N times Z equals fuzzy logic

In the current era of shrinking budgets, I’ve been reading more about the need to cost-justify IT expenditures. Vendors and users alike seem to be talking about total cost of ownership (TCO) and return on investment (ROI) more than ever.

This trend got me thinking about whether such considerations are real. To put that another way, how meaningful are TCO and ROI?

My jumping-off point here is the countless column inches that have appeared over the years about time wasted by staff in nonwork-related Web surfing, the cost of workstation downtime and other examples of people using their time unprofitably. Calculating what this costs the organization usually goes like this: Event X causes each user, on average, to waste Y minutes per day/ week/month/year. The average cost of Y minutes of the user’s time is Z dollars, and as there are N users in the organization, the cost per day/week/month/year of Event X is N times Z dollars.

Now on the face of it, this kind of argument sounds reasonable. But my problem with these calculations is that the aggregation of these personal quanta of time cannot be done logically.

The fact is that you can almost never take Albert’s, Bob’s and Charlie’s lost or abused chunks of time and lump them together in a meaningful way to show how much money is lost. People’s time can rarely be aggregated in the same way that you add up the cost of wasted power from computer monitors left on overnight. Any attempt to do so is a kind of fuzzy financial logic that says nothing about value.

So how does this relate to TCO and ROI? I contend that many of the calculations of both metrics are examples of the same spurious accuracy as the lost time calculations because they are usually founded on models that are based on idealized circumstances and shored up by faulty logic.

At the heart of the issue is that business is messy. You can build any model you like to analyze the cost of business, but at the end of the exercise that’s all you’ve got – a model.

For example, think you’ve got the cost of upgrading your desktops to the new release of some software package nailed down? How wrong you could be. You may well find that even though the software installs easily, on a significant percentage of desktops it changes a registry setting or overwrites a DLL and, poof! Half of your desktops become unusable.

So after unwinding the updates, you find that you just needed to change a configuration parameter, re-install the damaged application or sacrifice a chicken as you cold boot. Nothing to it, it just wasn’t factored into your model, thereby invalidating your calculations.

And over the life of a product estimating TCO or ROI will, most often, be very difficult. Your calculations will be based on idealized models of your business processes, and those models will be mostly wrong in many important aspects.

One of the main reasons that models of business processes will be out of sync with reality is that IT can have only a broad, simplified idea of what business processes actually involve. And much of this limitation is because computers usually support business rather than actually being the business.

I’m not saying you should completely dispense with TCO and ROI. What I suggest is that you treat them with disdain and suspicion, knowing that if you have to stake your job on their accuracy without being able to define value, you might as well start polishing up your resume.

Gibbs is a contributing editor at Network World (US). He is at nwcolumn@gibbs.com.