Split Testing 101: How to Get Started

TWhen trying to determine which elements of a web page will get you the best results, split testing is the way to go. Also known as A/B testing, split testing occurs when you test two versions of a web page simultaneously (say, they each have a different color scheme or headline) in order to see which one performs the best.

Usually, the original web page is A (the control) and the new design is B. When it comes to web design, split testing represents a valuable way to get insight on viewer behavior and web page performance.

Getting Started

To successfully split test, your website must have a decent amount of traffic. As with any experiment, you need a significant sample size to test on. The larger your sample size, the less likely it is that your results are a result of random chance. There is no magic number when it comes to how much web traffic you need, but typically you can get accurate results with a few thousand subjects. If you have a smaller amount of web traffic, then you’ll need to run the test longer to get an accurate result.

What Should You Test?

The web page elements you test depend largely on the goal of the web page. If you want to get people to sign up for a newsletter, for example, you might experiment with wording your web copy, a different number of fields, and whether to use a pop up or static sign-up form. Some other elements that can determine website success include:

  • Page layout: Should that picture be to the left or the right of the text? Does a double column work better than a single column?
  • Images: Do certain images perform better than others? How should you use product pictures compared to stock images?
  • Call to action: Should you change the wording of the call to action button? Should it be bigger? A different color?
  • Product pricing: Does one price point perform better than the others? What promotional offers should you use?
  • Amount of text on the page: Do visitors prefer longer or shorter text?
  • Headings and subheadings: Which text works best? Should you change the size or color?

How Many Elements Should You Test at Once?

Multi-variate testing is a variation of split testing that can achieve faster results in less time. Since just about any element of a web page can be split tested, you can change multiple elements of a web page– say a sign up button, different images, and a different headline– all at once.

The downside is that if the split test doesn’t go the way you want, you won’t know which elements are to blame for the shortcoming. You also need lots of traffic to get accurate results. It’s ultimately up to you whether to test one or multiple elements at the same time.

Create Your First Split Test

Now that you’ve got a good amount of web traffic and you’ve decided which elements to test, it’s time to create your first split test.

First you’ll need a split testing tool. If you have a Google Analytics account, you can use the Content Experiments tool. Log into your Google Analytics account and click on Standard Reporting. Under Content, click on Experiments. You can also use Visual Website Optimizer or Firepole Marketing’s free split test checker.

There are two ways to split test:

  1. You designate the two elements you want to split test (e.g., two versions of sales copy), and your tool will randomly replace the control page with the new page each time someone visits.
  2. You have your tool redirect some visitors to a new web page altogether. This is usually when you’re testing a larger element, such as a new color scheme or layout. To do this, you have to create a new version of the web page and upload it to your tool.

Once you’ve created all of your pages and/or variations, you need to designate a conversion goal. For example, if you want people to sign up for a newsletter then your goal is to get visitors to the “Thank you for confirming your subscription” page. You mark your conversion goal by including a small piece of code on that page. Whenever a visitor reaches the conversion goal, your tool records what version converted that particular visitor. You can monitor results as they come in, or just wait until a significant number of people have converted to review your results.

How to Know When You’re Done

How do you know when you have enough results to draw a conclusion? As mentioned above, a few thousand visitors is usually enough to tell whether A or B works best. If you use the Google’s Content Experiments tool, the program will automatically tell you when you’ve reached statistical significance.

If you’re using a different tool, you may need to determine your own results. To calculate your results you can use one of many statistics formulas. The Chi Square statistic is a popular one. If you want a comprehensive post on the numbers behind split testing, check out Paras Chopra’s post over at Visual Website Optimizer.

Or, if all of this math makes your head hurt, you can use a calculator like the one at User Effect to get your results (though it helps to have a basic understanding of the formula so that you know how and why the calculator came to that conclusion).

Tips for Successful Split Testing

  • Wait until you have a significant number of visitors who have converted to make your conclusions. It might be tempting to draw conclusions based on a few hundred visitors, but you’ll get more accurate results with a larger sample size.
  • Always test simultaneously and not one page after the other. You want to test your pages on the same stream of traffic.
  • Don’t stress over minute details. Using 12-point Helvetica instead of 12-point Times New Roman probably won’t make much of a difference traffic-wise. Focus on bigger changes that are likely to lead to better results.

You can apply all of the advice on web traffic conversion you can find, but the only way to really see what works is to test things out for yourself. Split testing really is quite simple to initiate and conduct. It’s a proactive way to optimize your website for conversions without leaving anything to chance.