Performance Data Shows Variances Between and Within IT Service AreasBy Doran BoroskiIT costs are confusing. On the one hand, total IT spend has risen dramatically over time, reflecting unbridled business demand for resources. At the same time, on a unit cost basis, costs for IT services are on a steadily downward trend. In other words, the total volume of business use of IT has outstripped ITs ability to reduce the cost of delivery.Looking specifically at IT unit costs, things get even more complicated. Unit cost trends vary significantly across different IT service areas, as well as within individual categories, with costs increasing in some cases. As CIOs increasingly focus on aligning spending plans with business requirements, a detailed and specific understanding of the full range of IT cost and performance trends is essential to developing budgets that address business needs.Mainframe Computing: 20 percent annual decrease per MIPSOn a cost per MIPS basis, mainframe personnel costs are declining by 10 percent a year, driven by productivity increases. At 25 percent annually, the declining unit cost of hardware (CPUs and storage) per MIPS is the greatest contributor to overall lower costs. Most data centers are reducing their unit cost of software per MIPS at a rate of 12 percent to 15 percent annually, but this remains the most difficult category of cost to manage. Since installed MIPS are often increasing as fast or faster as software consumption, total software costs are rising even though unit costs decrease.Unix Server: 18 percent annual decrease per processorMeasured by overall costs per processor, Unix costs are decreasing by 18 percent annually, and are declining by 30 percent a year if a measure of processing speed is applied. This is a direct reflection of continuing hardware performance improvements at similar price points. Measured by logical server, costs are declining by only 15 percent per year. Breaking down these trends, personnel costs are declining by 20 percent, directly reflecting productivity improvements in support services. Hardware and software unit costs are declining at a slower rate, since servers are getting bigger as measured by number of processors. Logical partitioning has not yet been widely implemented but is a leading trend. This will have the largest effect on unit costs over the next two years as implementation of partitioning raises hardware utilization and eases image support effort.Wintel Server: 10 percent annual decrease per processorMeasured by overall cost per processor, costs are decreasing 10 percent annually while when measured on a per server basis, costs are decreasing at 6 percent annually. A key driver of this slowly decreasing trend is that servers are getting larger and more complex as mission-critical applications continue to be ported to this environment. Thus, although unit costs are decreasing, hardware sizes are growing, thereby partially mitigating the apparent advantage. Software costs per processor are decreasing more rapidly, as core software has tended to be priced on a per server basis. Database costs which are not included in this base server operating metric are an exception to the overall pricing trend. Database costs are changing rapidly as vendors adjust their pricing methodologies to reflect technology changes. Price per server or per processor is evolving to price per user, active user, or other measures. As such, previous pricing trends may not accurately predict future costs. In general, pricing methodologies are responding to market pressures and declining operational costs.As with Unix, logical partitioning has not yet been widely implemented in the Wintel environment, but will have a major impact on driving down unit costs over the next two years, as implementation of partitioning raises hardware utilization and eases image support effort. Several leading Compass clients have realized significant cost improvement from implementing tools like VMWare.Storage: 40 percent annual decrease per installed GBDisk storage costs continue to decline rapidly, a trend that is driving storage acquisition strategy. Storage costs can be tiered based on usage type. For example, high-availability transactional data may have a different performance and recovery profile than PC file storage. In these cases, an optimal storage mix might be high-cost disk for transactional data and low-cost, lower performance disk for PC file storage. Costs per utilized GB are declining at a lower rate than costs per installed GB, however, due to increasing redundancy being built into the storage farm and (partially as a result) relatively low utilization rates. Key strategies for cost containment for storage include just-in-time purchasing and implementation of storage management tools.Desktop: 5 percent annual decrease per clientOverall unit costs per PC are declining at a relatively slow rate, with rapidly decreasing hardware costs offset by slow progress on support issues. While decreasing slightly or staying constant, service costs vary widely, based on the degree of standardization of services and hardware found in a given installation. Desktop support processes are maturing, characterized by increased implementation of remote support tools and more stable Windows software environments (e.g., XP). The productivity gains these initiatives yield for desktop support staff, however, are being offset by higher costs for salaries and process improvement implementation. Frameworks like ITIL are becoming popular as organizations seek to structure and integrate support processes, but most organizations are still in the planning and implementation phases. As such, expected productivity and quality benefits will only be realized once processes mature and are fully implemented. This has yet to happen in most organizations. Moreover, organizations should not necessarily expect drastic cost savings when these benefits emerge in several years. Rather than use productivity gains to cut people costs, many organizations use gains in one area to redeploy people from reactive non-value-add activities to more proactive work that provides more value to the business. These areas include planning, design, and engineering work, as well as proactive problem management. So, while investments in process improvement in one area will not necessarily result in lower overall unit costs, they can increase the efficiency of select process areas (most often those of a reactive nature). Moreover, process improvement efforts tend to improve the quality of services, thereby providing more value to the business.Generally speaking, hardware costs per PC are declining, but organizational refresh strategies are having a significant impact on overall costs. Some mature, cost-focused organizations are lengthening retention plans to four or even five years, but many more organizations positioning for the future have settled on a three-year refresh cycle. Software costs are generally increasing. While the cost of office applications and operating systems has remained relatively constant, organizations are investing more and more in support tools that complement new support processes. Service Desk: 15 percent annual decrease per contactPersonnel is the single largest cost driver in this area. Although stable workforces cost more year over year per person, productivity gains are offsetting these increased costs in high-performing organizations. Costs trends will continue to decrease as Service Desks push support to multi-channels such as email, instant messaging, FAQ, and knowledge bases to lower the direct interactions with agents. This increases the number of contacts but lowers the per contact cost.As knowledge tools become more sophisticated, agents will handle a higher number of contacts per hour. Alternative sourcing models, such as near-shore or off-shore migrations, are also significantly lowering per contact costs and will be a major trend for larger call centers in the next few years.Understanding Cost DriversIT unit cost trends continue downward. Within these overall trends, there are specific cost drivers such as hardware acquisition costs or personnel productivity that disproportionately impact these trends.Developing project plans based solely on overall cost trends may led to misleading cost predictions and miss improvement opportunities. Top performers optimize cost drivers to gain the most from change.Understanding internal business requirements and technology industry trends is key to using models to predict cost changes. These models can help IT organizations develop change plans while understanding the impact on costs. Modeling industry trends and company workloads at granular levels helps predict overall costs, define personnel staffing needs, and identify possible bottlenecks. Leveraging specific, detailed, cost changes will provide the biggest impact.Doran Boroski is a Compass executive consultant based in Naperville, IL.