People are picky by nature. What appeals to one person may not look as good to someone else. Unfortunately, when it comes to web design (and with most other things in life), you can’t please everyone—although you can certainly try.
In every scenario, there’s going to be a majority group and a minority group. Try as you might, you can’t get everyone to love your site design or layout, but you can try to make it as pleasing to as many eyes as possible. One technique you can use to do this is A/B testing.
Image Credit: Renjith Krishnan
What is A/B testing?
A/B testing, or split testing, is a method that you can use to test out different versions of a single element on your site to see which one produces the best conversion results. The name of the technique itself is self-explanatory: you present choices A and B to your audience, and see which one of the two gets picked the most.
How do you do A/B testing?
A/B testing allows you to go about website optimization in a step-by-step manner. If you’re going to be strict about it, then you can actually only change one element at a time. However, many webmasters decide to change several variables at a time at the beginning of the entire process. This lets them find the most promising design in a shorter amount of time. After several rounds of this, the actual single-variable testing is implemented to further fine tune the design.
Image Credit: Grant Cochrane
Step One: Choose which variables to test.
So you’re at the stage where your site is ready to go live. This is both an exciting and stressful moment, because while you’re happy that the site you’ve been working on for so long will be launching, you will also be worrying about those elements that you’re not very sure of. As you’ve probably already surmised, these are the elements that you should create variations of for your test.
Evaluate your website and list down which elements could use some improvement, or which variables can still be manipulated. In fact, you might have two ideas for one element of the site that both seem like good ideas. Narrow down the variables you want to test out, and create the variations.
Here’s a list of elements that you can consider testing on your website:
• Header – Size and positioning of the logo, margin and page border
• Navigation menu – Sizes of the menu bar, font size and type for the text, menu style (ie. drop down, horizontal, or animated)
• Content – Wording of the text on your site, tone with which the call to action was stated, font size and type used for the content
• Banners – Sizes and positioning of the banners, types of images used in the banners, inclusion or exclusion of a call to action in the banners
• General elements – Layout of your site, the positioning of certain elements, the inclusion or exclusion of features (ie. a search bar, a subscription form, etc.)
There are many more elements that you can test out on your site that aren’t included in the list. It’s important to remember that you really aren’t limited by any means on how you’d like to manipulate or modify each element that you want to test out.
Image Credit: arztsamui
Step Two: Create your test.
Once you’ve chosen which element (or elements) to test, create the variations that you want to try out. When you’re ready to do the test, you can set up a script on your site’s server to display the two versions of the site randomly for a certain period of time. You can have your programmer write a script that will record information that you can use to gauge which site was able to illicit a better response or reaction from your visitors.
You can also opt to enlist a third-party provider to set up the tests for you. In this case, you will only to provide the set of elements to be tested, as well as a criterion that can be used to gauge which design is the better one.
Step Three: Interpret the results from your test.
When the testing period is over, it’s time to take a look at the results and interpret them. In general, webmasters use conversionrates as the main factor in determining which of the variations is the better one. It’s important to have a large enough sample size so you can get data that is more reliable. Ideally, it is recommended that you push for a statistical significance level of 95% in your test.
The conversion itself can be taken to mean a lot of things. If you’re an online retailer, then this could be the number of sales made or clicks generated for each version of the site. If you’re a content provider or if you run a blog, then you might want to consider which version was able to make visitors stay longer and browse more pages.
After you’ve determined which version is the better one, then it’s time to implement the changes. However, if there are still more elements that you want to test out or if you want to optimize your site further, then you can do the tests all over again.
Image Credit: David Castillo Dominici
Do’s and Don’ts of A/B Testing
Here are some things to keep in mind when you’re running your A/B tests:
• Do conduct a series of A/B tests to fully optimize your site and come up with the best possible version of it. Doing just one or a couple of tests can help, but doing more at regular intervals is better. The Internet is ever-changing and new trends in web design, layouts, and applications are always being introduced. Because of this, there is a need to always keep your site updated and optimized.
• Don’t test out more than one element or variable at once. As mentioned earlier, you can test out several variables at the same time when you’re conducting your first few A/B tests. However, it is recommended that you then proceed by testing just one element at a time so you can eliminate the need to guess which variable was able to produce the change in conversion.
• Do set an appropriate duration for when you’ll be running your test. If it’s too short, then you might not have enough data to draw a reliable conclusion. If it’s too long, then you might already be running at a loss if the variation that you’re testing out is poorly implemented.
• Don’t test just one version of the other. The point of A/B testing is to test both the variable and the control simultaneously. At any given moment during the test, you should have two versions of the variable (or variables) being tested that should be displayed randomly to your visitors.
Guest Post By Mark Simpson