group, with a short deadline for submitting data. Participants were promised quick turnaround and preliminary analysis in exchange for their efforts in collecting and providing data. As an incentive, and to demonstrate this could be done, we provided the data for our own institutions (Hamilton College and Colgate University) in the initial e-mail request for each service. The agreement was that institutions must submit data to receive any detailed results.
Results of data collection for the first three areas--network services, desktop repair services, and administrative information systems--are reported below, using the grandmother hypotheses as a framework for discussion.
Network services refers to those activities of maintaining the infrastructure (wiring, fiber, hubs, and so forth) and the servers on the campus network, including the connection to the Internet. Cost of the initial building wiring, fiber runs, or electronics are not included. We asked participants to include the cost of the main servers for the library (OPAC and databases) and those that provide general campus information (Web server, public lab servers), but not the cost of departmental servers that provide only information for their departments. The costs of licensing and maintaining application software (e.g., statistical packages, administrative information systems) were excluded, while costs related to maintenance of operating systems were included. Data elements that were requested for network services are listed in Table 2.
Table 2: Network services data elements
- FTE staff
- Budget components,
i.e., dollars budgeted for:
- student help
- contractual arrangements
- professional development
i.e., number of:
- active ports
- used ports
- Total replacement costs and replacement cycles for:
- network electronics
- Description of network and service levels
Sixty colleges and universities from the U. S. provided data.4 These institutions are heavily networked, including residence halls. Network environments combine shared, switched, and fast Ethernet segments in buildings, with most of the campuses using FDDI as the core backbone technology and a smaller number using ATM. On most campuses there is a mix of technologies, as networks are in the process of upgrading to increase speed and support emerging uses of multimedia technologies.
Descriptions of the network environments and institutional characteristics such as number of ports, computers, students, and employees helped provide a context for understanding the institution and the necessary divisors for calculating unit costs. See Table 3 for a summary of these characteristics for the responding institutions.Table 3: Network services institutional characteristics
25th percentile Median 75th percentile FTE Network Staff 1.7 2.5 3.5 Active Ports 800 2,000 2,672 Used Ports 650 1,210 2,000 FTE Students 1,388 1,896 3,610 Employees 347 500 763
Caveat: As with any survey, estimates often are the best that can be obtained, especially when describing staff who provide a number of services.
We distinguished between active network ports and used network ports. Active network ports are connected to hub electronics--a computer could be plugged into an active port and connect with the network without any further work in the wiring closet. Used network ports have an actual computer or printer connected to them. For institutions in the early stages of networking their residence halls, the number of active ports may be considerably higher than the number of used ports. The number of used ports is a better predictor of the effort required to provide network services. Our analyses focus on averages, trends, and ratios, rather than absolute numbers. For example, a summary of the costs per port appears in Table 4.Table 4: Network services annual unit costs
Cost per 25th percentile Median 75th percentile Active Port $148 $180 $254 Used Port $200 $255 $381
We account for replacement costs, even though most institutions have not built these costs into annual operating budgets. We asked each institution to provide the total replacement value of servers and electronics and an estimate of the replacement cycle it would use if replacement costs were built into the annual operating budget. As expected, there was variation in projected replacement cycles with ranges from two to seven years for servers and three to ten years for network electronics. To apply a common estimate for the replacement cycle of electronics and servers, we computed the average cycle times for the entire group and then applied the average cycles (four and five years, respectively) to the actual replacement values for each institution. This helped to factor replacement costs into all subsequent analyses, regardless of whether the college had the money allocated in the annual operating budget. Using this approach, replacement costs represent, on average, 35 percent of the costs of providing network services, with salaries and contractual expenses representing 28 percent and 14 percent of the annual operating costs, respectively (see Figure 1).
Figure 1: Network Services--Budget components
Economies of scale
There is a certain amount of overhead necessary to provide network services to a college campus, regardless of the size of that network. This is most commonly reflected in the number of staff in the network services area. Correspondingly, economies of scale should result in lower unit costs as more devices are connected to the network. The data from the sixty respondents support this hypothesis. While variations exist, lower unit costs appear to be achieved for used ports as the number of ports increases (see Figure 2).
Figure 2: Network Services--Cost per used port
Colleges rely on outside organizations to deliver some portion of their network services, the most frequent being hardware and software maintenance and Internet connections to and from the campus. How are the unit costs associated with providing these services affected by the reliance on outsourced network services? The data for the institutions participating in our study do not reveal any significant difference among the unit costs for outsourcing versus in-house approaches.
