Taguchi Testing is now a remarkably common landing page optimization tool. A good deal of marketers are wondering what the heck is Taguchi Testing and is it all that it's cracked up to be? This article will answer all of your questions regarding Taguchi Testing with simple answers.
The Brief answer is:
Taguchi Testing is Fractional Factorial Multivariate Testing and is not a fantastic tool for landing page optimization. Let us elaborate on this.
What Is Taguchi Testing?
Taguchi Testing is a variation of fractional factorial multivariate testing which was designed from the 1950's and 1960's with a Japanese mathematician called Genichi Taguchi. His testing procedures, which were initially developed to improve manufacturing quality control, have gained recognition within the field of landing page optimization.
Multivariate tests are just tests which have more than one input variable. Complete factorial tests are tests that analyze every possible combination of inputs variables. Fractional factorial tests attempt to isolate and test only the subset of inputs which are deemed beforehand to be important into the output. The Taguchi method is a set of partial factorial methods that try to find out the best combination of characteristics in the presence of a lot of variance or noise.
As mentioned, the Taguchi method was designed to be used in manufacturing quality control. The Taguchi method has now been embraced as an instrument for landing page optimization. The differences between a manufacturing floor and also a landing page make Taguchi testing a wrong tool for landing page (LP) optimization. ) This will be examined later in this article.
Full Factorial Testing
The most famous landing page optimization tool that uses complete factorial testing is the Google Web Site Optimizer. This is probably the best free application available for multivariate testing (testing lots of factors simultaneously ) on landing pages. The Google Web Site Optimizer will statistically determine that combination of LP factors are likely to create the highest amount of conversions. The Google Web Site Optimizer in its present state does not analyze interactions between the variables being analyzed but merely tells how every analyzed combination of factors performed overall as compared with others.
The Google Web Site Optimizer is also a great tool for A/B split-testing. This entails testing only 1 variable. Typically only two variations of one variable are being analyzed against every other. One variation is announced the winner once the Web Site Optimizer calculates that it's attained a proven percentage certainty that it works better than the opponent. For instance, we might declare a variation to be the winner the moment we're 80% certain that this variation outperforms its opponent.
For A/B split-testing, I prefer to use an Excel model that performs exactly the exact same statistical test since the Google Web Site Optimizer (a one-tailed, two-sample, unpaired hypothesis test of percentage ) but doesn't require any of the set-up steps that the Google Web Site Optimizer does. My website has a link which will make it possible for you to download which Excel Split-Tester.
Taguchi Method Drawbacks
Taguchi's method of fractional factorial testing for LP optimization has important drawbacks in contrast to full factorial landing page testing. They are as follows:
1) Fractional factorial procedures assume that interactions between factors do not exist. This assumption is completely invalid for landing pages. Quite strong factor interactions normally do exist on landing pages. For example, any LP designer knows that a mismatch between a headline and the body will wipe out conversions.
Typically you will find lower order interactions occurring on landing pages. Reduced order interactions are interactions which occur between a small number of variables, usually 3 or 2. Higher order interactions (interactions between more than 3 factors ) are less common and generally not as important. In order for an interaction to be material and significant, one of the variables usually has to be important on its own If you ignore interactions during landing page optimization, you will probably not get the best outcomes.
2) Fractional Factorial techniques can only be used to check a small number of landing page mixes simultaneously. Typically the top limit of the number of separate landing page combinations that may be tested simultaneously using fractional factorial approaches is several hundred. Brainstorming marketers will quickly hit this limit after coming up with just a couple of factors and a couple of variations of each variable.
A few landing page optimization terminology should be presented here. A variable or variable is an element on the landing page that you're varying throughout the exam. A value is one of those countries that a variable or variable (these 2 terms both mean exactly the same) can occur during your evaluation. The branching variable is the range of values that an single variable or variable can take. Each variable has its own specific branching factor. A recipe is a exceptional mix of variable values offered for a test. Another way of expressing this point (#2) would be to say"Fractional Factorial methods can only be used to test small number of recipes simultaneously."
3) Fractional Factorial approaches are exceptionally restrictive to test layout. Frational factorial methods don't allow the evaluation designer substantially freedom when choosing the amount of variables or the branching factor for every factor. The Taguchi method utilizes a matrix arrangement that works with less than two dozen very particular combinations of number of variables and branching level for the factors. The test designer has to construct the test using one of these combinations of factor levels and branching factors. Total factorial methods have none of these limitations.
4) Fractional Factorial methods require imagining where variables to include in test. The restrictive character of Fractional Factorial test layout requires that the test designer select the factors that he or she considers to be most important. The single biases of this test designer will impact the choice of elements to include in the evaluation.
5) Allocating more bandwidth to the baseline isn't possible with Taguchi. The baseline is the present recipe which we are trying to beat with new recipes. It is very important that measurements of the baseline be legitimate since these measurements are the foundation for comparison against results obtained for every single recipe tested. To guarantee validity of the baseline's measurements, it is a fantastic idea to allocate at least 15% information collection (bandwidth) to sampling the baseline recipe. This sort of data throttling is not possible with Fractional Factorial methods such as Taguchi. It's easily done with Full Factorial evaluation methods.
Reasons For Your Taguchi Mismatch
Genichi Taguchi developed his testing approaches in the 1950's and 1960's to improve superior control on the production environment. His methods have become popular now in the specialty of landing page testing. The differences between making environment, for which the Taguchi method was designed, and now's landing page surroundings produce the mismatch which produces the Taguchi not the best choice for LP optimization. Listed below are the main reasons for the mismatch:
1) Expensive manufacturing prototypes vs. free landing page prototypes. Retooling a production line to get a brand new recipe is pricey. Among the greatest aims of Taguchi was to maintain testing cost down by reducing the amount of recipes to your minimum. In landing page testing, there is not any additional cost to make more recipes (new variations of a landing page which will be revealed to site visitors).
2) Manufacturing costs require a little test sizes, infinite landing page evaluation sizes. The high prices of manufacturing prototypes made little test sizes required. The Taguchi method keeps evaluation size small by guessing at and analyzing only the most important facets. On the other hand, Full Facorial landing page testing procedures and the low cost of producing new landing page recipes empowers simulataneous testing of millions of recipes.
3) Small manufacturing test sizes couldn't test, and therefore did not presume, interaction between factors. Landing pages are proven to have very strong interactions between variables. The Taguchi method was designed to assume no interaction between factors. That premise can easily result in incorrect results during LP testing.
4) Manufacturing environment tests are smaller since statistical significance is normally reached quicker. Landing pages generally have low conversion rates and therefore need much bigger test sizes to reach statistical significance. Manufacturing environment evaluations normally are intended to have a higher probability of success. Landing page success rates (conversion rates) are generally under 1%.
5) Manufacturing evaluation information is often constant vs. Landing page information which is discrete and unrelated. Continuous data allows the test researcher to have a smaller number of samples and interpolate effects for intermediate data points that were not gathered. The possibility of interpolating constant variable test results allows for smaller test sizes. Landing page factors are generally discrete, unrelated choices and therefore and don't permit interpolation for intermediate data ranges that were not collected.