Lesson 10: Split Testing

Split testing is the best method for determining which website features are most effective in generating clicks. Split testing, also known as A/B testing, is a method used to compare the success of two variations of a website simultaneously (where each variation, for example, uses a different color scheme or headline).

The "control" version of a website (called "A") is contrasted with "B," the modified version. Website designers can learn much about how visitors interact with their sites and how well they perform by conducting split tests.

Your website must receive many visitors if you want to run a successful split test. As with any experiment, an adequate sample size is required for statistical analysis.

If you use a larger sample size, there will be less of a chance that your results are due to random chance. Although there is no "magic number," several thousand subjects usually yield reliable results in online surveys. The test duration will increase in proportion to the lower traffic volume to your website.

In large part, the website's intended function will determine which parts of the website you decide to test.

A newsletter sign-up form is an excellent example of where testing web copy, the number of fields required, and the deployment of a pop-up versus a static form can yield positive results. Other factors that can influence a website's success include:

·       Which side of the page should that image go on besides the text? Is it preferable to have two columns rather than one?

·       Does one type of image tend to do better than another? For example, when deciding between stock photos and images of your products, how should they be used?

·       Should you alter the button's wording that prompts the user to act? Is there a need for a larger version? To use a different hue?

·       What happens when you vary the price of a product? How do you know which specials to promote?

·       Do site visitors prefer long blocks of text or shorter paragraphs?

·   Titles and subtitles - what's the most efficient wording, exactly?

How Many Should You Put Through a Single Test?

Multi-variate testing is essentially a more efficient form of split testing. Since it is possible to split test nearly every aspect of a website, you can simultaneously make multiple variations to a web page, such as a new call-to-action button, set of images, and headline.

If your split test results aren't what you expected, you can't pinpoint which variables caused the failure. Accurate results also require a large volume of visitors. You can choose to test a single component or several at once.

Types of Split Tests

Multiple types of split tests can be conducted to ensure that the time and effort spent on the experiment will yield valuable results. For example, when conducting A/B tests, marketers often aim to accomplish the following:

- The number of people who click on a hyperlinked title to visit a website can be increased by conducting A/B testing on the titles of your blog posts and webpages. A/B testing can help increase the number of people visiting a website.

- Improved conversion rate by experimenting with call-to-action placement, color, and text. As a result, more people may "convert" into leads after submitting their information through your website's lead generation forms.

- Test out various introductions to your blog posts, fonts, and featured images to see which ones keep readers engaged and ultimately result in a lower bounce rate.

- About 70% of a website's visitors will abandon their shopping cart without purchasing anything.

The term for this is "shopping cart abandonment," and it's bad for business for any e-commerce site. However, this rate can be lowered by experimenting with the placement of shipping costs, product images, and other design elements on the checkout page.

Making Your Very First Split Test

Now that you have a sizable amount of site visitors and a list of potential variables, you can set up your first split test.

Obtaining a split-testing tool is the first step. The Content Experiments tool is available to anyone with a Google Analytics account. Standard Reporting can be accessed through your Google Analytics account. Pick Experiments from the Content menu. Split test checking can also be done for free using Firepole Marketing or Visual Website Optimizer.

You can split the test in two ways:

You can split the test into two different elements (say, two different sales copies) by telling the tool to alternate between the control and test pages on each visit.

Some of the users of your tool are directed to an entirely different website. This is done when a more substantial change, such as a new color scheme or layout, is tested. To accomplish this, update the web page and add it to your program.

You need to set a conversion goal after making your pages and variants. For example, if you want people to sign up for your newsletter, one goal could be to direct them to the "Thank you for confirming your subscription" page.

Then, add a line of code to the page where you want to see a certain percentage of conversions. Each time a user converts, your tool will track which version was most effective for them. You can review your results immediately or wait until a sizable number of people have converted.

When You've Reached the Finish Line

When do you have enough data to decide? In most cases, the difference between A and B can be seen after only a few thousand visitors, as was mentioned above. However, when you've reached statistical significance, Google's Content Experiments tool will let you know.

Results may need to be calculated manually if you are using a different instrument. However, many statistical formulas can be used to compute your results.

Optimizing A/B Testing

Don't draw any conclusions until you've seen a sizable number of conversions from your traffic. Though it'd be easy to jump to conclusions based on data from just a few hundred users, a larger sample size would yield more reliable results.

Never test a single page at a time; instead, test the entire document simultaneously. Your pages should be tested using the same volume of traffic.

Don't worry about minor things. There probably won't be much difference in viewership if you switch from Times New Roman to Helvetica in size 12. Instead, focus on the more substantial changes that will likely improve outcomes.

There is no substitute for trying different strategies to increase conversions from website visitors. It takes minimal effort to set up and run a split test. However, proactively optimizing your website for conversions ensures nothing is left to chance.