How statistics manipulate you

14.11.2006

Look at the Fibre Channel switch market ; there have been three players for the last three years: Brocade, McData, and Cisco. While I'm no statistical whiz, I know that in your shop you run McData in the core, connected to your mainframes. You probably run Brocade switches. You might be talking to Cisco, or even have implemented their director in non-mainframe environments. So how come the Fibre Channel networking market is listed as a $2.7 billion space? I add up the total revenues of Brocade and McData I come up with about a $1.3-1.4 billion. I know anecdotally that Cisco adds $300-400 million to that - so even if you call it $500 milion, it's at most $1.9 billion. Where's the $800 million? Or are those the invisible switch ports?

Since we never say what we mean in the IT industry, why should the numbers be any different?

The numbers should be a guide, but if they are just stupid, you should ignore them. Don't do anything because of some statistic alone; use them as a reference to point you in a direction. Real references from real people are better ways to buy anyway. Real people don't speak in statistics (normally). They say things like "that product sucked. It caused a fire alarm" or "this was the best money I've spent in years."

Now, having said that, you should absolutely be thinking about creating classes of files (or blocks, for that matter), and putting some high-level assumptions for each class down on paper. Then you should socialize those assumptions with other IT folks, business line folks, and even mucky mucks if you have the juice. I guarantee it will be a worthy exercise, and you'll find out that there are a lot of misaligned assumptions. Just being able to get folks back on the same page makes it a worthy endeavor. The mucky mucks probably think you already have your file data completely under control, and that Mr. Spitzer will never be able to find fault in the system. Or, you may find they want to begin the exercise and start to solve some real problems. I figure you have about a 4 percent chance of that.

Send me your questions -- about anything, really, to sinceuasked@computerworld.com.