Slashdot Mirror
Building Better Spam
henbane writes "Cringely is plugging a new method of advertising from Dr. Jim Kowalick and Mario Fantoni. Their book entitled 'E-Mailing Your Way to Sales With
the Taguchi Approach' is out in the autumn. What could be worse than a method which increases the returns on spam?"
Thank you. That's exactly what I was about to say. I'll be the first to pull the trigger when we get the spammers against the wall, but just because it's email marketing doesn't make it spam. I get plenty of marketing mail for games, telescope equipment, and other stuff I'm actually interested in because I opted in to the lists after reviewing their policies on sharing my address and confirming that they wouldn't. This prevailing idea that every commercial use of email is spam is raving nonsense. It is sad that the spammers have managed to so thoroughly hijack people's perception of what can be a useful marketing vector.
Trouble making decisions? Just flip for it.
Because if they do, they'll figure out quickly that sending me spam won't increase their returns.
Spam isn't hated because it is targetted advertising; precisely the opposite - SPAM is hated because it is untargetted. That is, people get spam for things that they would never buy. Personally, I do get targetted emails - I've given my address to local retailers, and I get their specials via email. I'm not annoyed at them. I'm annoyed at the folks who spam me with stuff that I would never even remotely be interested in.
If making spammers more effective means that I won't get 50 emails a day for stuff I'll never buy, I'm all for it. If it means that I'll get discounts for stuff I do buy, then I won't mind too much.
Cringley points out how standard engineering tools, in this case Taguchi's Design of Experiments (DOE) methods, can be used to increase the effectiveness of advertising. Claiming that he "plugs spam" is a complete mis-reading of the article. He points out the original study used "spam" in order to prove it's effectiveness; the study isn't dated.
DOE is how engineers make complex design decisions with as few experiments as possible. Mostly, he uses eBay as an example. He slightly mis-reads what Taguchi's DOE is about when he says that the old eBay data can be mined to re-create an orthogonal array. The whole point of DOE is a priori deciding what experiments to run, instead of the shot-gun approach used in the past. If you're gonna use data mining, then you don't really need Taguchi excpet for data reduction.
Personally, I recommended this approach to a high-volume eBay seller a couple years ago. He sells widgets with 3-4 different features (style, size, color), and uses a variety of terms to describe them (i.e. [stunning|beautiful|awesome] [rare|unique|one-of-a-kind]). Basically, he could run 16 or so tests using these various terms in the right combination, and determine which combinations were likely to work best. Ultimately, he didn't go down that route, but I'm pretty sure this is what Cringley was getting to before he got it confused w/ data mining.
Using data mining to do the Taguchi stuff is tough, b/c there are too many uncontrolled factors. I'm sure he'll get 100 letters on the topic from DOE experts and write a follow-up column next week.
As for spammers, I bet they start using DOE techniques, as they'll have to as fewe & fewer emails are getting through, making it a less profitable venture. Of course, legitimate advertisiers should be using the same techniques, and maybe they do. But DOE can be applied to any process, whether it is building cars, designing rockets, baking cookies, selling on eBay and, yes, sending spam.