TCO: How does the total cost of ownership (TCO) of PC-based desktops and Apple Mac's compare

Table 2. Costs per PC per year ($).

ASK COMPUTERWIRE REFERENCE MATERIAL

Computer Finance - 04/01/99
Servicing costs of the managed desktop
What is the value-add of managed desktop services based on total cost of ownership (TCO) analysis methodologies? This report considers the current thinking in the light of commercial managed desktop outsourcing offerings.

By now, Gartner Group's total cost of ownership figures on conventional, LAN-connected PCs showing annual costs of nearly $12,000 is old news. Significantly, it does not seem to have budged much as PC purchase prices have declined; the former $2,500 customary investment for a business desktop is now at least $1,000 cheaper.

So why haven't the TCO numbers changed? The stock explanation is that the cost of ongoing support has not yet declined for the 'unmanaged' PC because the degree of complexity in varying machine configurations, OS version incompatibilities, and software distribution hasn't changed. Although Microsoft and Intel are phasing in measures for remote control of machine configurations (e.g., Wired for Management, Zero Administration Windows), the take-up has been slow.

About 18 months ago, Gartner acquired Interpose, a developer of a cost tracking tool, which has since been rebranded TCO Analyst. Gartner has since stated that its TCO Analyst product is indeed a tool designed for vendor sales teams. So what use is this to users? We studied the experiences of two users who took advantage of the tool through a third-party consulting service.

Managing By TCO Numbers
Managing complexity is part of the 'soft costs' that, critics claim, is a gray area. These costs are not on the list price of any software or equipment, but are to be found in the intangible costs of LAN management and formal or informal technical support that is usually buried under overhead, if counted at all.

TCO Manager is a comparison, forecasting and management tool that provides information to IT and financial management for both day-to-day management and IT procurement decision making. It factors in some of those 'soft' costs. The software collects asset, cost, service-level and IS performance data that is then used to calculate the TCO of the organization's IT infrastructure. The organization's actual costs are then compared to organizations of a similar size, type, geographical location and technology.

Table 1 details potential savings, as measured by Gartner Group, on the cost of desktops as a percentage of the TCO. Some of these estimates are frankly meaningless. For instance, that '90% of the power of PCs not used' - just how is this assessed? However, based upon these figures Gartner Group believes that a massive 40% - 50% saving on a typical desktop is achievable with the right mix of technology and management.

Table 1. Potential desktop cost savings
(Source: Gartner Group).

Many of the areas of potential savings highlighted represent best or simply good practice. And there is of course no reason why users can not begin to reduce the cost of the desktop independently of any TCO methodology. Adopting good practice for instance can reduce costs in the following areas:

* Asset management: No CIO can rein in costs without having an extensive knowledge of the desktop asset base.

* Standardization: This lessens the risk of software clashes and allows the use of more cost-effective desktop tools.

* Use of desktop management tools: This is the big-ticket item in PC TCO. These tools include products that help increase efficiency and automate tasks associated with PC audit, PC maintenance and helpdesk support.

* Lack of control or poor supervision of policies: For instance, many organizations fail to police downloading software from the Internet. This may result in under performing desktops or may even act as a conduit for viruses.

* Centralized purchasing: In recent years there has been a trend towards allowing individual business units to hold responsibility for the procurement of hardware or software. These purchases have then to be maintained by IT. Often IT finds itself maintaining technologies that it is ill equipped to manage.

The Infracare Prescription
ICL's managed desktop service, dubbed Infracare, begins with an initial TCO assessment. This assessment is carried out by a senior ICL consultant and involves an audit of desktop cost data from around the organization. The resultant findings are presented to the client together with industry average benchmarks and recommendations on the development and improvement of the IT infrastructure and support service, that promises to reduce the TCO of the desktop.

The second stage involves negotiations over an approach for implementing the recommendations. Generally three options are presented. The first of these has ICL working with the client to prepare the business case for investments that address some or all of the recommendations made. The client maintains the flexibility to prescribe which tools, technologies and resources are to be used. The ICL consultant will develop investment plans for the initiative, and with the assistance of TCO Manager, provide ROI analysis.

The second option provides the client with priced options for the delivery of specific systems and service solutions, which address one or a number of recommendations.

The third option proposes terms for the complete outsourcing of IT support with a commitment to reduce TCO to an agreed figure over the contract period. In this option ICL will design and implement services tailored to the requirements. In consultation with the client, ICL will then undertake subsequent assessments to measure progress in terms of TCO reductions over time and to identify further target areas for improvement.

In meeting its commitment to reducing the TCO, the ICL methodology depends upon the following core strategies:

* Standardization with a minimum of six PC configurations.

* Full asset management using software from US vendor Asset Pro.

* Remote management of software and network with rigorous change control.

* Pro-active incident reduction through improved training and preventative intervention.

Infracare will not implement these systems/practices overnight, rather they will approach improvement step-by-step. Phil Murray, Director of Infracare, says that typically they will commit to a reduction of between 15% and 25% of TCO within three years. Murray points out that few finance departments are, as yet, willing to include reductions in soft measures as part of the package. However, he expects this to change in the near future as confidence in the TCO approach grows.

User Experiences
UK utility company: This organization with around 6,000 PCs underwent an initial six-week assessment. ICL analyzed all the costs associated with running the company's desktop environment and calculated the TCO. It also observed the company's IT service provision and delivery mechanisms. Finally, ICL delivered recommendations as to how the TCO could be reduced to the organization's directors.

In the first instance, details of the utility company's size, geographical location and technological complexity were fed into the TCO Manager software. This provided the ICL consultant with a best match - or 'typical' sample - from the TCO Manager database. This means the costs associated with any 'what if?' scenarios relate directly to comparable organizations' situation-scenarios generated by subjective assessments. Obviously, the data generated by TCO Manager provided only a general picture that was largely based on qualitative observations.

The assessment excluded the costs of developing business applications - with the exception of mail server systems. ICL argued that although a significant amount of PC usage involved accessing and using mainframe and mid-range applications, large variations in application development costs made it difficult to devise a benchmark. Anyway, we don't believe that development costs are relevant to desktop TCO; what's more important is deployment and maintenance.
Table 2 shows the feedback following the Infracare initial assessment. Typically direct costs are budgeted costs of the IT groups within an organization plus IT assets and software owned by the business.

Table 2. Initial assessment of utility company desktop TCO
(Source: ICL Infracare).

In comparison with other organizations, the company showed higher costs of support. This was attributed to costly, labor intensive processes and service delivery, cultural demands for large on-site support resources and a failure to leverage the efficiencies of service-delivery technology.

The hardware depreciation costs were lower than for typical target organizations because no asset register existed. Lacking such data, we believe that the analysis is useful to track internal costs, but would lose relevance if any attempt were made to compare costs with those in other organizations.

The study also analyzed indirect costs. In most cost accounting systems, indirect costs are those that are delineated separated and charged to specific projects or groups. They are usually lumped under 'overhead' or 'burden rates.'

In this analysis, indirect costs were otherwise unattributed costs resulting from end users delivering informal technical support that was not formally charged to any budget, and the general loss of downtime that, again, does not wind up in the budget.

In this company, indirect costs proved considerably lower than for comparable organizations. That was because, according to end user survey feedback, PCs were not considered essential tools. Therefore, little time was lost because relatively little activity of any value was performed on PCs. Given those attitudes, the company's significant investment in high-end PC hardware and comprehensive software tools could be questioned.
Based on the initial assessment, ICL Infracare delivered the following recommendations:

1. Establish a central help desk: Existing level one support was inefficient, primarily serving as a means for dispatching technicians to the user's machine. The assessment team recommended the creation of a formal corporate help desk that would strive to answer more support questions over the phone on the first call, and proactively coordinating the efforts of problem resolution teams.

2. Improve training for IS and end-user staff: Through better knowledge, end users would become more self-reliant, solving their problems more quickly. And if technical support staff was necessary to solve a problem, improving their knowledge would deliver similar results. (Note: ICL has found that company directors are generally very resistant to the idea of increasing training budgets because of claims that the benefits cannot easily be measured. For instance, it could be difficult to attribute reductions in downtime to improved computer literacy.)

3. Create a comprehensive asset and software register and management platform: The assessment team recommended the establishment of a single hardware asset and software register to include identification, configuration, financial and TCO-based event data. (Gartner Group predicts that call times can be reduced by 30% if help desk staff has immediate access to end user system configurations.) Obviously, success would depend on how well hardware and software configuration updates could be automated.

4. Pro-active capacity planning: ICL believed that this would make the company better placed to make smarter and more timely decisions regarding IT procurements, deployments and disposals. Such measures would include regular reviews of server and client capabilities in terms of storage space, processing power and memory, network (LAN and WAN) traffic management etc.

5. Establish a centralized, locked down client environment: Standardizing configurations and enforcing them through lock-down measures helps simplify support burdens. But a lot of pieces must come together for this to happen -including careful capacity planning, meticulous shakedowns of proposed hardware and software configurations, and moves to take full advantage of features being introduced by Wintel.

6. Consolidate file servers: Although the company was in the process of migrating off legacy platforms, it had not considered this idea. ICL suggested significant areas of rationalization.

7. Streamline procurement processes, using workflow technology: The procurement process encompassing all activities, from a user's first request for information through to the successful commissioning of the equipment and/or software, should be automated using predefined hardware and software choices (a byproduct of enforcing standard configurations), forms-based automated workflows, electronic order placement, and effective interfaces between purchasing and related applications such as training, asset and change management, support teams (helpdesk), financial accounting, budgeting and capacity planning.

8. Mandate client data security: Through cultural and technology, mandate storage and backup of data on central servers.

Naturally, ICL recommended that TCO assessments not be a one time thing. Instead, it should be used to track improvements. Ideally, the organization should be able to undertake this on its own, with modest outside consulting.

