Conversion optimization with A/B testing
How do you know which version is the best? An A/B test gives you quick and effective answers to small adjustments you want to make to your website.
What is A/B testing?
A/B testing is briefly explained a method for testing small changes to a web page against an existing design, to then see which version works best. In other words, what the name implies is a test with two versions of a website—an A version and a B version. You test these by sending traffic to the two different versions, then seeing which version works best. The winner, the one with the best performance, is the one that one uses next.
What can be tested with an A/B test?
What you test with an A/B test depends on your goal with the website. One traditionally tests different content, design and functionality, and measures what works best. For example, one can test different headings, placement on the buy button in an online store, the number of fields in a sign-up form, or different images on a web page.
Conversion Optimization
To understand A/B testing, it is important that one is familiar with the concepts of conversions and conversion rate, which will govern the A/B test. The starting point for conversions is goals. Measures on a website can be, for example, sales, inquiries, registrations, downloads, and the like. A conversion is when a visitor accomplishes the goal of the website (for example, buys a product in an online store). We use conversion rate as a measure of how many percent of visitors convert. For example, if 10 out of 100 visitors buy a product in an online store, you have a conversion rate of 10%. In other words, A/B testing is about increasing the conversion rate of a website—in other words; conversion optimization.
So, how does it work?
With the goal defined, the purpose of the A/B test is to check out assumptions about changes that you think will have a better effect on a website's conversion rate. Further, an example follows with an imaginary test for an online store.
Example: Buy Button Test
In an online store, we want to test changes on the buy button. It is through the buy button that we encourage our visitors to take action, and the buy button is crucial in creating conversions. In our online store we will check out which location of the buy button works best. You can also test with other changes to the buy button, for example by using different colors, sizes and text on the button, but for simplicity, we now only test the location of the button.
We're setting up a challenger version of the product page with a changed location on the buy button -- a change we believe will bring more purchases.
Once the challenger version is in place and ready for testing (version B), we choose to direct part of the traffic to this version. The measurement parameters we use are clicks on the buy button, and it is the version with the percentage of clicks on the buy button/highest conversion rate that wins.
In our test, we choose to send 50% of the traffic to the existing version A, and the remaining 50% of the traffic to the new version — version B. In the event of major changes, which entail a greater risk of a decrease in the conversion rate, it is appropriate to assess exposure risks. One can then choose to send a smaller share of traffic to the new version, so that there are less consequences if one does not succeed with the challenger version, version B.
What kind of traffic should be used in the test?
There are further opportunities to define other characteristics of the target audience. Be it, for example, running the test on users who receive a link in a newsletter, selecting out users who use a given browser or platform, or users who have a given cookie installed in their browser (for example, a cookie that they are logged into the online store).
In our example, we choose to run the experiment against all visitors, coming from desktop, and we do not want to include smartphones and tablets in the test, as the website looks different on these platforms.
We see that 2% more people click the buy button on the new version. We have enough traffic to be able to say that the data is valid and reliable, so we are phasing out version A, directing 100% of the traffic to version B which is the most optimal version.
Here's how to get started
To conduct an A/B test, the easiest way is to use an existing tool. There are several solutions on the market, such as Optimizely, Visual Website Optimizer, Google Optimize and Mixpanel. At Increo, we have extensive experience with Optimizely It is the market leader in A/B testing. The solution is used by 6,000 companies and is highly streamlined and easy to use.
Do you want help optimizing your website with the help of A/B testing? Get in touch with us at increo@increo.no for a nice chat about the possibilities of your company.
