Multivariate Testing Using Taguchi Method

The biggest diffrence between GWO multivariable testing and the commercial MVT solutons like Offermatica, Optimost, SiteSpect is the use of Taguchi Method/Fractional factorial methods.

Taguchi methods are statistical methods developed by Genichi Taguchi to improve the quality of manufactured goods and, more recently, to biotechnology, marketing  and advertising. Learn More about Genichi Taguchi at Wikipedia

In multivariate testing….if there are 30000 combinations to be tested, using taguchi method, we can select only 10 combinations to test and predict for the rest of them And finally predict which perform the best out of all 30k.

This method has been in use in non-it industires. What people are saying on web is taguchi is an accurate method for selecting the samples, the 10 they select using this method are logically represent all 30k combinations so it is like testing all 30k but not sampling.

All we need now is to find a tool or tip to that can predict which 10 combinations need to be tested….watch this space

Sphere: Related Content


You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.

AddThis Social Bookmark Button

4 Responses to “Multivariate Testing Using Taguchi Method”

  1. Vijay,
    Love the attention on Taguchi. As a clarification, GWO does not employ the Taguchi method, instead using a full-factorial approach. Nor does Optimost, which uses Optimal Design. Vertster and Offermatica do.

    Optimal Design and Taguchi are both “fractional factorial” approaches, meaning that you test a subset of the total permutations (or “recipes”) and analyze the results to determine the best recipe and the contribution of each element in the test.
    Matthew Roche

  2. Thanks Matthew for making it clarify. What I don’t understand is with GWO we would need to test all combinations but any of the full-factorial approach we would test only a subset of total recipes. You were mentioning GWO uses full-factorial approach, I think that is predecting the performance of the variations, or does that also limit the no. of combinations to be tested?

  3. Story added

    Your story was featured in WebSense! Here is the link to vote it up and promote it: http://www.soundsgood.in/dpl/OnlineSearchMarketing/formula_for_search_engine_optimization

  4. Story added

    Your story was featured in WebSense! Here is the link to vote it up and promote it: http://www.soundsgood.in/dpl/Design_Usability/Hahaha_kichkcihckch

Leave a Reply