UK finance company: This organization is one of the UK's largest banking groups. Prior to the initial assessment, ICL conducted a workshop to set out the terms of reference for the project. It was agreed that by measuring existing desktop TCO, it should be possible to quantifiably demonstrate that investments in new technology would either reduce TCO or generate improved returns on IT investment through greater user productivity. With this focus, soft costs such as costs of use, training and lost productivity are clearly on the menu.

The initial TCO analysis was restricted to a specific section of the company. This covered four individual businesses with a total of 2,053 end-users. The costs set out here were analyzed as part of the assessment project and included hardware assets, software assets, internal IT staff, and end user metrics.

1. Hardware assets: There was some disparity between the number of client systems recorded on the client's inventory database and the numbers of users recorded. It is recognized that some users have the use of two PCs for valid business reasons. There was also a significant disparity between the number of assets recorded on the fixed asset register and inventory database. The fixed asset register was about 60% accurate while the inventory database was 80%+ accurate, based on data that was eight months old. To obtain the cost of client system asset depreciation taken by the businesses over the past 12 months, ICL sampled the fixed asset database, and then applied the volumes recorded on the inventory database. This was thought to be the only way of obtaining this figure, although with an estimated 20% margin of error. All IT consumable costs were estimated based on reported usage, industry standard and manufacturer guideline costs.

The total IT purchasing data for the preceding 7-month period was annualized to obtain the total expensed purchases for client and server systems, as follows:

* The estimated value of new asset purchases and consumables were subtracted.

* The costs of software upgrade purchases were identified from license tracking information. This figure was also subtracted with the remainder being apportioned across client and server systems as the cost of upgrades.

2. Software assets: In order to obtain a view of the annual software spend, the annualized number of installed new PCs was multiplied by the cost of the standard software build. This assumed that there was therefore no redeployment of licenses from old systems.

3. Internal IT staff: All staff were counted except those involved in Y2K or external business system work, because they were not available to respond to typical end user problems. (For the sake of the assessment, they were instead counted as end users.) The results fed a calculation of the ratio of IT staff to number of supported desktops.

4. User data: The average burden rate, number of hours worked and average salary was obtained from HR. The average salary of IT staff was also calculated, as this was higher than the average user figure.

A total of 220 surveys were received from users across the businesses being assessed, equating to just over 10%.

Table 3 shows the TCO results of the assessment. The typical figure refers to an industry typical figure for TCO based on the company's size and complexity. The actual figure refers to the actual costs accumulated during the assessment.

Table 3. Initial assessment of finance company desktop TCO
(Source: ICL Infracare).

From this data the ICL consultant made a number of observations and suggestions of programs to reduce TCO. Once again the establishment of an asset management system was strongly advised as was the improvement of the organization's training provision. Following is a summary of the suggestions:

* Redefine help desk processes: Consolidate access to relevant personnel and asset databases onto a single system accessed by level one support. It would provide rapid call logging and diagnosis, aided by a help desk tool linked to asset, personnel and knowledge databases. This should allow higher level staff to be more effectively used as second and third line support.

* Redefine IT procurement processes: Use EDI or similar methods of supplier communications, on-line catalogs, and order management systems.

* Define a standard 32-bit desktop configurations: This would involve a balanced integration of Pentium PCs, Windows thin clients, and browser access. This approach would deliver extended central management, resilience, and diagnostic capabilities, which should show a positive impact on TCO and support any automated asset management initiatives.

* Specify WAN support contract goals: Delineate availability service levels based on downtime, response time, and performance. Although performance with the current IT environment was deemed adequate, introduction of new technologies such as videoconferencing or WTS (Windows Terminal Server), and the impact of server consolidation, would lead to a significantly increased demand for network bandwidth.

* Perform network impact assessment: Validate design requirements of future infrastructure strategies, and adjust WAN service contracts as necessary.

Beyond the desktop service needs, the analysis also recommended consolidating development resources as well. While there might be merits to this, given tight labor markets, we believe that such factors were outside the scope of the assignment - lowering the desktop TCO.

Usefulness of the Results
Although it often relies on subjective criteria, using a tool such as TCO Manager can help an organization paint a picture of how effectively it is managing its desktops.

Regardless of the accuracy of an initial assessment, there is some obvious business value in conducting follow-ups to get a relative picture of how well the cost picture is improving - or not.

Both companies studied here had plenty of headroom for improvement, since they didn't conduct proactive asset management prior to the studies. In the case of the first company, we would question either:

1. The belief that PCs were of minor importance to business processes.

2. And if they actually were of minor importance, why then was there a need to mount such an investment if the returns would be trivial to business effectiveness.

Admittedly, much of what TCO Manager can be used to recommend is common sense, like asset management, configuration control, etc. It only provides an arbitrary set of numbers to prop up a business case. But in some cases, having some numbers may be better than having none at all.

Computer Finance - 03/01/98
Desktop Asset Management Tool Payoff
With PC support costs seen as the black hole in the IT budget, more companies are turning to asset management solutions to reduce ballooning administrative costs and are realizing a quick return on investment in the process.

High LAN/desktop support costs have plagued IT departments for years and for many organizations, the challenge of reducing the LAN/desktop support budget has become a top priority. LAN/desktop management concerns itself with the management of individual clients on the LAN rather than the management of the network itself.

The problem is especially acute for small-midsized organizations; although their management challenges are clearly not as complex as the Global 2000, they usually lack the resources to gain effective control over their assets.
We surveyed six midsize organizations with desktop populations of from 300 to 2,500 PCs to assess the contribution of PC/LAN asset management and related tools.

The Job Remains Labor-Intensive
With corporations placing a greater reliance on end-user access to information, applications and enterprise business systems, IT managers are forced to throw more resources into LAN/desktop support and management to deliver required service levels.

That support and management effort means a fatter support staff, or more IT technicians for help desk, trouble shooting, repairs and general care of the LAN/desktop environment.

In other words, IT managers have and continue to throw IT personnel at the problem of keeping the LAN/desktop environment delivering the required functionality to keep the corporation humming. Meta Group estimates that the cost of operating a distributed LAN/desktop environment for 2,500 users is $11,900 per node/per year based on a five-year total cost of ownership (TCO) model. Labor costs, consisting of technical support, administrative support and end-user operations accounts for approximately $9,400 per node per year.

To make the IT challenge of managing the LAN/desktop even greater is the fact that this task is growing more complex rather than less complex, given an increasingly mobile end- user community.While IT staff has enough on its plate managing fixed hardware and software assets, troubleshooting remote or mobile users or doing software updates in an efficient and timely manner is increasing the management burden.

The Role of Tools: A Summary
Meta Group recommends that organizations investigate the following solutions for taming desktop costs:

* Inventory/autodiscovery agents: Understanding inventory is the first step to controlling costs, and it is essential when planning upgrades. Ideally, the tool should use agent-like autodiscovery methods. Meta Group's estimates that volume purchasing of desktop inventory asset tracking tools range from $30-50/desktop.

* Performance Monitoring/Analysis Tools: These reside on the end-user's client, providing continuous, real-time analysis of software/hardware interactions. They are useful for determining end-user efficiency (e.g., showing what times of day a machine is used, which applications are used, how much time is spent waiting for CPU cycles, etc.). Such tools can help plan equipment upgrade needs. Meta estimates that these tools cost about $70-$100 per desktop, and $2,000-$3,000 per server (although $150,000-$250,000 site licensing is common).

* Planning Database And Analysis: Using data fed by inventory agents and monitoring tools, these database planning tools can be used for providing the big picture on enterprisewide desktop upgrade needs. These tools are expensive however ($250,000 and up).

According to Tally Systems, a provider of asset management integration tools, corporate managers express several key areas of interest for LAN/desktop management:

* Inventory (what hardware and software resides on each PC).

* Electric software distribution (ESD), or the ability to upgrade or install software globally.

* Software metering (determining actual software usage).

* Software acquisition (purchasing only those software licenses that are needed).

* Hardware acquisition (effective deployment of all PCs, including older models.

In reality, LAN/desktop asset management sits at the heart of several network concerns: network management, financial management and help desk.

What this means is that LAN/desktop management tools offer IT managers a way to more efficiently address each one of the three key network concerns.

Vendors tend to take a suite approach to asset management offering solutions that address inventory, electronic software distribution, software licensing and remote control software. Software averages $20 to $40 per desktop.

Some providers, like Tally, provide links to other vendors' system management products, and help desk, so IT managers can access functionality outside the realm of the company's Cenergy solution, which includes NetCensus, for inventory, CentaMeter, for software usage management and WinInstall for electronic software distribution.

In a study done by Microsoft, which looked at 10 companies, with installations ranging in size from 153 to 14,625 desktops, who implemented its Systems Management Server (SMS) the savings achieved, using SMS to perform hardware and software inventory, software distribution and remote diagnostic services, averaged $1,200 per desktop, annually. The vendor did not factor in 'soft' cost benefits such as less downtime, increased user productivity, or other intangibles, although it noted that these benefits are considerable.

The average initial investment required to plan, pilot and deploy SMS was $142 per desktop. In terms of ROI, average payback time was 14 months. The fastest payback was measured in eight months for an initial installation of 2,500 desktops.

Getting started with Assets
Without knowing a company's desktop assets, or what you have and what you use, businesses tend to overspend on LAN/desktop management, whether that means supporting software applications that no one uses, over buying software licenses, or repairing PCs with a history of trouble tickets.In a mass corporate effort to do more with less, management is looking for accountability. For IT managers that means knowing what's on the desktops and the cost of installing software, etc.Yet, according to Mark Tecca, vice president of North American sales for Tally, industry reports indicate that only one-fifth of all corporations have done a baseline inventory of their LAN/desktop environment.

