The battlefield of Microsoft v Sun, IBM, Novell, Netscape and Oracle has been redefined in terms of the companies' Total Cost of Ownership (TCO) claims.
The missiles in this battle are press releases and consultants' estimates of the TCO of network computers, network personal computers and personal computers.
The basis for the widely quoted estimates of TCO remains shrouded in mystery. Granted, everyone includes the costs of equipment and operating system software. But those numbers aren't decisive because they can account for as little as 16 per cent of a TCO.
The various estimates don't explain what else is included. For instance, the "cost of end-user operations", which makes up 56 per cent of Gartner Group's TCO, doesn't reveal how much of that is for unplanned downtime, how long it takes to recover from system crashes, time lost in recovering data, the negative revenue impact, how that affects customers and how much time is wasted before someone calls the help desk.
Instead, one finds vague references to "futz" factors. Yet blaming futzing has little meaning in the absence of explaining who, when, why and how it happens. There's no reason to believe any of that would diminish with a network computer or a NetPC.
Consequently, IS managers are in a quandary. How can they reconcile Gartner's estimate of $US9784, Zona Research's $US3679, Forrester Research's $US2680 and Entext's Information Services' $US1500? How can one believe pronouncements from Oracle that its network computer would have "costs equal to that of a TV"? How can anyone accept that Microsoft can field PCs that have "zero administration costs"?
There's a simple answer: you can't reconcile any of the conflicting claims. In the absence of adequate disclosures about what makes up the various cost components, any comparisons of TCOs are guesses. No one is itemising the verifiable costs of what matters the most: the people who try to use a network computer in their jobs.
The total cost of ownership depends on what one counts. Just consider that an equipment purchase of $US3000, amortised over four years and operating for 52 five-day weeks, would cost US36 cents per hour. A typical information worker earning $US40,000 per year would cost at least $US40 per hour.
The biggest contributors to idling workers aren't failures of semiconductors or disk errors. When work stops, it's predominantly a reflection of systems management practices, software reliability and user behaviour.
Identical networks that have identical equipment configurations will experience widely different failure characteristics, depending on the quality of network management and the capabilities of the people.
I'm on the board of directors of a network management company that operates client/server networks for a fixed fee per client. I obtained the following listing of customer help desk calls for a typical time period:l Ninety-eight calls were made primarily because the PC users didn't know what to do. I doubt that swapping computers would overcome such a lack of employee training.l Ninety-four calls reported that the printer didn't work. That's not anything that a downloadable Java applet or a diskless computer could cure.l There were 43 incidents of failed log-on attempts. Only improved administrative practices and reduced employee turnover could remedy that.l In 40 instances, somebody kicked or loosened the connecting cables. I have yet to hear an argument that lower-cost computer equipment will simplify the ever-present cabling hazards.l Only 84 calls conceivably could have been fixed by choosing a more robust and reliable client/server configuration, such as eliminating operating system crashes (31 incidents for Windows 95), automating resolution of error messages (28 incidents) and resolving NT 3.51 Pentium server downtime (25 incidents).
The above network is exceptionally well-managed, with a competitive outsourcing price per seat of less than $US3000 per year. Even after applying superior administration and automated diagnostic aids, 77 per cent of the cost of ownership was largely attributable to human errors that were only vaguely related to choosing different equipment.
Computer executives shouldn't believe that mere changes in network configuration will reduce their TCO. Those responsible for justifying IS spending would be well-advised to examine not only the technology expenses that show up in their IS budgets but also the personnel and business consequences if computer-aided people don't do what they're supposed to do.
As for Microsoft, Sun, Oracle and a long list of consultants, IS managers should be aware that the public relations missile barrages are mostly vapour trails. Until verifiable operating results become available regarding the ability of employees to complete their work efficiently, one should suspend judgment about any expected savings from a proposition that doesn't recognise that humans will always cost more than computers.