Testing is one of the most important techniques when it comes to website optimization. It shows us what the users like, how they react to design, which content they are looking for, which offer is more attractive, and so on. However, "users" are not a bunch of robots with the same needs and tastes. Each visitor to your website is a person that lives in a certain country, works in a specific field and got the word about your website from a different source. All these affect profoundly the way they behave in your site and whether they will buy from you or not.
Avinash Kaushik writes about segmentation analysis, a fundamental practice for website testing. If one user comes from London and one comes from San Francisco, how can you have them both in one test? Most probably, they will respond differently to different designs. Even though they speak the same language, they live in very different cultures and should be treated differently by you. Below are a few simple segments that may show significant different behaviors:
However, choosing the right segments is not a trivial endeavor, it takes a lot of thinking. In the Online Marketer Segmentation Guide (link to pdf), a paper published by Omniture, the following are quoted as the requirements for successful segmentation:
These guidelines are very useful when deciding which segments should be chosen for testing and optimization. But the choice must also take into consideration the ability to manage the segments. As the paper explains:
The truth is, while one-to-one marketing is a worthy goal in many cases, it just isnít practical. Just as the catalog or direct marketer looks at cluster codes, lifestyle overlays, the online marketer needs to create meaningful groups or segments that are significant as well as manageable.
The most common practice nowadays is to launch tests that includes all the visitors of a website. While aggregate AB and Multivariate tests are a great start, it is far from optimum. When you aggregate traffic, you are optimizing for the lowest common denominator, not for the highest conversion rate. This leaves a lot of opportunities on the table. Maybe this is still a common practice because there was no easy (and free) way to segment traffic and test only a few specific segments.
However, there is one relatively new and simple way to segment your tests by a series of variables...
BTBuckets, a free Behavioral Targeting and Segmentation tool, has recently launched a Firefox plugin that enables the segmentation of Website Optimizer tests. The plugin enables using specific segments created through BTBuckets tool on Google Website Optimizer experiments. This can be highly valuable, for example, if you want to test only US visitors, or returning customers, or a specific traffic source, or almost anything your imagination can come up with. From within the GWO interface, you will see a bucket button that will simplify the tags implementation process and allow the segmentation of your testing. An example of how to do it can be seen in the video below.
One website for everyone is a short-run solution. In the long run, websites must meet the needs of the customers and provide them with what they are looking for, not with what is the best for the average user. That's why techniques like Testing, Targeting & Segmentation are essential for website success.