Your Step-by-Step Guide to A/B Testing With Google Analytics

What Is A/B Testing?

The idea of A/B testing (also called split testing) is simple enough. Create a new version of your current web page. Then put them both online to see which one is the best. The one that brings the most business is the best design, and you update your website with the new page to enhance your business.

The idea is simple. A or B. Better or worse. But the logistics behind the actual process of split testing are just a bit more complex.

Consider the steps necessary for an effective A/B test.

• You create a new variation of a current webpage.
• You put both of the pages online at the same time.
• You split your traffic between the pages.
• You check for traffic stats and conversion rates.
• You update your site to remove the page that performs more poorly and you’re done.

Creating the initial pages is easy enough, although there are factors to consider including how much you change between the two pages. Too many variables and it’s hard to tell what is increasing or decreasing conversions.

But if you’re new at this, you may not know how to put both pages online and start directing traffic to them. And once you have the pages online and traffic flowing, how do you monitor your stats and conversions?

The Goal of A/B Testing

It’s very easy to get bogged down in the nuts and bolts of making split testing work. Then you forget what your actual goal of the test may be. A/B testing is a process of learning – emphasis on process. A good split test takes months, and you’re going to learn a lot more than how many sales each page makes if you’re able to track data well.

Through split testing, you learn whether you should be focusing on a single outcome or conversion or if there are several conversions that may take place on your page. By making small changes between your testing pages you can learn very specific things about your audience. You’ll see which design elements and content make a big impact and lead to conversions. In some cases, you can pinpoint exact percentages of increase from changing things as simple as image placement and font spacing. You can also test different ad types or even dynamic content.

Before getting into the logistics of split testing, determine what the specific goal for your A/B test will be. Are you looking for subscriptions? Sales? Sign-ups? Conversions are defined differently for different companies and websites. You should know exactly what you’re looking for before beginning the process to make the full test productive.

Ultimately, however, all of your split testing boils down to boosting your conversion rate optimization, or CRO. This boosts revenues. It doesn’t help your business to boost initial traffic and clicks but lose sales. It’s the bottom line that matters. Always.

But you can’t get starting on all of that useful A/B testing without knowing how to do it.

Fortunately there are tools available that are not just straightforward, but are free as well.

Google Analytics

One of the many webmaster tools offered by Google, Google Analytics allows for A/B testing between two pages or multiple pages simultaneously. Google calls its A/B/n testing a Content Experiment.

To get started with a “Content Experiment”, aka A/B test, look in Google Analytics for Behavior and then select Experiments. 

To get started you’ll click on the button for “Create Experiment.”

From there you’ll be taken to the experiment screen. Your first step will be to name your experiment. Your name should be very descriptive to make it easy to remember what – exactly – you are testing in your A/B experiment.

Define Your Experiment Objective

After a well described name, you’ll select an “Objective for the experiment.” This is where you’ll set up the metrics you want to use as part of your test. You can choose to evaluate based on Adsense, Ecommerce, Site Usage, Bounces or more.

The metrics should match your goal, obviously. If you’re tracking sales, you’ll want to see “eCommerce.” If you’re looking at impressions and ad clicks, select “Adsense.”  Use “Goals” to keep an eye on your own measures like event sign-ups or session duration. Track average page views and the time spent on the site through “Site Usage”. Select the metrics you’re interested in, and you aren’t limited to a single metric – you can track multiple things simultaneously.

Split Your Traffic Accordingly

On the same screen select the amount of traffic you’d like to direct to the test page. Remember an A/B test is a test of a variable against a constant. Your standard website is the A – the one you already use. The B is the new version of the webpage that you’re testing.

A 50/50 split of traffic will take half of your current traffic to the new page. This may be ideal for small changes that won’t make dramatic changes in visitor behavior. If you’ve made dramatic changes to your website, diverting half of your current traffic to the new page may be risky. It might be better to send a smaller amount to the new site for the purposes of your test.

A classic A/B test uses just two pages for testing. Google Analytics allows for multiple pages of testing, however. If you choose to test more than just two pages at a time, be sure you have enough traffic to split between the pages to ensure good results.

Email and Advanced Options

At the bottom of the experiment screen there is a toggle to turn on email notifications of changes. It’s always a good choice to know what your traffic and website are doing, right? Turn notifications on.

Under the toggle for emails, there are Advanced Options.

You have the choice under Advanced Options to choose how to distribute the traffic to your sites. Select this option to assign an equal amount of traffic to each page variation for the life of the experiment. Otherwise, leaving that button disabled will allow Google to adjust your traffic dynamically based on performance.

Plan on testing for at least three weeks to a month for the best results. You can leave even the timeframe up to Google Analytics if you fix your confidence threshold to a minimum level that must be achieved before a “winner” of the A/B test can be determined. If you allow Google to crown a winner, the test may take more time to reach the desired threshold.

Codes and Experiment

Once the experiment page is set up, the next step is to configure the experiment and add in your current website and the variations you’re trying to test. This simply requires entering the current URL of your webpage and then the URLS of the pages you’d like to test. The preview image will let you be sure you’ve entered the right URL before you “Save Changes.”

Once you save changes, the experiment code for your A/B test should become visible. If it does not show up immediately, check to be sure that your Google Analytics tracking codes are installed on your current page and test page.  

Copy the experiment code and follow the instructions to paste it directly after the opening head tag at the top of your current web page. Save your changes and move on to the final step in setting up the experiment.

Start the Experiment!

Once the code is in place, Google Analytics will check for mistakes and let you know what errors you might need to correct. Once any errors are resolved, your content experiment will be ready to launch and results will start being tabulated after a day or two.

You can check results while the experiment is running, of course, and you should. But the best time to look for results is when the full experiment is played out. Google Analytics will use the metrics you set to determine a winner that meets your confidence threshold. The winning page is the one you use on your site permanently.

And once you figure out the winner of the first A/B test, you simply go on to your second. Continue experimenting on various pages of your website to increase conversions and continuously improve.

It’s important to note that Google Analytics cannot handle multivariate testing, which is a more complex technique where multiple variables are changed and tested in different combinations. Google Analytics isn’t designed for that. It is, however, very good at what it is designed for: straightforward, easy to use A/B Testing.