That means that four-fifths have never looked at their PC assets. 'At this stage of the game, asset management is the low man on the totem pole, usually spurred as some kind of crisis intervention,' Tecca said. Some examples of situations that spur organizations to implement asset management tools include:

* Major software upgrade of the desktops.

* Year 2000 urgency.

* IT budgeting.

* Software licensing.

Whatever the impetus for management to bring in asset management tools, IT managers agree that some one must take responsibility for the software.
In fact, industry participants agree that asset management projects fail without the proper commitment for resources to get the project going and to keep it going. Essentially, asset management has to become a way of life.

Companies like U-Haul International, Phoenix, Ariz., used outside consultants, one time a year for three years to do a manual audit of the firms desktop assets at the cost of $6,000 to $7,000 a pop. 'By year end the information on the company's 1,200 PCs was out of date,' said Mark McCardle, supervisor for PC support.

Today, after spending approximately $11,000 for an asset management solution, the company can cost justify the purchase simply by eliminating the need to hire outside consultants for a couple of years. Furthermore, IT management is more up to date on its asset reporting with inventory automatically updated every 60 days. MIS owns the asset management project with one person spending about four to five hours a month getting reports and applying updates.

Fortis Inc., a Minneapolis, Minn.-based international financial group, had a different initial experience with its asset management project. According to Joe Hays, PC LAN project analyst, upper management decided several years ago that it was time to know what PC assets the company owned. 'The company didn't know what assets it had, where they were located and software configurations were all over the board,' he said.

While the initial asset management inventory went well, the project faded because there was no on-going project champion to carry it forward. 'It was definitely a loss. No one lost any sleep but in some cases IT or upper management didn't have the information if needed when it wanted it,' said Hays.For the people who needed the asset inventory information, there was a time loss and productivity loss, the value of which is undetermined.

For the past few years, since claiming ownership for its asset management solution, Fortis has benefited greatly, seeing the initial, approximately $45,000 for 5,000 desktops, pay itself back multiple times over.

Getting started with asset management involves up front costs, to plan, pilot and deploy the project. Tally estimates that hiring a software consulting service to get the project going can range $70 to $80 per desktop, a figure that goes up or down depending on the timeframe allotted for the project.In its investigation of ROI on Microsoft SMS, the company calculated that an enterprise roll out of SMS averages $142 per desktop, including labor, training, consulting fees and initial hardware and software costs.

Inventory Management Register
Talk to most IT managers and they have little or no idea of what PC assets the company owns. This lack of information becomes evident when disaster strikes, resources need to be replaced, or software needs to be accounted for.

Some attributes in an asset management tool include:

* Platform specific information.

* How the asset is networked to other hardware or software.

* Information regarding vendor support and contracts.

* Financial information.

* Location of the asset.

* Responsibility for the asset.

Keeping pace with desktop assets pays back in more ways than one, according to vendors:

* It enables IT managers to plan hardware and software upgrades.

* It enables IT managers to uncover hidden assets.

* It enables businesses to fulfill software license registration requirements.

* It allows IT departments to control internally developed software.

At Briggs & Morgan LLP, a St. Paul, Minneapolis-based law firm, with about 200 desktops and 100 notebook computers, the IT department just completed a physical inventory of the organization's PC assets, using Support Magic from Magic Solutions. The driver was the firm's IT director who wanted to upgrade the company's laptop computers but had no idea how many they owned.

Wayne Yarmoska, IT director, said the software is modular with applications for internal help desk, inventory, purchasing, warranty tracking, on-line vendor information, service agreements, etc. The company recently paid $35,000 to upgrade its asset management software and add an additional 12 users to track a total of 300 PCs, 150 on-line printers and track 20 networked servers. The law firm uses the inventory information to plan purchases, assist the help desk and distribute software. One IT project recently completed was to identify remote dial-in software for license compliancy. The software is used on PCs used from home. It required technicians physically visiting about 100 users and approximately 120 man hours over a six month period because up until recently, the firm didn't have a clue as to how many software packages were being used.

Paying a technician about $15/hour, Briggs & Morgan shelled out $1,800 to account for the $79 software. Today, the inventory database is used to help IT identify who is running what software, who requires a software upgrade, etc. The firm hired an entry level clerk, at less than $20,000/year to handle all requests.

In the past, three or four technicians would have to pull records or look up original invoices to track down software information. These professionals included two application support specialists and two network administrators, whose salaries ranged from $35,000 to low $50,000.

Hiring the database clerk frees up the higher level IT professionals to spend more time focusing on their areas of expertise.
At Fortis, Hays estimates that each physical desktop analysis cost about $30,000 in labor. That figure is based on paying a temporary technician $15/hour to visit each of 1,300 desktops for about 1.5 hours per desktop.

Having the inventory information has saved Fortis the $30,000 multiple times, as the financial company has used it for several software projects.

The company has also used the inventory information data base to track down licenses for internally developed software, as well as using it for its Year 2000 project which required knowing every piece of hardware and application software within the organization.

Smaller installations also find big savings using inventory tools. At Firestone Textiles, Woodstock, Ontario, Glen Farrell, senior LAN technologist, is responsible for 90 nodes that run 24 hours a day. The company, has been using Intel's LANDesk Manager asset system to take automatic inventory for more than six years.

Not only does the inventory database save the company time but it also saves it money, only buying the number of software packages it needs when upgrading the desktops or installing new software. According to Farrell, he's able to run a query on desktop hardware/software information and get results in less than five minutes. He compares that with having a technician visit 90 desktops to get the same hardware and software configuration information.

Even if the technician spent 15 minutes per machine, that adds up to 22.5 hours, or almost three days in labor at $15/hour. Farrell estimates that he spent $1,300 for the initial software.

Software Metering Matters
According to Tally, software metering, a core component of asset management, has developed from a simple license compliance tool to an enterprise-wide usage information repository.

While metering is still a key part of honoring license agreements, the reporting aspect of metering provides IT managers with key information for decision support, usage and planning. For cost savings, metering helps IT managers track overspending on software by identifying underutilization.

Dixie Regional Medical Center, which supports three hospitals and several clinics in the St. George, Utah area, uses Intel's suite of desktop management products for 450 desktops.

The product enables the company to track software usage, restrict software usage, prevent unauthorized software from being installed and keep track of employees who play computer games.

According to Lance Bedingfield, systems engineer at Dixie, cost savings from metering are substantial. The medical center had a site license for 80 copies of WordPerfect. Software metering revealed that at any one time only 15 people were utilizing the software, meaning that the company was vastly overspending on its license.

This and other incidences of overspending has saved Dixie thousands of dollars.

Software Distribution Savings
Despite the need to move to new applications quickly, actual software roll outs, which are often done manually, can be time consuming and inefficient. The biggest cost factor in manual software distribution is labor.

According to a Microsoft study of a Windows 95 roll out to 1,000 users, a manual roll out including preparation and deployment took 125 person-days while an automated roll out using SMS took 27 person-days, a savings of 98 person days.

Dialogic Corp., a software developer in Parsippany, N.J., and Network Associate customer with 1,200 supported desktops, has seen savings from automatic software distribution several times over given the high number of requests to install or update software.

Dialogic initially invested $120,000 to hire a consultant to install the initial asset management software for the 1,200 desktops.

Here are two examples of how Dialogics saves money with automatic software distribution: The company recently rolled out an Microsoft Office 97 upgrade to 800 desktops. Chris Tokazowski, help desk system analyst estimates that a manual installation would have taken two to three temporary technicians about eight months.

Based on an annual salary of $35,000/each, one technician's time costs about $2,900/month. Eight months salary runs to more than $23,000. Using two or three technicians to do the roll out would have cost Dialogics well over $50,000.

Using the automated software distribution tool, it took three weeks to complete the software upgrade and involved no outside help.

Another incident at Dialogics involved installing Microsoft's Service Pack III on 300 desktops. Tokazowski estimates that done manually, it would take two technicians about 1.5 months to weed out which PCs required the software and to do the installation.

Based on an annual salary of $35,000, six weeks of a technician's time costs approximately $4,000. Two technicians run to more than $8,000. An automatic rollout was done in a matter of hours.Automated software distribution is also done at Dixie Regional Medical Center. The organization estimates that manually rolling out software to its 450 desktops would take four technicians upwards of one months time working five to six hours a day.

Paying a technician between $10 to $14/hour would cost the company over $8,000 for a month worth of labor. Using automated distribution tools, a software roll out takes several days.

Help Desk and Troubleshooting
A great benefit to organizations who deploy asset management software is the link between the inventory function and help desk.

Having desktop configuration information as well as a history of the desktop asset at the tip of their fingers, support personnel have instant access to data that they would otherwise collect during a manual visit to the user.

Products that provide remote control capability allow support technicians to observe what a user is doing and, if necessary, take control of the user's desktop to help diagnose and fix a problem - again, eliminating the need for a visit to a PC to make the fix.

Labor savings is the key benefit of this type of linkage. In fact, some industry analysts estimate that reactive technician dispatching constitutes between 30-40% of the total ongoing labor costs associated with supporting end-user environments.
Additional savings are realized by lowering the cost of the duration of support and reducing the number of calls placed to the help desk.

The second saving is the result of proactive management which is made possible by having desktop information on hand.
Dixie Regional Medical Center is a case in point about the potential labor savings associated with asset management software and help desk.

Prior to linking help desk tools to a desktop inventory database, IT need six technicians to support 200 machines and 600 users. Since creating the linkage, the organization has reduced support staff by two and now supports over 400 machines and 800 users. That reduction in staff represents a saving of more than $70,000 annually.

