Ecommerce Should be EZ

EZ Split Testing - First Steps

Obviously testing is critical for effective ecommerce -- and you've prbably heard a lot of noise about split testing (usually called "A/B Split Testing"). As you may have guessed, split testing can be as simple as randomly rotating between several sales pages to see which one is most effective.

Unfortunately, the field of split testing has become very cloudy because of all the vendors trying to differentiate their particular split testing product. Mr. Taguchi did not help the situation by applying matrix algebra to create a multivariate testing methodology (the "Taguchi Method"). Zealots of multivariate testing highly fine tune each aspect of a sales page to maximize action rate.

All of this is well and good but for the fact that the initial learning curve for multivariate testing is pretty high. If you are already split testing and wish to explore multivariate testing -- by all means do so. But if you are still 'flying blind' with no split testing at all because of the complexity... well, let's get over that right now.

Every ecommerce company should split test. The only questions are 1) which pages to test and 2) how to get started initially.

Pages to split test

Clearly we want to split test all of our major product sales pages. But hopefully you've been convinced to build an an email-marketing-centered ecommerce machine -- and therefore you will want to carefully split-test your name capture pages. (A "name-capture" page is a landing page which convinces the visitor to exchange their email address for a special freebie or bargin.)

If you follow an advertising model (say, by using Google Adsense to earn revenue), you might want to create a split-test to measure different Adsense layouts for clickthrough effectiveness.

Getting Started, requirements

So what do we need for very basic split testing? Let's list our simple requirements:

• Rotate pages randomly but show the same page to the same person
• Record unique visitors with the "displayed page"
• Record actions (say a sales click or form submit)
• Report with each test, and the stats for each displayed test page

And, just for icing on the cake:

• Self-optimizing by displaying the "best" page most of the time

Roll your own or use something "off the shelf"

As you can see, basic split testing need not be terribly complicated -- and yet it can make a huge difference if your site's effectivness!