Google Analytics Experiments (Split Test) Setup


How do you setup a split test (or experiment) in Google Analytics? Find out in this short video from Elementive.

After you've setup the test, be sure to check out our video to find out how to review experiment results here.


In this video, we're going to discuss how to set up experiments or split tests in Google Analytics. Split tests let you compare two different versions of one page to see which page gets more people using or converting on your site.

To begin, open up Google Analytics. In the sidebar, click on Behavior, then Experiments. To create an experiment, click on Create Experiment. On the Create Experiment form, you first want to enter a descriptive name for your split tests. For this video, I'm going to use Test Experiment.

Next, you want to select your objective. This is what you're trying to improve with this test. You have two choices, goals or site usage. Under Goals, you will see all available goals. You'd use Site Usage to test ways to reduce bounce rate, increase page views or increase session duration. For this experiment, I'll select a goal of people ordering a competitive analysis from Elementive. You can then decide how much traffic you want to be involved in this task.

This requires a bit of explanation. If you select 100% and you have two versions you're testing, your original Page A and the test Page B, then 50% of your visitors will see one version and the other half will see the other version. But you can also select a lower percentage of traffic. This will reduce the number of people who end up being involved in your test. For instance, selecting 10% of traffic would have 10% of visitors in the test group. In that case, the test page would be shown to 5% of visitors, and then 5% of the test group would see the original version. The other 90% of visitors wouldn't be part of the experiment, and they would just see the original page. If you select a lower percentage, then this means fewer people are involved in your test, so it's gonna take longer to get a big enough group to have conclusive results. Typically, you just want to set this to 100%. Advanced Options let you change how traffic is distributed and also would let you adjust the confidence threshold. Typically, you can just leave those alone. 

The next step is specifying the pages you want to test. You can include up to 10 variations, but that gets pretty hard to control, plus it really reduces how many people see each version of each page, which means you're just gonna take longer to get results. For the next step, you need to input code. If you're comfortable copying pasting JavaScript into your website then you can click to manually insert the code and follow the instructions. Or, you can click Send the Code to the Webmaster to send all those instructions off to whoever manages the code for your website. After creating the experiment, you can go back to the experiment list and you'll see your test experiment listed. Once the code is installed, click the experiment name, scroll to the end of the page and make sure it validates. If it validates, click Start Experiment and your split test will be running.

If you have other questions about Google Analytics or other aspects of measuring your website's performance, please contact us below.