The initial impetus to purchase asset management software was to improve troubleshooting and help desk. Dixie also uses remote control software to enable technicians to respond to problems without having to leave their desks to fix the problems. Fortis also reaps benefits from linking its help desk software to its asset management solution.

According to Hays, the software has been linked for one year and he estimates that the company has saved the cost of one technician during that time.

How? Running about 700 trouble tickets per month, technicians save about 12 minutes, on average, per call by having desktop information at their fingertips. Paying a technician $20/hour, the company saves $2,800/month or $33,600/year.

In addition to saving technician time, there are end-user time savings as well. Every time a technician doesn't have to bother the end-user, that employee can be much more productive. Hays also noted that in some cases, having the desktop information on hand, allows a problem to be resolved without ever having to visit the desktop.

At Briggs & Morgan, a law firm that bills clients up to $235/hour, end-user savings add up fast when PC problems are quickly resolved.

Help desk linkage enables technicians to see trends in calls, so reoccurring problems can be quickly resolved.

The log of information on each desktop also enables the firm's two technicians to be on the same page when responding to calls.

By eliminating what Yarmoska calls triage, this helps control staff costs by shaving 30 to 40 minutes off of each and every call.
If, according to conservative estimates, the help desk gets 20 trouble tickets per week and saves 40 minutes per call, that's a savings of about 13.3 hours per week. For a technician that earns $38,000/year, the firm saves about $243/week or $12,636/year.

Most firms report that any time saved by technicians is time that these IT professionals can spend working on other projects or doing proactive network management.

U-Haul reported that help desk linkage has stabilized help desk hiring. In fact, by reducing desktop visits by 20 percent, with each call averaging half-an-hour, the firm has saved $3,900/month or $46,800/year, more than the cost of one technician.
The company generates about 650 trouble tickets per week and has about nine help desk technicians to attend to the company's 1,200 PCs.

Freeing up second level help desk technicians, who are also responsible for software installation and other proactive management, gives them more time to be proactive in maintaining the network for end-users.

While the main benefits realized from asset management at Dialogics is for automated software distribution, the company also realizes benefits from help desk linkage.

In fact, asset management software has enabled the company to eliminate three full-time temporary technicians and three part-time technical positions at an annual savings of $130,000/year.

Other Benefits Appear
It appears that organizations buy asset management tools looking for one or two specific functions and overtime, explore the additional functionality offered.

As one IT manager put it, the product keeps reinventing itself.

At Briggs & Morgan LLP, the organization uses its asset management software for facilities tracking, ie. desks, chairs, telephones, etc., financials, i.e. depreciation, etc., as well as IT desktop management. Other examples of complementary asset information include:

* Completed and planned maintenance.

* Lease information.

* Contract information.

* Acquisition value.

* Associated costs.

* Warranty information.

* User information.

Table 1. Summary of user benefits - all labor figures (hourly rates, salaries) are not burdened.

Computer Finance - 02/01/98
Cost Cutting With Standard Desktops
Most desktop TCO models are undergoing some sort of re-evaluation at present. A true guide to desktop cost effectiveness might be the 'running cost' of systems - and a move to a standardized desktop perhaps the surest way of lowering operating expenses.

'Anyone with half an eye on support costs just has to do it' insists Cap Gemini's Andy Mulholland, talking about the growing trend towards standardization of the desktop PC. In the UK, British Steel's massive $640 million contract with Cap Gemini to outsource 7,000 PC systems is centered around this very notion of the standardized desktop.

As well as moving toward a common system configuration that should make management of its desktop systems easier and cheaper, the British Steel deal is dependent upon the successful introduction of industry standard desktop browser technology for the access of corporate data. Its supplier is confident the adoption of this form of standardized environment will reduce license requirements and 'could save as much as 50% on the software license costs of the desktop' Mulholland told us. Accessing data held in Oracle Financials via the Web, for instance, is expected not only to reduce the complexity of operations for British Steel staff but should have a knock-on affect in helping push down support costs and lowering training bills.

The move to a common desktop environment may appear commonsensical, but not everyone is confident about the best route to take.

Bill Smith, director of information technology and planning at Baltimore Gas & Electric has outsourced the procurement and maintenance of 6,000 desktops, but he claims there is nothing currently available to help him assess the return-on-investment from any would-be standardized environment.

For other users like Edgar Sosa, IT manager for Grupo Maseca, the largest producer of corn flour in Mexico, a move to a standard desktop seems to have been an act of faith. 'We really had very little idea of how much our desktops were costing us' he told us - and even after implementation of a standardized environment it is still to be settled, as the company currently has a team of Microsoft consultants working on site trying to figure it all out. In his defense, Sosa maintains the decision to outsource the company's 1,200 desktops as part of a 3 year multi-million dollar deal with Hewlett-Packard was driven more by a desire to get the IT arm of its operation doing more, than as part of any bid to reduce costs. The company has an annual turnover around $700 million and has recently experienced a period of rapid growth as it expands its markets into North and South America.

Admitting to being 'guided solely by the desire to reduce cost' Smith of Baltimore Gas & Electric seems a long way off being able to make any decision about standardization. He says he is unsure at this stage even if he should be leasing, renting or buying all his desktop systems. In order to reach a judgement on the value of going to a standard desktop the company is making extensive use of asset management packages from NetBalance Inc. and Janus Technologies Inc. This should allow the utilities firm first to start collecting all the necessary financial data - culled from disparate corporate sources - and then to tie expenditure to hardware contracts and purchase orders and details of software licenses. It should all help make decisions about desktop purchasing more realistic, but it seems there is much to do before the cost-benefits of a move to a common desktop can be accurately analyzed.

Vendors would protest that the extensive research they have sponsored into the total cost of ownership (TCO) should help evaluate such a move. The problem is that this research has been so varied with findings seeming to tell a slightly different story. So this month Computer Finance steers a way through the minefield.

A Focus on Running Costs
TCO models are essentially sales tools. They offer comparisons of differing architectures and assume a range of fixed management practices. Normally TCO analysis will include calculation of the so-called 'soft costs' of peer support and the 'hidden costs' of lost business opportunity through downtime ... calculations which render the analysis useless for all but the pseudo-analyst sales pitch. Estimates of downtime, for example, include assumptions on unavailability against some supposed industry average of the cost, regardless of the wide discrepancy between the outage expense of an electronic point of sale (EPOS) system and that of the desktop on a brokerage's trading floor.

Similar distortions occur when abstractions such as shadow staffing and informal peer support are taken into account. Are they a cost in terms of lost productivity or a benefit in terms of increased team spirit.

The decade-old, industry stalwart TCO model which adds a new factor called 'business risk' to the estimated cost of stuff like computer downtime is, it is claimed, able to supply the actual business cost specific to an organisation. Disappointingly, however, the new TCO Analyst reverts to the infinitely malleable world of averages to provide a factor for the variance in risk across industry sectors.

Other approaches such as calculations of Total Economic Impact (TEI) focus upon the merged or holistic value to business of IT. This argument centers on the fact that most TCO models measure only the cost side of an IT investment. TEI takes in cost components and an appreciation of best practice guidelines for cost reduction and then extends the whole exercise to take account of business benefits and flexibility balanced against corporate risk. In other words, the view is that while TCO measures cost efficiency, TEI is a yardstick of cost-effectiveness.

Of greater value might be use of an operating cost benchmark (see Table 1 for some hypothetical ballpark figures which might provide a basis for your own cost model) against which the efficiency of the desktop spend could be judged. To do this CIOs need only establish the steady state or running costs of a system - strip away all the qualitative costs normally included in TCO analysis and consider only the hard spend, the capital and operational costs.

Such a notion is contained in what Meta Group refers to as the Real Cost of Ownership (RCO). It has developed a cost-benefit methodology to evaluate the services IT provides to line-of-business (LOB) units, a key aspect of which is the separation of easily quantified components (capital costs, and IT staff expenses) from subjective factors (peer support or shadow costs, and business benefits).

RCO differs from TCO by separating LOB costs from IT costs. Because most IT budgets do not incorporate shadow support or recognize the productivity cost associated with users operating client machines, Meta has captured LOB costs in a separate cost-benefit analysis category. While IT costs can be accurately quantified (e.g., in dollars, in full-time equivalents [FTEs]), the LOB cost and benefit sections of the RCO model are often determined through subjective analysis. As a result, IT organizations must often compare quantitative cost projections against qualitative benefits when making technology decisions, instead of traditional ROI analysis.

In this RCO model, there are two major categories for IT costs - capital and operational costs.

There are some aspects of computing costs which just cannot be managed away (most of the 50% premium paid for notebook versus desktop PCs, for instance, simply have to be carried - remote dial-in infrastructure, shorter life cycle and the higher purchase price of notebooks are good examples). But while there are many cost factors over which IT has no control the degree of desktop standardization will influence operational costs to some degree, though Meta suggests that IT departments should not go to extreme measures to eliminate heterogeneous products, considering the realistic futility of such an exercise and the potential dangers of vendor lock-in.

For PC hardware, the suggestion is that organizations should narrow the list of suppliers to two or three preferred vendors. Applications (e.g., e-mail, calendar) should be standardized within individual business units, but this becomes considerably more difficult across the entire enterprise.

Comparison of internal estimates of running costs with the figures shown in Table 1 wil help give a comparative measure of cost-effectiveness and hence some guidance as to where attention needs to be focused when looking for cost savings. By targeting specific elements of the desktop spend and excluding soft costs - which by definition are intangible, abstract and usually not quantifiable - managers are free to concentrate upon those elements of the desktop spend on which they can have an impact.

Such a running cost methodology assumes that the skill of the driver, the IT manager, is paramount. Used in tandem with an appreciation of some best practice desktop management methods, running costs could become an invaluable management tool in lowering the desktop spend by allowing the better management of IT, en route to a common or standard corporate desktop environment.

