Facts are hazardous in this business, and I do my best to avoid them. Opinions are much safer.

Occasionally, though, something tangible belongs in the dialog, so here’s what an InfoWorld reader has to say. I’ve done a bit of editing to ensure anonymity. Otherwise, the text reads pretty much as I received it.

“I run IT for the Acme Corporation (yes, the one Wiley Coyote orders from). We’re semi-autonomous from the corporate DP department. DP wants to correct this condition. We don’t.

“My area manages 120 PCs, and have a total budget of $1.3 million, so there’s your $11K per PC. ‘Grand total’ includes a $400K tithe to DP (they figure their costs and send a non-negotiable bill). Do I split it out as ‘mainframe’ costs, ‘network’ costs, ‘PC service’ costs? No way. DP pretends to, but admits the numbers are largely made up.

“(I sometimes wonder how much is going to pay for their middle managers sitting around talking about ‘architecting the meta-data in the Zachmann Framework’, but then, I’m sometimes overly pragmatic.)

“Even in my own area, how do I separate the costs of projects from the infrastructure to make the projects work? Not everybody really needs networking, yet everybody gets Banyan, TCP/IP, SQL*Net middleware, because it’s cheaper, smarter, and pro-active to make sure anybody can run anything.

“And besides, the Big Bosses don’t want to hear breakdowns into cost categories they don’t understand in the first place. They take the grand total IT budget including all salaries and divide by the number of seats.

“And now the real bottom line for the big cheese: HE DOESN’T CARE. He has a $100 million/year business to run, so I’m 1.5% of his budget. He’s being nice to give me about 3% of his time, and he only gives me that because my area is ‘strategic’ and rapidly growing. As a line-business person, he’s much more concerned about the benefits than the costs, and he can see with his eyes closed that the benefits are much greater.

“It’s the full-time DP types who aren’t even connected to the business world who care about the costs. Every new machine, every new application, every new layer of middleware is another chore to them, another cost, another problem. They want to hold these down and speak much of simplification, homogenization, and limiting numbers of tool sets. Users agree in principle, but principles drop to the bottom of the priority list when a useful new application for their machine appears.

“I restrain the most enthusiastic end-users, but basically I’m on their side – even though their new uses are also new problems for me. I’m out in the business area and can also see the benefits.

“Sorry for the rant. I’m just trying point out the complexity. The accounting is the least interesting and important part.”

When you put your mental graphs on your mental wall, which of these measures do you chart: corporate value, IS costs, or your migraine level?

Look, I sympathize. Your headaches are my headaches. I feel your pain. (For that matter, my knee hurts, but that’s another subject.) Unfortunately, you’re paid to have headaches, and a reduction in your headache load doesn’t automatically translate to corporate value.

Reducing your costs doesn’t translate to corporate value either. If it did you’d resign immediately, reducing the cost of you to zero. You’d get rid of your headaches, too.

IS typically represents between 1% and 3% of a company’s total budget. Now be realistic here – how much of that can you really save by reducing your PC cost of ownership? The guess from here: even if you replace all your PCs with NCs and work really hard, your savings won’t amount to even a blip on the radar screen.

You have to decide where you’re going to focus your attention. You get much more leverage out of value creation than cost reduction. As reported last week, I once generated a 4,000% ROI with a single spreadsheet. Think of how much return you can generate by providing really great support for all of your company’s spreadsheet users.

And that’s just the tip of the old iceberg.

My first involvement with financial modeling was the result of procrastination.

The guy on the other end of the phone was the budget manager, asking me why I hadn’t filed the production budgets. Not wanting to tell him I’d forgotten all about it, I thought fast and responded, “How can I give you the production budgets when we don’t have any volume projections from Sales?”

This generated a good-news/bad-news situation: I bought myself a week, but I had to build a really nasty financial model of production. It reduced our unaccounted budget variance from over 10% to under .5%, and also produced one of the best lines ever heard from a production manager. When asked to explain what caused the remaining half-percent variance (which was favorable) this gentleman quietly answered, “Good management.”

With this model, we demonstrated that hiring another thirty union workers would save the company nearly $200,000 per year. I don’t care what that PC cost the company – the ROI was awesome.

The model was so successful that I ended up doing a bunch more financial modeling. One of the things I learned: When to include only marginal costs, and when to use fully loaded costs.

Which points out one of the basic flaws with current Total Cost of Ownership (TCO) models.

Yes, it’s time for more of my continuing tirade on the subject of how much it really costs to put a personal computer in front of an employee. My last column generated quite a bit of correspondence. With one exception, the letters I received unanimously agreed that it would take an act of Congress or military procurement to drive expenses anywhere near the $10,000 to $12,000 now reported as annual costs. The exception found himself rebooting several times a day, and clearly anyone who drives a lemon experiences a high cost of ownership.

What does the difference between marginal costing and full cost loading have to do with the cost of a PC? If you’re unacquainted with the jargon, “marginal” in this context means “incremental”. Marginal cost accounting only takes new costs into account. When you fully load costs, you include everything, including costs you’d already experienced but which also apply to the new item you’re analyzing.

From what I can tell, TCO models use full cost loading. They ought to use marginal costing, because otherwise they’re including money spent before PCs came in the door. That’s wrong: any money spent before PCs showed up isn’t a cost of the PC at all. Seems obvious, doesn’t it?

So let’s start listing what we should leave out of our cost model. Item one: the cost of futzing.

Futzing allegedly extracts a high toll – potentially hours per week goofing around installing screen savers, changing Windows wallpaper and so on. If it’s an hour per week, that would be about 50 weeks X $50/hour = $2,500 per year in costs.

As it happens, I take the futz factor personally, because readers tell me they sometimes post copies of my columns on their cubicle walls. What’s the connection? Putting Dilbert, Cathy, pictures drawn by your kids, and my column on the wall amounts to futzing, too. So here’s a magic question: Does management in your company worry about the time and resources spent in non-technical futzing?

There are only two answers: yes and no. If the answer is yes, then the futz factor shouldn’t be included in the cost of the PC – it’s the cost of futzing, however it’s expressed, and it isn’t different just because it’s on the PC.

If, in contrast, the answer is no, then the futz factor shouldn’t be included in the cost of the PC. Management doesn’t worry about time spent futzing anywhere else, so why should it suddenly matter when it’s on the PC. Going back to our cost accounting jargon, futzing isn’t a marginal cost, so it shouldn’t be included.

There’s a bizarre mentality that comes into play when people put financial models together. They acquire selective amnesia, suddenly forgetting everything they know about how real people behave in real offices. Including the futz factor in the True Cost of Ownership is an example.