Site   Web

October 5, 2015

A/B Testing with Google: A How-To Guide

You are the operator of a website. You are proud of your creation but it isn’t performing to the standards you had anticipated. Or maybe you are just looking to try out something new without tampering with the results you are currently receiving. Across all industries and site styles, A/B testing is an immensely powerful tool to test what scenarios resonate best with your customers.

There are some pretty large-scale benefits that can be reaped from A/B testing; everything from conversion rate increases to more alluring design elements for your audience. The problem most beginners face when entering into A/B testing is difficulty in determining which tools are best to use, how to get the testing set up and configured, and how to know when everything has run its course.

Recently, Kapital increased conversions by 44% through A/B testing. And they did it all with Google Analytics.

Google Analytics is a powerful (and free) tool that every business owner should be using to study and optimize websites. Below I outline how to set up an A/B test so that you too can enhance your site’s performance.

Step 1: Select Your Objective

So that you’re aware, Google has combined the terms A/B testing and split testing into the single term they call “content experiment”. This will be found under “Behavior” and “Experiments”. Now simply click “Create Experiment” to get started.

The first thing you will need to do is name the experiment in accordance with the metrics you are gauging. If you are testing out a new sign-up module or a new CTA for selling products, set the name as a descriptive identifier for that particular element.

Next, define the metrics that will be used to evaluate test results. This will be found under “Objective for this experiment”. Various metrics can be selected, including:

  • eCommerce – This is used to measure the number of transactions and revenue.
  • Adsense – Measures ad clicks or impressions
  • Goals – Measures predefined goals such as session duration, page clicks, or event attendances
  • Site Usage – Helps measure average page views or time on site

What you select is all dependent on the elements that you wish to try and improve. Also, do note that multiple metrics can be selected and applied.

Step 2: Divide Your Traffic

Now that objectives have been defined, it is time to divide percentages of your web traffic in the experiment. This controls how many individuals will see each version of the site you are testing. This is established through the “Advanced Options” tab under “Distribute Traffic”. Here you can assign the values you determine appropriate for each site variation.

If this portion is left untouched, the experiment will revert to its default behavior and adjust traffic dynamically in accordance with performance.

Set the experiment to run for a minimum of three weeks to obtain more definitive and accurate results. It is also advisable to set the confidence threshold. The higher you set the threshold, the more certain you can be of the winner’s capabilities against the other design.

Step 3: Configuration

Here, the test will be configured by adding the URLs of the original page and the page variations. Once the URLs have been entered, inspect the preview image to ensure that all of the information you have entered is correct. Once accuracy has been assured, press “Save Changes” and proceed to the next section.

If the Google Analytics tracking codes are installed correctly on all versions of the web pages, an experiment code will generate in the box below. Open the head tag, and place this code at the top of the original web page and hit “Save Changes”.

Step 4: Review and Launch

Now that all of the information is in place, review the data and ensure accuracy. After entering in the tracking codes, Analytics will confirm this information and display any errors that might have occurred.

From time to time Google is not able to locate the code. In this instance, check your work for any errors that might have occurred. If you are still unable to find any issues, the validation process can be skipped granted that you are certain the code was entered properly.

Do note, however, that Google advises the validation processes be skipped only as a last option. If there are no errors and the code has been applied correctly, Google will provide you the go-ahead to start the experiment and you will begin to receive data within the next couple of days.

Step 5: Evaluate the Results

Once the entire experiment has been completed, Google will determine the most effective version of the web page in accordance with the previously assigned metrics. You can now publish the dominant version of your site page, dive even deeper into the elemental successes and failures, and test new versions and aspects all over again.

For a site to evolve and mature, it must be tested, and tested often. A/B testing is a fantastic tool to help you determine which aspects of your site work, which don’t, and how you can improve upon these areas. The Google Analytics A/B testing tool can help transform your page into a powerhouse performer; you just have to be willing to consistently test, review, and test some more.

Have you used Google’s A/B testing tool? What improvements did you gain out of the experiments?


Conscious online marketer, web executive, and multi-faceted writer Tina Courtney has been creating and fostering online innovations since 1996. Tina has assisted many clients in maximizing online production and marketing efforts, and is a staff writer for SiteProNews, one of the Web’s foremost webmaster and tech news blogs. She’s produced and marketed innovative content for major players like Disney and JDate, as well as boutique startups galore, with fortes including social media, SEO, influencer marketing, community management, lead generation, and project management. Tina is also a certified Reiki practitioner, herbalist, and accomplished life coach.  Learn more on LinkedIn, Facebook and Google+. Visit My Google+ Profile