There are all sorts of advantages of comparing internal and benchmark running cost models, it:

* Gives a measure of scope for cost savings.

* Gives a working methodology for achieving cost savings.

* Allows users to target specific elements of cost for reductions.

* As a technology migration exercise, such as a move to a standardized desktop, the model would help gauge costs before, during and after standardization.

Common Operating Environment
By viewing the PC as a commodity item capable of being standardized, Hewlett-Packard believed it would realize substantial savings in terms of reliability and a reduction in the cost of desktop systems management. Back in 1992 the computer giant had an estimated 100,000 seats worldwide which was estimated to be costing $4,750 per seat per year. This realization led the company to adopt a more pragmatic attitude toward IT and the sacrifice of certain freedoms commonly associated with non-standardized environments.

A company wide strategy to establish a Common Operating Environment or COE was set up. It viewed IT as a process, as a people thing rather than as an IT/technology issue. It aimed to maximize the efficiency of the then emerging client-server mode while attempting to rein in the spiraling cost of ownership of desktop computing. In addition, the establishment of a global intranet accessed via a common desktop browser would result in unspecified business benefits and qualitative gains. Some key features of the Hewlett-Packard PC-Common

Operating Environment included:

* Standardized PC configurations... lessening software conflicts.

* Common set of desktop and system tools.

* Common procedures for integrating operational applications.

* Managed configuration via the network.

* Development of a structured training program.

* Cultural changes to the organization.

The software used for remote maintenance and management of PCs played a key role in the move and was the main source of untold cost savings (see Table 2 for a view before and after COE implementation). 'We obtained favorable license terms due to effective management. We reduced the risk of litigation and license infringement. And knowledge of the target platform kept down to the bare minimum unexpected costs due to installation problems and incompatibility issues,' Hewlett Packard's Lawrie Kimble claims.

Five year cost comparison assuming conventional three year lifecycle. Traditional purchase consists of LAN card and five year on-site warrenty, captial cost = $3,440. Upgrades $208 each at 18 months intervals, except 36 months (replacement). 8% compound annualinterest on capital. Excludes procurement, Installation, support, asset tracking/management, upgrade process and other administrative costs associated with PC ownership. TimelessPC costs assume Standard TimelessPC, $140 per month (3 year contract + renewal on same terms).

However, he stresses that cost savings are not necessarily made solely by implementing specific tools or technologies. 'I know a number of major corporations that have tried and failed to implement a COE-like solution because they have treated it purely as a technology issue rather than as a change management project'. A move to a standardized desktop will affect how people work. 'End users will have to give up some of the perceived freedoms they've had in the past. But for IT, it could mean staff can spend more time focusing on business issues and less worrying over technical issues.'

For HP the move has clearly paid off. Three years after its introduction, the Common Operating Environment has a 99% implementation with 136,000 nodes and has:

* Reduced the support staff to end user ratio from 1:50 to 1:200.

* Reduced the time required to install new PC configurations by 75%.

* Cut the time it takes to distribute new software worldwide from one month to three days.

* Improved response times to users' questions by a factor of 10.

* Saved HP $40 million a year on PC management expenses.

The figures cited in Table 2 exclude the cost of migrating to the PC-COE, which were not made available to us. Tools used were developed in-house and some have since been commercialized for the market (thereby paying for themselves, in HP's view). It is noteworthy also that within a well managed environment soft costs decline almost as a matter of course. Downtime reduces naturally where it is less likely to occur as a result of the deployment of tools designed to pre-empt such events. This lends weight to the argument that hidden or soft costs are a function of management rather than of architecture.

Desktop COE in Practice
Grupo Maseca Inc., of Monterey was unwilling to tell us how much it is paying Hewlett Packard to put in place its new common operating environment. We do know, however, that the company is paying $2,500 a shot for the initial purchase of a 500 batch of 100MHz, 32Mb RAM, 1.2Gb HP Vectra's (which seems a bit on the high side) and if this is at all indicative of the kind of price the business is being charged the overall bill will not bring the cost of Maseca desktops within the range achieved by Hewlett Packard itself internally.

Nonetheless, the initial focus in Monterey will no doubt be on shifting towards better desktop management practices as a means of controlling expenditure and the roll out of better desktop management tools. The introduction of extensive asset management tools has according to Sosa, 'made life so much easier.'

Acknowledging the move rests squarely on successfully communicating the philosophy of the change to a common desktop, Sosa says much was done to involve everyone in the process. As part of the initial phase of the implementation Maseca was encouraged to set up a steering committee made up of representatives from various departments to decide upon the mix of services to be included as part of the COE.

In summary, the Grupo Maseca environment takes in:

* Standard 100MHz, 32Mb RAM, 1.2Gb drive PCs.

* Centralized systems support for all 1,200 office desktops based on Windows NT and a manufacturing desktop PC capability based around Windows 95 and SAP R/3 running on HP9000's.

* A dozen Unix servers running Informix DBMS supporting some 1,200 clients, over 40 regional sites.

* Use of Microsoft System Management Server (SMS).

* Use of Tally System's Cynergy system for asset management.

Toward a Commodity Approach
En route to some common Windows NT desktop environment, Smith of Baltimore Gas & Electric says that while he has already decided to outsource the procurement and maintenence of all the company's 6,000 desktop PC systems, the next decision is over whether to start leasing rather than buying them. Such debate has prompted another vendor to pick up on the standardized/commodity approach to desktop operation and launch a scheme that is somewhere between a traditional PC lessor offering and an outsourcing contract. With Siemens Nixdorf's Timeless PC a business can hand over the installation and management of standard issue PCs for a fixed fee that includes a limited help-desk service covering operating system and hardware issues (though a more extensive support service is available at a premium).

While the service goes only so far, there is a small return to be gained from such a basic arrangement and some organisations will only be too happy to hand over to a third party the purchase and management of such a commodity item as a standardized hardware platform. Afterall as René Caryol, IT Director at IPC Magazines and one of Siemens Nixdorf's customers confirms, 'routine management of desktop hardware isn't going to provide a business with any competitive edge nowadays. All the time, I'm wanting to move away from worrying about actual deployment and instead try to focus on what applications can actually do for the business. We can't do that if we spend every minute managing the desktop, managing the network and looking after hardware maintenance.'

To this degree, there may well be IT management benefits in a scheme like TimelessPC. It is meant to look financially appealing, as well. PCs are installed at a fixed monthly cost based on a banded structure for pricing. There are three bands: basic, standard and advanced for an upfront unit monthly fee starting at around $60 (?5) and rising to $200 (?25) on a 3 year contract basis. The fee covers PC procurement, installation and upgrade costs. And there's a monthly technology assessment that 'index links' the performance of the desktop machine installed under the scheme and results in timely upgrades.

The fiscal advantages of the commodity approach that something like Timeless PC offers includes: no immediate capital outlay; smooth payment stream; off balance sheet finance. The disadvantages are that it offers no asset value and carries penalty clauses for early termination.

Reducing Running Costs
Some pointers begin to emerge here as to the looks of a best practice shortlist that would help first establish and then incrementally reduce running costs of desktop systems: Asset management: As the situation at Baltimore Gas & Electric illustrates, no CIO can rein in costs without having an extensive knowledge of the desktop asset base.

Standardisation: As the cases of British Steel and HP have shown, this looks a sure way of reducing costs by lessening the possibility of software clashes and allows the use of more cost-effective desktop tools.

Lease and vendor financing schemes: When money is tight, or accounting treatments dictate lease options could be way of funding desktop investments.

Use of desktop management tools: The big ticket item in PC TCO is the cost of management and support of desktop systems. As a first step to keeping them in check, the situation at Group Maseca indicates that deployment of a good selection of desktop management tools is essential (a focus for reports in forthcoming editions). These include products that help increase efficiency and automate tasks associated with PC audit, PC maintenance and helpdesk support - the likes of: fPrint IV: An asset management package for creating an inventory of hardware/software in use across an organization ($35,000 for 1,000 - 2,000 desktops).

Cybermedia Support Server Repair-Engine (CSS) for workgroups, 1.0: Detects and prevents known software incompatibilities and has the potential to lower the break/fix support element of PC running costs ($90/seat).

Platinum's DeskWatch: Performance monitoring/analysis tool which evaluates equipment needs, upgrade needs and application requirements (identifies users that require more training or support by picking up on recurrent user errors and is of use in helpdesk cost reduction) (around $175/seat).

AssetPro: Establishes a database of previous purchases, recording levels of service obtained from various vendors such as price, delivery times and installation charges. Also has links to a variety of catalogues that it can search and generate purchase forms to deliver an economical mix of products from disparate sources ($30,000 base unit).

Table 1. Typical breakdown of running costs of a desktop environment in Global 2000 businesses ($/year)
(Source: Meta Group Inc.).

Capital costs ($1,372/PC annualized for a typical PC/LAN environment) are easily quantified because they are already available in dollar amounts. Traditionally, capital costs have been thought of as insignificant versus operational costs, but when all costs are factored (including clients, networks, servers, and applications), capital costs virtually equal operational costs. As a result, organizations should leverage volume purchase agreements for all hardware and software procurement. The operational costs ($1,417/PC annualized for a typical PC/LAN environment) are measured in Full Time Equivalents (FTEs), then given a dollar value based on the average fully burdened cost of an FTE. These are the cost estimates based on a hypothetical 10,000-user organization, distributed across multiple geographic locations and running applications across both clients and servers. The automated management infrastructure of this organization is typical of Global 2000 firms, including knowledgebase and trouble-ticket engines for end-user support and automated tape backup for storage management, but no extensive use of electronic software distribution, remote control, or inventory tools.