Service levels and network complexity
Most campuses described similar service levels for network services. Coverage for network outages was generally during business hours, with some personnel on call in the evenings. Most institutions had T-1 connections to the Internet.
Although responding institutions have a variety of network topologies, use a variety of equipment, and are supporting a diversity of protocols and hardware platforms, there is no discernible way to confidently relate these components of complexity with effects on the unit costs of providing network services. While it seems reasonable to assume that more complex networks would exhibit higher unit-service costs, it is also possible that increased complexity would be reflected in lower service levels.
Some further observations
Our preliminary analysis suggests that the full cost of network services is not yet reflected in the annual operating budget for IT organizations. Participating institutions are investing heavily in wired campuses, providing high-speed connections to the students in residence halls as well as throughout academic and administrative buildings. Early adopters of networking are in the process of upgrading their campus network infrastructures to support higher speeds. In general, the costs of upgrading the infrastructure will represent a significant portion of the ongoing cost of providing network services.
As instructional applications of technology become increasingly dependent on campus network and Internet-based resources, campus residents will expect their networks to function on a seven-day-a-week, twenty-four-hour-a-day basis. The implications for staff coverage, now generally provided on an informal call-in basis or outsource arrangements, can be substantial. The challenge for campuses will be to define service levels associated with various unit costs. At this point, the data collected do not provide substantial insight into the relationship between service-level expectation and unit cost.
Desktop repair services
Maintaining desktop computers and printers has become a regular part of providing IT services, and the ninety-three institutions that contributed data5 on this service area have been providing this service in two basic ways--using significant outsourcing or using their own staff. The data elements requested were basically the same as for network services, with computers substituted for ports. Institutions were instructed to report only the costs of providing repair of institutionally owned computers and printers and not to include costs or revenues related to repairing personal equipment or internally charging departments for repairs. Because this service focuses on maintaining existing desktop equipment, no replacement cycles were requested.
For all participating institutions, staffing levels are generally low, averaging 1.8 people per institution. On average, salaries, materials, and contractual arrangements make up 47 percent, 25 percent, and 17 percent of the annual budget, respectively, but these averages mask some significant variations among the institutions in terms of budget components. Annual unit costs per computer ranged from $53 (25th percentile) to $145 (75th percentile) with a median of $102.
Economies of scale
The data collected support the general conclusion that economies of scale result in lower unit costs per computer as the number of computers being serviced increases.
There are significant differences in the way the responding institutions provide desktop repair service. Forty-eight of the colleges use some minimal contractual services (less than 10 percent of the budget). On average, there is about a 15 percent cost difference between the group that outsources more than 10 percent and the group that outsources less than 10 percent, with the in-house group averaging $122 per computer and the outsource group averaging $108 per computer. The difference is more dramatic when considering those institutions that devote more than half of their budgets to contractual arrangements versus those that spend less than half on those services. The cost per computer for the outsourced group was less than half that of institutions that primarily provide the service using their own staff.
Service levels and complexity
It was difficult to relate the service level associated with various unit costs, as most institutions do not maintain detailed records on the time-to-repair function for desktop equipment. Most respondents said computers are generally repaired within a certain time period if parts are on hand and within a longer period if parts had to be ordered.
One recent study comparing technology support in educational institutions to that in business environments suggests, "When an educational PC fails, it simply gets taken out of service for several days. A business computer is usually repaired within a matter of hours. Therefore, downtime for educational computers is two to three times higher."6 This is more likely to be true for public and departmental computer labs than for equipment on faculty or staff desks.
While participating institutions report generally similar expectation levels for equipment repair, there could be considerable variability in meeting these expectations; however, no data submitted support this hypothesis. Institutions in the study generally support both Macintosh and Windows desktop environments and alluded to a range in age of equipment on campus. We were not able to confirm or deny the hypothesis that a less diverse hardware environment results in lower unit costs.
Some further observations
Desktop computer maintenance is an essential service in any technology environment, not only for institutionally owned equipment but also for the increasing number of computers being brought to campus by students. A number of institutions indicated they can generate enough income by providing repair services for students to subsidize the maintenance of institutional equipment. The potential for lowering net unit costs by providing repair services to students might be examined in more detail. Such a service is similar to generating revenue by providing long-distance telephone service to students.
As students, faculty, and staff become increasingly reliant on technology for teaching, research, learning, and operating the business of the college, expectations for service levels will rise. One participant summarized the expectations at his campus by noting, "The users call when they have any kind of problems and they expect immediate attention." Another indicated, "Our users have come to expect almost immediate response (during working hours) if they have a machine that is not working."
