All Posts in Technology

|In: Strategy and Tactics|Last Updated:

Dear Dad,

Thanks for the e-mail. I enjoyed the joke, although I don’t really get the part about the nun and the rabbi. What you described them doing seemed out of character somehow.

Glad to hear your European trip was such a success. A German version of our new book, Selling on the ‘Net would be great, but since neither of us speaks German I’m not sure how we’d collect and analyze a bunch of German Web pages. Any ideas?

Speaking of the book, I know I promised to shamelessly promote it in my column. It just seems so crass, weaving in the title and publisher and pretending my readers really need to know it’s called Selling on the ‘Net by Herschell Gordon Lewis and Robert D. Lewis, published by National Textbook Company, for $39.95.

It would be one thing if there was any possibility of subtlety. Anyway, I’ll try.

You asked about my description of information theory a few weeks ago. I know it was a long shot. To answer your question, though … I may be, but since that’s hereditary you may not want to push the subject too far.

And besides, if it hadn’t been for the typesetting problem it would have been far clearer. Information entropy really is a great measure of complexity. Let me show you how you’d use it. You use four programs for most of your work: WordPerfect, AOL, the American Heritage Dictionary, and that CD-ROM with the world’s famous literature. WordPerfect is critical – give it a 5. AOL is important, so it gets a 3. You use the other two, but you could easily do without them, so they get 1 each. Add ’em together and you get a total of 10 points.

The formula should have been printed as:

-SUM[p(i)log(p(i))]

(You need the base 2 logarithm, so divide the regular, base 10 log by log(2) to convert it. Last time I used algebraic notation, but typesetters have a hard time with that.)

To get p(WordPerfect), divide WordPerfect’s score of 5 by the score total of 10 to get .5. Plugging all of your scores in:

-[.5*log(.5)+.3*log(.3)+.1*log(.1)+.1*log(.1)]/log(2)

which comes out to around 1.7. That’s a pretty low score – a pretty reasonable work situation. When it gets much over 2 things are getting complicated.

I know most business people avoid any statistic more complicated than a percentage, Dad, but that doesn’t have to be an absolute rule. If the best measure takes a bit more algebra, that shouldn’t be a hard barrier, should it?

Look at the formula for Return On Investment (ROI). That’s a polynomial series. A lot of people don’t understand why you put all of your costs into an artificial “Year 0” (you pretend all of your costs happen up front before you start figuring annual cash flows) but if you don’t, the polynomial could have more than one right answer. That gets hard to explain (“Well, boss, the return on investment is either 15.3% or -5.8% and we have no way to choose which one is more valid.”)

Or compound interest. You take the interest rate (plus 1) and raise it to an exponent equal to the number of years you’re compounding. People who don’t understand compound interest don’t even know if they’re losing ground to inflation, so it’s pretty important.

I know Americans take pride in being mathematically challenged, Dad. That doesn’t mean I should encourage the “Gumping of America” (GoA, to use the technical term) in my column. Ignorance may make good box office in Hollywood, but it’s a bad way to manage a business.

Okay, I’ll stop. Talk to you soon.

Love, Bob

P.S. I promise, I’ll find some way to flog Selling on the ‘Net in my column. Sigh.

Comments

|In: Policies and Procedures|Last Updated:

Journalists and professional marketers know that if all the statistics in the universe were piled in a stack, the whole awesome mass of them would lack the persuasive punch of a single anecdote. Why? Because most people, for most issues, make decisions emotionally and use logic to justify their emotional decision.

Anecdotes appeal to the emotions. Statistics appeal to reason. It’s not a fair fight.

Think about your organizational rival in the next cubicle. He picked the help desk problem tracking system your organization uses, and it’s a disaster. And the clod chose it because the sales rep happened to be pretty, and he’s having marital problems. If he’d looked at the performance statistics, industry ratings, and other objective facts instead, he wouldn’t have made such a bad choice.

We all know of stories like this, and they prove my point, don’t they?

Actually, they don’t. I just used an anecdote to “prove” my point. I haven’t proven anything. If I’d said, “The Farfinkel Institute just released a study showing that 74% of all product decisions do not include a fact-gathering phase,” I’d have proved my point (or the Farfinkel Institute would have proven it). Of course, you’d have fallen asleep in the middle of my explanation.

Next time you listen to a political debate, keep your anecdote-detectors on full scan. No matter what the issue, you’re more likely to hear illustrative stories than statistical facts in our marketing-driven policy dialogs.

Yes, I know what Mark Twain said about lies, damned lies and statistics. Yes, people can contrive misleading statistics. That’s nothing compared to their ability to script tailored anecdotes. Think of it this way: an anecdote is just a statistic with a sample size of one. You can “prove” anything at all that way.

We fall for this stuff all the time. Want an example? How much time, energy, and budget have you expended on disaster recovery planning and virus protection? Compare that to the time, energy and budget you’ve spent researching hardware design problems and software bugs, in implementing preventive maintenance problems, and in instituting administrative quality assurance programs.

Pause to reflect on this question before continuing. Then consider the following two items:

Item #1: Ontrack Data Recovery (the gods of pulling data off of trashed hard drives) recently published a study attributing 44% of all problems to hardware problems, 32% to system administration mistakes, 14% to software bugs, and 7% to computer viruses, and 3% to natural disasters. (Reported by Investors Business Daily, 9/18/96, summarized in Edupage.)

Item #2: In a recent editorial, BugNet (www.bugnet.com), made a similar point, demonstrating that problems from software bugs are 100 times more prevalent than problems from viruses.

(Why the discrepancy between the two reports? Ontrack only counted episodes leading to data loss, while BugNet tallied all problems.)

Do these two items lead you to think your efforts may be misplaced? Good. Without a doubt, computer viruses have been overhyped as a threat. They’re characterized as digital AIDS or Ebola, when in fact, as with biological viruses, most cause minor, annoying symptoms – computer colds and flu.

(These statistics, of course, don’t reveal the overall seriousness of the different threats. Natural disasters, for example, may only contribute to 3% of all episodes of data loss, but I’d bet they contribute to more business failures than all of the others put together. That makes business recovery planning worthwhile regardless of its statistical rank. Statistics devised by other people often won’t serve your purposes, which was Mark Twain’s point.)

Don’t fall for manipulative opinion-shapers who use story-telling as a substitute for facts. On the other hand, when you’re trying to persuade, make sure you do illustrate your points with examples that add some punch to your dry statistics. You need to engage both halves of your audience’s brain. That’s a matter of clarity, not distortion.

Comments