Table 2. Costs per PC per year ($).

Computer Finance - 11/01/97
Finding The Elusive Desktop TCO
Now that thin clients are being proposed as an alternative to conventional self-contained networked desktops, the issue of total cost of ownership has advanced off the back burner. In this report, we examine the methodology of one new tool.

What's the real cost of personal computing? In the past few years, attention has been riveted on so-called total cost of ownership (TCO). The findings of virtually every industry study points to the fact that upfront costs are but a fraction of the real costs of operating PCs.

For instance, Meta Group points out that continued pressure to cut overheads often results in reduced support budgets that may lower the reported costs of support - one factor in TCO. But those costs don't go away. Meta notes that these costs end up being absorbed by the 'shadow help desk', a power user sitting nearby who helps colleagues solve routine and not-so-routine PC problems on his or her own time. The costs are still there, but they grow hidden.

We analyzed the returns on help desk in a previous issue (see index, p. 24). However, even if organizations come to terms with the cost of support, many wonder realistically where those (and related) costs will ultimately end. In this report, we examine one of a new breed of tools, TCO Adviser from Interpose, which claims to help answer those questions.

Cost Factor Breakdown
Total cost of ownership analysis tools are designed to put a front end onto TCO models, providing IT managers with an easier way to analyze their organization's IT data. One company addressing this issue is Interpose. This company has just launched the latest iteration of its TCO analysis product. The product, called TCO Advisor, was previously called client-server Solution Advisor. According to Kevin Auger, vice president of marketing at Interpose, the tool is being used by corporate IS departments and IT vendor companies including Microsoft to report on TCO figures.

Auger says that a number of different factors have driven the market to focus on total cost of ownership, including the year 2000 problem and the rising popularity of the network computing initiative. 'I think that these things are making companies pay attention to their own environments. They are making IT managers aware of what their assets are, what they are costing today and how they can move forward with improvement plans,' he says. 'Outsourcing is another thing that is coming of age. People are considering it a lot more. All of those things put together are pushing the IS organizations to look for sensible business decisions and to make those business decisions they need the information.'

Auger explains that client-server Solution Advisor was more of a return on investment tool than a total cost of ownership assessment product. It was only with the second version of the product that the company began to include total cost of ownership parameters in the back end logic. The product uses total cost of ownership parameters and industry wide data on companies' total cost of ownership metrics gathered by a specialist market analyst group, which has been collated into an index. This is referred to by the front end tool to see how close customers' costs are to the industry average. Under the direct elements, the Interpose model lists (see Table 1):

* Hardware and software. The capital expenditures and lease fees for new installations, and upgrades of servers, clients, printers, and network communication devices.

* Management. The network, system, and storage management IS labor and professional services outsourcing fees.

* Support. The help desk support labor, training labor and fees, procurement, travel, maintenance and support contracts, and overhead labor.

* Development. The application and content development, test, and documentation including new developments, customizations, and maintenance of non-business applications.

* Communications fees. The lease line, server access charges, and allocated WAN expenses.

Indirect cost factors include:

* End user IS. The cost of end users supporting themselves instead of relying on IS support (peer and self support), casual learning (non-formal training), and end user self-development of applications.

* Downtime. The lost productivity due to planned and unplanned network and system unavailability, measured as lost wages.

The model also breaks down the potential benefits to the organization into economic benefits, detailing tangible financial gains, and business benefits which focus on less quantifiable gains.

The economics category is comprised of productivity increases, organizational efficiency, revenue and cost avoidance. The latter is a measurable reduction in current or future spending, outside of the TCO model, particularly the elimination of fees and expenses, or the elimination of a planned future expense.

The business benefits addressed by the Interpose model include capability, competitive advantage, flexibility, performance, risk mitigation and finally service.

Market Differential Difficulties
Any reporting tool is only as good as the data that is entered into it. Ashim Pal, an analyst for Meta Group says that for any product to be effective, the model behind it must be comprehensive. Keep in mind, Meta has a similar model that competes with Interpose's. We will examine this model in a future report.

In spite of Auger's claims for the comprehensiveness of his product's model, he admits that there are some parameters that are not included. Some of these parameters would be particularly important to companies making large purchases in the US.

One such omission is the lack of consideration given to regional tax variations from state to state. If a company is spending hundreds of thousands of dollars on the components of a new IT strategy, then the last thing it needs is to have its TCO estimation thrown out of whack by the loss or gain of a few percentage points on the balance sheet.

However, sometimes it's not always cost effective to track everything. Pal explains that the administrative overheads spent gathering this data may be too great to justify the distinction in state tax.

Consequently, companies may find it more cost-effective to analyze things from a more abstract perspective, ignoring the fine details. 'We are in the process of fitting in state to state variations in tax in our own model,' he explains. 'That is good in terms of the accuracy of the model but it may add so much to the complexity that it may be best to stick to a generic assessment. That has some power in terms of simplifying the model.'

One argument is that for companies with offices in many states, the variations in regional tax may balance themselves out to create an average TCO.

Perhaps a more serious flaw in the product model is the lack of distinction between different industry sectors. The industry TCO data provided as part of the Interpose model incorporates figures from a number of different companies operating in different markets. This creates problems, because IT can be seen as a core competence by some companies but not by others, and corporate attitudes to IT investment will depend largely on industry sector.

A bank for example is more likely to want to invest a large proportion of its revenue in cutting edge computing technology because of its focus on the transfer of information and on a high quality of customer service.

A manufacturing company, on the other hand, may want to make less of an ongoing investment in IT in certain areas and in some cases may decide to outsource more aspects of its computing operation.

Auger responds that the next version of the product would support not only more variations in geographical TCO factors, but also data which differentiates companies operating in different industries. The company has started to address this issue now, by taking data from some of its customers after they have used its software.

He parries concerns about the accuracy of the data in the current product by arguing that the broad spread of different TCO parameters including many indirect costs enables companies to measure themselves with a fair degree of accuracy against an industry average. The question remains, however: which industry is he talking about?

A further worry about a product such as TCO Adviser is that in order for it to work properly, the data plumbed into it has to be accurate. Gathering accurate data is a problem that increases exponentially with the size of the organization. (In fact, this is also a challenge with another metric, function points, also described in this issue; see page 1.)

Potential customers should not underestimate the cost of gathering an accurate inventory of their systems and other IT resources. Monitoring the hardware that exists in a company is bad enough, but monitoring the software that resides on that hardware is even more difficult.

Auger argues that it is impossible to establish a return on investment for this process, because until you have the data you will not be able to work out how much of a saving you can expect to make by altering your IT strategy.

IT managers are therefore in a Catch 22 situation. Much to their consternation, they realize that financial directors like to see figures in black and white before they will commit any budget to an enterprise-wide initiative such as inventory and asset management.

Meta's Pal believes that gathering data on a survey basis can lead to inaccuracies. 'A lot of the data that you're dealing with in a TCO study will be somewhat soft and estimated. It may be impossible to get it all,' he says, arguing that people involving in the thinking process will probably end up owning the model results, which could affect their responses. 'A reality check is worthwhile. You compare costs with another company of the same type. You then get one of the other consulting firms to verify that what you have done is logical. You have a certain amount of hard data gathering. On top of that there is some soft data gathering. You then need to test parameters in the mold to see whether they are applicable.'

Rod Franklin, vice president of Interpose partner Entex, does not think that the data gathering process is very difficult, unless discrepancies are found in the data. If that's the case, administrative costs can easily skyrocket because of the need to spot and resolve errors or omissions. He recommends spot auditing the results, and says that typically, sampling responses from a 4,000-user organization will take two staff days out of a six-week engagement period.

If the results are satisfactory, the spot verification process shouldn't be too expensive, he says. If not, then extra time must then be allotted for detailed audits. He notes that many of the companies using the product have their own inventory services, however, meaning that it will be easier for customers to attain a fixed cost for this part of the project.

Although TCO Adviser could logically be perceived as a tool which focuses purely on cost savings, many users say that the tool comes into its own when used to analyze the potential for functional enhancement in different parts of the IT operation. Many dismiss the cost savings element of the software altogether as purely a means to an end.

Entex's Franklin says that he uses the tool for honing functionality to reduce cost savings in the long run. He explains that although the product can be used for finding out where costs are too high, if you take that to the logical next step you can't reduce the cost unless you make functional changes. 'We use the Interpose tool to hone in on those areas that need improvement. Then you have to dig below and ask why these areas incur these costs and what process changes need to be made so that they get lower costs and more efficiency in this area,' he says.

User Experiences
Because TCO Adviser is such a new product, there were relatively few users who could point to any results when we compiled our analysis. Even fewer organizations were willing to go on the record.

Sears Termite and Pest Control: Sears used TCO Advisor to reduce costs and increase functionality within its IT department. Kemp Anderson, director of management information for the company, explained that the firm used TCO Adviser to look at both direct and indirect costs.

Anderson, who says that his administration system has been DOS-based rather than Windows-based for a long time, explains that one of the cost-heavy areas that the product found was Sears' support department.

It was discovered that there were some unaccounted for support costs, because many people were making up for the lack of central support resources by conducting support on an ad hoc basis. This meant that support costs were disorganized and decentralized and difficult to monitor.

'The TCO is all based on the corporate network. We struggled a little bit getting the data required because we had to use surveys. Having to pull the information together internally is difficult to do but it is a good exercise. You have to pull things from a lot of different sources,' he says.

Anderson explains that he had already made the decision to move from a DOS to a Windows-based environment before using the TCO Advisor software. However, he says that the software helped him to ascertain the cost of the move. 'Most companies out there, everyone walks around with laptops or has systems on their desks, but no-one can quantify how much it's costing you to run these systems and TCO Advisor helps us do that,' he says. 'It has good graphics, good charts comparing you to the industry average and it was simple to use.'

