Ever since I first read about the Taguchi method in Robert X. Cringely’s PBS column a year ago, I’ve been fascinated by it.
Since then I’ve seen a few more references to it on the Web, but frankly I think I need an engineer and a statistician to explain the formula to me.Here’s the theory behind Taguchi: Let’s stay I’m trying to design a high-quality chair. I can make lots of decisions when I design it: the type of wood to use, the brand of sealer to use, and what color to paint it. Now any idiot ought to know that white paint versus green paint should make no difference on the durability of the chair. But if it’s less obvious, how do you figure out what variables matter?
Enter Taguchi. Mathematically you can figure out which variable matters, so you can reduce the number of tests. Let’s say I have 12 different types of wood I can buy locally, 12 brands of sealer, and 144 different paints. If I’m remembering high school algebra correctly, that gives me 20,736 different combinations to test. Taguchi will let me quickly figure out that the type of wood I use is the most important factor, letting me build and test 12 chairs instead of 20,000+.
If you’ve noticed that cars are a lot more reliable today than they were 20 years ago, Taguchi has a lot to do with it. The Japanese learned about Taguchi first, which was why Toyota and Honda came from out of nowhere. I remember when people made fun of Hondas. Nobody does that anymore.
Cringley talked over a year ago about applying Taguchi to advertising. Well, I found a free Taguchi ad comparator. That’s a lot better than the $499 product he was talking about a year ago.
I haven’t had time to test it yet, but if I ever decide to try to sell something on Ebay, I’m going to search for old ads for the same thing and run them through it to see if I can figure out why some listings fetch more bids than others (assuming it isn’t an obvious reason, like the item being in poor condition, having a bad photograph, or being listed in the wrong category).