Website Optimization with Multivariate Testing

ab split testing

Source: http://opensource.washingtontimes.com/blog/2011/jan/25/4-lessons-learned-about-multivariate-testing/

You’ve already launched your website, where the design process involved poring through books and consulting with professional designers. But there are obviously some things that still aren’t working at this point, because your sales are coming short of expectations.

At this point, you’ve got to stop and think about your next step. Don’t immediately assume that what you need is more traffic; what’s equally important is figuring out a way to increase the conversions from your current traffic. This time around, you might be better off asking for the opinions of the people who matter the most when it comes to your site and your business: your site visitors.

You could do surveys, but the odds that a good number of people will respond are slim. Describing how a website would look better in words and collating data from such surveys would undoubtedly require more effort than necessary, and the results are mediocre at best.

What is Multivariate Testing?

This is where multivariate testing comes in. It’s a technique where you ask your users which version of your website they prefer quite indirectly, because you’re basing your results on the response that your website will elicit from them. It’s one of the test methods that webmasters can use in behavioral targeting to offer personalization to their site visitors as well.

Multivariate tests are like several A/B tests that have been combined into one. A number of elements are tested out in combinations for each test run, so webmasters can come up with an optimized design in a shorter period of time.

Where Can Multivariate Testing Be Used?

You can run a multivariate test on all sorts of online content and pages. Here’s a quick list to show the diversity of applications for this method: landing pages, search results, sign-up forms, navigational menus, category pages, promotions tests, product pages, and multi-stage checkout pages.

Aside from websites, multivariate testing can also be used for email campaigns and newsletters.

Advantages of Multivariate Testing

As mentioned earlier, multivariate tests can be used for website and email marketing content. By optimizing the content that is being used for these campaigns, conversions can be increased, whether it’s in the form of clicks or sales. Aside from that, two other benefits that multivariate testing offers is an accelerated learning curve and breakthrough thinking.

Multivariate Testing

Source: http://www.websiteoptimization.com/secrets/metrics/10-4-website-optimizer-diagram.html

How is Multivariate Testing Done?

Multivariate tests involve four basic steps: assessment, design, testing, interpretation and implementation.

Site assessment. Before anything else, the content to be optimized must be assessed critically. In this stage, major design flaws are spotted and weeded out. This is a crucial step of the process because this is where elements that are to be changed are noted.

Test design. Once a list of elements to test out has been generated, it’s time to design the multivariate test. This involves grouping together elements that can be tested out together in batches. Alternative designs and variations of the elements are then created. It is recommended that the original and alternative versions to be tested vary to a noticeable extent.

Testing. During the testing period, different versions of the website will be served randomly and simultaneously to site visitors. The test will be run for a certain period of time, where relevant data regarding the response of the users will be recorded.

Interpretation and implementation. Once the test has been completed, the results are then processed. This is the stage where the data is collated and the website version that was able to get the best results is implemented. Further multivariate tests should then be conducted to continue optimizing the website or marketing emails.