Is the web the ideal medium for Taguchi?

The answer... it just might be. Marketers have been playing around with Taguchi methods for a few years now, and during that time, we've learned a number of things. Taguchi DOE was developed for manufacturing environments where almost all external (noise) variables can be identified and controlled. This means things like temperature, humidity, air pressure, etc.

In marketing, very few of these so called noise factors can be controlled. In fact, very few can be identified at all. Marketing is extremely noisy. One of the best ways to combat noise is to run multiple tests against the same (or similar) sample. On the web, this is easy. Run a two week test cycle (to balance day of the week) and then repeat. Unless there has been a world catastrophe or your first cycle included the holidays, you should have a pretty similar audience.

Taguchi has also been used for direct mail and email optimization. Getting good results in this medium has proven elusive, yet there are some "experts" out there claiming to be able to do it. Personally, I have seen the results and they are dubious. The reason is simple... you get one shot with a given audience. Once you have mailed your list, it starts changing. People come and go from the list. Folks have been exposed to your brand one more time, altering the way they would react to a subsequent mailing. This is on top of the huge amount of environmental noise that is present in marketing.

Although it is possible to control all these things and get a good result in email or direct mail marketing, it is definetly trickier than on the web.