Further work needs to be done to study the relationship between unit costs and quality of service provided as measured by average downtime for equipment.
Administrative information systems
Centralized information systems continue to play an important role in transacting the business side of the college. Regardless of the technologies used, these systems provide fundamental information that supports decision-making by admissions, financial aid, registrar/student affairs, business office, and alumni/fundraising. Support services include systems analysis, programming, and operations. Twenty-nine institutions submitted data in this area.7 Not all of these institutions are supporting the same collection of applications; four of the institutions are using software developed in-house, while the others are using packages from vendors. The major components of the annual budget for this service area are salaries (63 percent) and software maintenance (17 percent).
Economies of scale
The basic functions performed by an administrative information system are similar across institutions. For most of the institutions, the total cost of providing these services falls in a narrow band, from $200,000–$500,000 (see Figure 3). Therefore, regardless of size, a certain amount of overhead is necessary to provide basic administrative support services.
Figure 3: Administrative information systems--Total cost
None of the institutions in our study contract with a company to manage their administrative information systems (commonly called facilities management). However, a common issue faced by all colleges is whether to develop software in-house or to license integrated packages from a vendor. Nearly all of the institutions responding in this area have chosen this second route. It is commonly believed that institutions choose this approach with the goal of increasing functionality, providing higher levels of support, or minimizing the need for technical support staff. Among the responding institutions, the average staff size is smaller for those that license integrated packages, but average unit costs are approximately the same for both groups (see Table 5).Table 5: Administrative information systems average staff and unit costs
Averages In-house License FTE Staff 8.5 4.7 Cost/User $126 $123
Service levels and complexity
Supporting integrated administrative systems can be a complex process. These systems are generally built on a database foundation and often require significant technical knowledge for effective use. It is likely, therefore, that the improved functionality that vendor-provided solutions promise (hence, higher service levels) comes with lower staffing needs, increased software licensing fees, and demanding hardware requirements. This appears to be reflected in the proportion of budgets devoted to these service components. We have not been able to determine whether the improved functionality is actually realized, or whether reduced staffing levels create a roadblock to achieving higher service levels.
Some further observations
The data template components for this area reflect the ongoing cost of operating these systems. Hence the initial, and often substantial, acquisition costs for hardware and software were not generally reflected in the data provided by the institution. In the future we intend to incorporate an amortization analysis for the servers, similar to that done for network services.
While the data template reflects centralized costs, it is often the case that the newer administrative systems require powerful desktop computers to utilize the full functionality of the software (e.g., access through the Web and use of sophisticated client/server architectures). The upgrade costs at the desktop level are not reflected in this template. As with the other two services, our analyses raised as many questions as they have answered. This is consistent with the goal of promoting better understanding of the complexity of these services.
Conclusions and future directions
A growing number of colleges and universities are participating in the COSTS project (including more complex and larger institutions), suggesting that with further participation it will be possible to develop benchmarks for unit costs that may be useful across institutions, even among a variety of institutional types. The simple analyses we have done provide valuable insight and point to some interesting avenues for further investigation. Relating unit costs to service-level expectations is particularly promising for understanding the differences exhibited by institutions and has the potential to help identify exemplars for further study.
Further progress depends on a substantial increase in the number and variety of institutions willing to make the effort to be introspective about their IT support services. The project goals are focused on developing a common understanding of campus IT support needs. In this era of the support services crisis, a better understanding of our options is essential.
1 “A Modest Proposal,” AAHESGIT listserv, 25 November 1996. This listserv has over 6,000 subscribers, including CIOs, faculty, campus technology administrators, and others interested in the impact of technology on teaching and learning. To subscribe, send e-mail to [email protected] (with the subject line left blank) containing the following message: SUBSCRIBE AAHESGIT yourfirstname yourlastname.
Back to the text
2 These are small, private, liberal arts institutions. Two-thirds of all institutions of higher education have fewer than 2,500 FTE students. Understanding services for this group can be of significant planning value.
Back to the text
3 Parker is Director of Academic Computing at Harvey Mudd College.
Back to the text
4 In addition, three Canadian institutions provided data; we do not include their data in this summary to avoid issues related to conversion of currencies.
Back to the text
5 As with network services, we provide data only for the U.S. institutions, which numbered eighty-four.
Back to the text
6 Understanding the Total Cost and Value of Integrating Technology in Schools, International Data Corporation, 1997. 7 We report data from twenty-eight responding U.S. institutions.
Back to the text
Karen Leach ([email protected]) is the chief information officer at Colgate University. David Smallen ([email protected]) is director of Information Technology Services at Hamilton College.