One decision Anderson made as a result of using TCO Adviser was to recruit more IT staff to the central support department. This should free up valuable time for non-support staff to address their allocated tasks. He explains that the organization is centralizing and is growing at roughly 20% per year.

TCO analysis tools like TCO Advisor seem to be few and far between. An emerging product category, the TCO analysis tool relies very heavily on the comprehensiveness of the back end TCO model and the process used to gather and assimilate data. It also relies on the granularity of any industry-wide index data so that companies can produce results that reflect their business environments.

The most important thing for any company to remember is that the results should not be taken as being entirely accurate. While they may be able to give an approximate result, no company is exactly the same, and key assumptions that may be made about one company, such as its level of decentralization, will be different in others.

Accurately collecting the volume of data which will highlight these differences is a costly task with a naturally occurring margin of error.

A Major US Food Retailer: We also came across a TCO analysis conducted by a major US food retailer using TCO Advisor. The retailer had been conducting its own TCO analysis for three years and wanted to compare the actual TCO figures to an industry average.

The six-week analysis was conducted on an environment consisting of 5,693 clients, 204 servers, 612 printers, 184 hubs and 28 routers. It used 150 person hours of consultant time, and 195 customer personnel hours. The aggregate results are shown in Figures 1 and 2.

The company found that it had a 20% help desk call abandon rate, and an average 7-minute hold time. Each call lasted only five minutes on average, but frequently failed to solve the problem: less than half of the IT issues were solved at the help desk. The company found that it lost 8% of its productivity due to self support or peer support and that it spent ten hours of self/peer support per user per month (compared to an industry average of four hours). Users in the company involved 1.6 neighbors in solving each problem and received little or no training.

Overall cost comparisons: direct versus indirect costs (source: Interpose)

The recommendations made as a result of using the product included re-staffing and adding tools to leverage existing staff, standardizing the desktop environment and considering a refresh program or leasing and implementing a more manageable and controlled desktop environment.

The company was also advised to implement an end-user training program and to retool its help desk with new systems, staff, training, and publicity. It needed to implement asset management to get awareness and control of the environment and put in place a TCO lifecycle management process to track progress.

Cost Factor comparisons for a major retailer

ROI isn't the whole answer to TCO
Although end user corporate customers may use the TCO Advisor product on an ad hoc basis, professional consulting firms with a core competence centered around increasing IT efficiency often take a more holistic approach to using analysis software. Tim O'Brien, director of business consulting at services company Comdisco, explains that the TCO concept has been around since the mid to late 1980s, but at that time there was no toolkit available to measure it.

'You can't manage what you can't measure,' said O'Brien. He notes that organizations have traditionally shied away from TCO measurements because of the difficulties of conducting them in distributed computing environments. 'To develop integrated solutions you need a baseline of service levels, but also a handle on cost.'

Significantly, O'Brien tends not to speak about return on investment when assessing customers' IT costs. He says that return on investment is a measurement that lends itself more to revenue generating projects, and contends that very few CIOs believe that the technology represents a positive return on investment. Return on investment metrics are full of assumptions about the company's IT operation, he says, adding that he is not comfortable with the term. He follows an assess, implement and manage aim methodology to improve the customers' business benefit from IT. He uses the term 'business benefits' advisedly, because many customers tend to want to treat computers as assets that generate cash flow, he says, adding that this is not an accurate way of looking at things.

The assessment stage is little more than a fact-finding method. From this stage, the company developers make recommendations and implements them. O'Brien explains that there are three ways to find things out about an organization: a physical inventory, plus an estimate or an auto-find product. As for the physical inventory portion of the job, O'Brien maintains, 'that is a painful process and the only real benefit of a physical inventory other than the accuracy is if you're building a central asset repository and you want the data in that repository to be 100% accurate on the day that the new solution rolls out.' He notes that the window of error [from an estimate] varies from industry to industry, noting that a 95% accuracy rate is 'reasonable'.

O'Brien says that when speaking to end users directly about their systems, the volume of imports can be staggering. Consequently, he generally puts a consultant on site with the client for four to six weeks. It helps to have some synergy with the customer's business, understanding what its aims are for the project. If the customer wants cost savings, then before the project begins in earnest, Comdisco get a feel for what their points of pain really are. Alternatively, they may be focusing on service agreements as part of an outsourcing project.

O'Brien says that it is up to the customer how granular the data collection is on a regional basis. The product gives them the ability to measure cost on a very small scale or at the enterprise level, he explains. Companies that manage costs may decide to do a cost assessment on five regional offices in parallel for example.

As a final thought, O'Brien explains that one of the primary purposes of TCO is to identify the areas that need the most work. He says that Comdisco prioritizes problems in customer environments, putting first those problems which would incur the highest cost by not being solved. 'I want to emphasize that Interpose has a terrific tool. Its value is maximized by building processes to develop cost and service level metrics and tracking those costs over time,' he concludes.

Table 1. Interpose TCO cost factors
(Source: Interpose; http://www.interpose.com).

Computer Finance - 09/01/97
Targeting TCO: The Wintel Strategy
As part of an ongoing series to investigate the total cost of ownership (TCO) of thin clients, we examine in this report, Microsoft's Zero Administration for Windows (ZAW) initiative and Intel's Wired for Management initiative, for NetPCs.

By now, we've all heard the mantra: network computers (NCs) are cheaper than fat PCs. As the target of bloated total cost of ownership (TCO) analyses, that have estimated the cost of fat desktops at $10,000 per user annually, Microsoft and Intel have responded with several initiatives and specifications for managed, or thin PC clients.

Their message: you can cut the TCO of desktop clients with the Wintel architecture.

Specifically, we are referring to Microsoft's Zero Administration for Windows (ZAW) program and Intel's Wired for Management scheme. Both are designed to complement each other to support the NetPC solution: in effect, a 'dietary', as opposed to a fat or thin client, because they are designed to offer a more flexible solution than the ultra thin network computing clients, which contain little or no intelligence or storage on the desktop.

Will the NetPC provide relief for swollen IT PC support budgets, and how does it stack up to the NC and PC?

Neither side has much real data to support their hypotheses, but that won't stop us. In this report, we examine Microsoft's Zero Administration for Windows (ZAW) initiative and Intel's Wired for Management proposals - what are they based on, and what experience is there to date to validate or refute these models? We also look at two early NetPC users.

Microsoft's ZAW Offering
Microsoft's ZAW initiative will culminate in features integrated with NT 5.0 and Memphis, the next version of Windows. For now, users can download the Zero Administration Kit (ZAK) for Windows 95 and NT Workstation 4.0 from the Web which offers limited Zero Administration capabilities. ZAW offers key benefits in four areas:

* Automatic management of changes. This allows system administrators to install the operating system and application from a central server in a similar fashion to NC models, reducing the resources needed to upgrade end user software.

* Centralized administration. Administrators can manage their environment using a new management interface through a Web browser, adding snap-in modules for more functionality.

* Application flexibility. A key differentiation from the NC model, this feature allows administrators to specify the capabilities that a client operating system has.

When Windows NT 5.0 emerges, the ZAW features will also include the following:

* Intellimirror. A technology which replicates the state of a PC and its data onto a server. This allows users to relocate and access their data and applications should their client fail. The ability to 'roam' also allows users to log onto any machine on the network.

* On Now. A technology which enables companies to leave NetPCs in a dormant state to be 'switched on' instantly from the server. This enables administrators to perform prescheduled administrative changes without leaving their client machines on for periods when they will not be used.

Wired for Management Initiative
Intel's Wired for Management initiative, announced last September, encompasses a variety of technologies and products which allow corporates to manage the state of their PCs more effectively from a central site. Three key features of the scheme are:

* Hardware and software-based facilities that allow PCs and servers to be remotely recognized and manipulated. This enables remote software distribution and upgrades.

* Desktop management applications for the remote management of PCs including software installation, configuration and repair, based around Intel's LANDesk management technology.

* Integration with enterprise management systems such as IBM's Tivoli, enabling IT departments to fold PC management into existing management programs.

The Intel scheme is designed to provide better thermal management of the hardware and more effective mechanisms for creating alerts, and better procedures for handling the Desktop Management Interface (DMI) technology, which is a standard for feeding back client information to a central site. It also includes technology such as Wake On LAN, where you can 'wake up' a PC to perform pre-defined processes.

The initiative saw the introduction of a new product, LANDesk Client Manager, a desktop management application designed to gather and present information from the managed hardware systems.

The company also launched a motherboard including an environment management processor that monitors parameters such as voltage and system temperature to feed back to the central management site. The NetPC is a subset of the Wired for Management initiative.

Advantages and Disadvantages
One large advantage presented by the ZAW initiative is that it is flexible.

Unlike its nemesis the NC, the NetPC enables companies to define their own level of rigidity at the client level. Using the ZAK (support for which is mandatory in the NetPC specification) administrators can choose one of two levels.

They can force clients into TaskStation Mode, where the desktop is completely locked down and one application is accessible through a browser; or they can opt for AppStation mode for a knowledge worker who runs three or four applications but does not need to alter the system configuration.

A black mark against the ZAW/ZAK and NetPC initiatives is their ambiguity when it comes to mobile workers. Because the systems are designed to give administrators more control over end user PCs, mobile workers who are away from the network and the server for varying periods of time, will be difficult to manage. (However, the NC has not managed a satisfactory response for mobile users either.)

In spite of this, a positive aspect of the ZAW/ZAK and the NetPC schemes is that they reduce support overheads. In particular, help desk support for users whose clients fail will be more manageable because the ZAW initiative will enable users to relocate and use another client. This means that the need to interact with the help desk would decrease, and the impact of client failure will also fall as end users find themselves able to recover their data and application state very quickly, saving time and therefore money.

