April 2003

When is cheaper not really cheaper? That depends on what the definition of the word "cheap" is.

One clue as to what "cheaper" means is to remember that in any calculation of Return On Investment (ROI), the time factor cannot be ignored. What good would it be to report that the ROI of an investment is 250%, but it will take 75 years to achieve that rate of return? Such an investment might sound wonderful if the number 250 is emphasized, but ignoring the 75 is foolhardy. You can get a better annualized rate of return by buying bonds!

It is vital to any technology cost comparison to calculate all returns on an annualized basis, so that you will not be doing an apples-to-oranges comparison. Buying a piece of software at 50% discount from list price is not necessarily a bargain. What happens if it takes five years to recoup the investment, but the vendor cancels support after three years? Then you might feel compelled to buy the next version before the five-year break-even point has been achieved. Not very smart, is it?

You might want to get the vendor to bring the price down further, but why bother? Do you really want to spend money just to break even? Once again, it is simply foolish to buy software, use it for a couple of years, and then throw it out when the vendor dictates that the next version must be purchased (or else an add-on support contract must be signed, after the ROI calculations have been filed away and the extra cost is thus ignored). Any software purchase that has a break-even period of more than a couple of years is essentially throwing the company's money out the back door by the shovel-full.

In order to make a positive impact on company profits through implementing a software package, the product lifecycle must significantly exceed the break-even period. That means that the smart purchaser will know the break-even period and will negotiate a firm commitment that vendor support and free upgrades will continue throughout that period, as well as a few years beyond it. Otherwise, you might be better off telling the boss to just invest in some bonds.

Now comes the hard part. What if that vendor is Microsoft, and that product is Windoze? As far as I know, nobody has provided a solid, realistic analysis to prove a net positive ROI over the lifecycle of a Windoze version. Sure, some slick consultant might say, "Buy this new version of Windoze, and it will save you 20% compared to the previous version of Windoze." But that is a false comparison. What if that previous version had a negative ROI? Without knowing what your baseline or benchmark of comparison is, how do you know that you're not simply buying another product with a slightly-less-negative ROI? And if that is what is actually being sold, then anyone who buys it is just throwing away a little less money than last time.

As one example of how subtle this whole process is, Microsoft recently commissioned a survey that claimed that Windoze is cheaper than Linux over a five-year period. Research group IDC compared about 100 companies and claimed that, for various tasks and applications, it was cheaper to use Windoze2000 since supposedly the support cost was so much lower than Linux, that after five years the zero-entry-cost advantage of Linux disappeared. But there is a hidden assumption involved, isn't there? Doesn't that comparison presuppose that Microsoft will continue to support and sell Windoze2000 for five more years -- just to get you past the break-even point relative to Linux? And that five-year window must be counted from when your company actually implements the new regime -- not just when Microsoft begins selling it.

I would like to see a comparison that showed three, five, and ten years' breakpoints comparing Windoze2000 to OS/2. But that will never happen. OS/2 eCS, for example, will not force a new licensing regime on your company every two or three years, whereas Microsoft has announced plans to discontinue Windoze2000 availability on March 31, 2004. (System builders can get it for one additional year.) Now why in the world would Microsoft release a comparison touting a five-year break-even period, for a product with only 11 months left before termination? Do they think we can't count to five?

Apparently some people can count. Microsoft's own software lifecyle page shows that so many complaints were received that Microsoft had to extend their support cycle to seven years for most products. But don't think that there won't be heavy pressure to change more often than that. "Support," like "cheap," can mean almost anything that a slick salesperson wants it to mean.

Most recent revision: April 3, 2003
Copyright © 2003, Tom Nadeau
All Rights Reserved.