The AppStation mode offered by the ZAK offers an alternative to a problem cited with Java-based NCs: the compact Java applets on which Java-based NCs are based offer minimum functionality, which may pose constraints for knowledge workers and senior executives. In short, the application is best suited for structured tasks that are repetitive and do not deviate from a set routine.

By contrast, the AppStation mode offered by the ZAK gives access to more applications.

The fact that applications can be processed on either the server or the client in a NetPC and ZAW environment, means that customers can create their own TCO profiles. There are advantages and drawbacks to server and client processing. One of the biggest disadvantages facing NetPC users that want to use the system as a completely thin client is that Microsoft does not currently offer customers the facility to do this within its own product.

This problem will be resolved when it releases a version of its own multi-user server core, Hydra Server, for Windows NT Server 4.0.

Until then customers wanting this option have to rely on Citrix server software which is only available to run on NT Server 3.51.

We examined several Citrix WinFrame sites in a previous report (see index, p.24). One of them, Holsten Medical Group, wanted to enable its physicians to key in patient data direct from exam rooms. That would have averaged five PCs per physician and a resulting cost of $600,000 (200 $3,000 machines for 40 doctors). Instead, Holsten installed 175 Wyse WinTerms at a cost of $900 each.

In spite of popular concerns about network overheads, Chip Childress, Holsten's IS director, says that there are no network problems when running the applications over his 10Mbit Ethernet infrastructure. Instead, he finds that performance increases when all of the Citrix servers are centralized alongside the database server on his 256Kbit Frame Relay WAN rather than administered locally.

Part of this optimal performance is provided by the 3Com switch that he has installed on his network, and the other is due to the very small packet sizes that Citrix uses, he says. There is also the fact that Citrix divides the client screen up into a 64 x 64 grid and sends over only the screen data that has changed, he says, adding that this also provides good security because no true data goes across the network.

Childress estimates that at the time of installation, he saved roughly 50% on the cost of each of his 175 client devices, based on a required client PC specification of a 100MHz Pentium with 24Mb RAM and Windows 3.5 Workstation. He believes that he is getting similar performance to a 133MHz Pentium processor now.

Potentially, the all-Microsoft approach would more effectively tap the enhanced features within NT 4.0. These features include faster Web server facilities with automatic content indexing, server-based DCOM support and built in support for Novell's Network Directory Services.

Microsoft has made no announcements about when Hydra will ship, however.

The other drawback to using a totally server-based environment where users access application code centrally is that it places extra strain on the server, according to Ashim Pal, an analyst at Meta Group. In particular, Pal warns that customers will find Microsoft applications such as Office 97 to be particularly power-hungry when run on the server, causing noticeable performance hits.

This will cause infrastructure costs to rise sharply as more server power is bought in to compensate.

The NetPC specification allows client devices to operate in a hybrid manner too, so that users can download some application logic from the server and cache some on the disk. Intel and Microsoft argue that the provision of a hard disk in the NetPC specification can reduce the bandwidth requirement on the network because it enables the client to cache application code that otherwise would have to be bought down from the server every time it was used.

Theoretically, this increases performance, but in reality it threatens to place strain on network bandwidth according to Pal.

ZAK works by downloading pieces of Active X code that it needs from the server and running them on the client. Pal argues that in spite of the caching facility, different Active X controls will often be downloaded across the network to gain access to application components. Active X controls are often very large which means that the network will still come under strain, he warns.

Alternatively in an NC environment any client-side server space is small and generally located in RAM, meaning that companies have to be more disciplined about their downloadable applets.

Another consideration that Microsoft and Intel have overlooked according to Pal is the social aspect of the computing model, which is nevertheless easily quantifiable. Microsoft quotes a blanket saving of 48% of management costs, using a comparison with a 'loosely managed' implementation of Windows 3.1 or 3.11. Microsoft's TCO cost reduction figures in the context of ZAW, and the Wired for Management program do include support costs. However, Microsoft has not broken out support costs based on the geographic location of the support staff.

This oversight means that Microsoft cannot present an accurate picture of the cost of support which includes the cost of employment. Labor costs in Germany and Switzerland are greater than in the UK and US, for example, because of factors such as employee rights.

Another geographical variation according to Pal is that companies in non-English speaking countries will generally pay 10% more for their software than in English speaking countries because of the need for international software versions.

Another factor affecting employee cost is holidays. US support staff generally have 15 days holiday as opposed to UK staff who take four weeks on average and staff in Germany and Scandinavia that take eight. To apply one number to support costs is therefore na?ve, asserts Pal. This point is valid because variations in support costs means that the support cost as a proportion of the overall TCO will also vary. Because the cost of support is emphasized so highly in Microsoft's justification of the ZAW initiative, this is particularly damaging to the company's argument.

Given Microsoft's emphasis on cost savings, it is worth examining alternative cost saving figures from Meta Group which clearly show a higher cost saving from NC installations than from NetPC installations. According to Meta Group's data, the total cost saving gained from moving to a pure NC environment from a conventional PC-LAN environment is 26% compared to a mere 12% gain from migrating to a managed PC environment (effectively a NetPC environment). Although Microsoft's ZAW initiative promises to support Windows-based terminals and it is working with third parties to develop such systems, none are yet available. In the interim, NCs are a smart option for the task-based worker from a cost point of view.

In conclusion, companies employing a high percentage of task-based workers can achieve the best savings using NCs for those employees. Where more functionality is required, perhaps at line manager level, a NetPC implementation supported by ZAK will provide a hybrid of cost savings and flexibility. For mobile and senior executive workers, conventional PCs may be the most appropriate option, offering a high degree of functionality.

As most companies are a mixture of all three categories of worker, diversity appears to be the most sensible option.

User Experiences
Pennzoil: The motor-oil giant has taken advantage of the Intel NetPC implementation, just finishing a pilot scheme with the product at its headquarters in Houston, Texas. The company, which began moving its IT infrastructure to a client server environment in 1993, piloted the NetPCs in its IT department.

Britt Mayo, IT director for the company, has nothing but praise for the NetPC process and says that the NetPC implementation has enabled the company to avoid any increases in server or network management costs because no additional servers or network capacity is required. The NetPCs could just be dropped into the infrastructure according to Mayo, who adds that his company was an anomaly because of the stringent standardization process it had already adopted. This prevented the company's staff from upgrading conventional PCs with additional PC cards, for example.

The company's software is all loaded from the network, Mayo explains, adding that network users generally run the software from the network too, in the same way that they would have done from their conventional PCs before.

There were only a couple of drawbacks to the system, he adds, one of which was that mobile users still had to run their software from the client and could not take advantage of the NetPC features because they were often not attached to the network.

The other drawback for the NetPC is that new units are potentially difficult to add to the system. Whenever a NetPC machine was added onto the network, the Pennzoil network infrastructure had to cope with the complete download of the operating system to that machine. Although adding one machine was not very difficult, adding a lot of machines at once in the form of a complete departmental or company upgrade to NetPCs would present a network load issue, Mayo warns. He adds that he typically replaces his PCs every three years, meaning that only three percent of machines are upgraded a month. This would work out at a tiny fraction per day, he explains.

Because the NetPC market in the US is just emerging and machines are not yet widely available, Pennzoil will not opt for the full distribution of NetPCs in its corporate environment yet, explains Mayo.

Prudential Healthcare: A unit of the Prudential Insurance Company of America, Prudential Healthcare is one of a core set of companies that piloted Intel's pre-production NetPC computers in a live environment to assess the cost benefits of the units. The company provides health insurance to over 15 million customers, processing 30 million claims every year.

Prudential piloted the NetPC devices in the Customer Service and Dividend Units within its Woodbridge, New Jersey offices. The Customer Service unit handles queries from customers and benefits officers while the Dividend Unit carries out year-end accounting information for member organizations that provide healthcare for their employees.

Myles Trachtenberg, Prudential Healthcare's VP and CIO, explains that he uses a separate Pentium 166 NT server with a 2Gb hard drive and an 18 Mb RAM NT server to run Intel's LanDesk Configuration Manager (LCM). This system remotely manages the NetPCs. The company's production servers are Compaq Prolinea and Prosignia boxes with 32-96Mb RAM.

Unlike Pennzoil, Trachtenberg circumvented any potential problems with network load by configuring the machines in an isolated facility when they first arrived, before attaching them to the network at the end user location. This saves the company having to send roughly 30Mb of data as a system image across a busy network during peak hours, he explains. Because the prototype NetPCs were available only with Ethernet connections, Trachtenberg used a Cisco switch to connect the NetPCs to his existing 16Mbit Token Ring network over CAT 5 cabling.

The greatest benefit of the NetPC machines is their remote configuration facility, he explains, adding that the remote control function within Intel's Wired for Management initiative enabled technicians to fix problems remotely without misunderstanding descriptions of problems from non-technical users.

One of the best applications that Trachtenberg can see for the remote management and configuration facilities of the NetPC is in pharmacies across the country. Because support and administration costs rise sharply when dealing with remote nodes over a wide area, the company would benefit from being able to manipulate PCs across remote links, he says.

Even inside the organization, he estimates that he has cut setup time by 50%.

In spite of his enthusiasm for the NetPC, Trachtenberg explains that it may not be the only option for workers within Prudential Healthcare.

The company has 7,000 3270 terminals that it wants to replace across its national service centers, he says, and it is possible that many of these positions will be more appropriate for NCs. This will depend on the application base, he explains. Applications that simply need to access large amounts of data that have been pre-processed at the back end, will be more likely to be accessed using an NC device.

Another possible use for NCs is as low cost access devices for doctors and physicians that need to retrieve customer healthcare information from a central site.