5 Tips for Testing with Google Content Experiments

This entry was posted in Reporting and tagged , on by .

Have you had a chance to use Google Content Experiments to optimize a website? Formerly known as Google Website Optimizer, Content Experiments was introduced in 2012 as a feature within Google Analytics. Content Experiments allows marketers to conduct A/B testing of an original page (A) against a variation (B).

During testing, Google will send a pre-determined percentage of new visitors to the test page and compare metrics for clicks, conversions, time on page, and pages per visit.  If you are new to A/B testing or Content Experiments, then follow these best practices for running successful tests below:

1. Establish a Measurable and Realistic Goal
Content Experiments has four types of goals or visitor behaviors that you can track including: URL destination, visit duration, pages per visit, and events (such as adding something to a cart). Don’t just test alternate web design elements for general usability. Instead, set a measurable and realistic marketing goal and use website testing to reach that goal. Why are you testing? What kinds of results do you hope to achieve? Maybe your goal is to send 10 percent more visitors from the home page to the highest-revenue generating page on your website. Or you could test two different call-to-actions in an effort to increase sales conversions by 3 percent or more. 

2. Keep it Clean
True A/B  testing means experimenting with one variable at a time against your original page. For example, here are three separate variations you can test for in a call-to-action:

  • Changing the language of a call-to-action. For example, “Get Started Today” vs. “Signup Today.”
  • Changing the position of the call-to-action. Testing an upper right hand corner placement vs. a middle of the page placement.
  • Changing the color of the call-to-action. Testing a red button vs. a yellow button.

Each of these individual tests can provide a lift for your marketing results. For accurate test results, keep your experiment clean and limited to one variation at a time. If you test a new color and a new placement for your call-to-action in one experiment, then you will not be able to isolate which change impacted visitor behavior.

3. Set The Right Parameters
If your website is highly trafficked, then running a test for one or two days may give you time to collect a relevant sample size and draw conclusions. However, if your site receives only a few hundred visits per day, then you may want to run your test over the course of a week or longer. Likewise, those sites with fewer visitors should send as much as 75 percent or more of new visits to the test page to help speed up testing. Use this A/B split test calculator from VisualWebsiteOptimizer.com to determine the right timetable for your test.

4. Use Content Experiments to Optimize AdSense Revenue
In September, Google announced Content Experiments integration for AdSense users. If you are a publisher running ads on your web properties, now you can leverage Content Experiments to optimize those ads for the greatest revenue. First link your AdSense and Analytics account, then test ad size or placement variations to determine which ad types  generate the most clicks and ROI for your site. Publishers may appreciate Google’s multi-armed bandit algorithm, which looks at live data and sends more traffic to the winning variation for maximum revenue. Or, publishers can override this option and send a predetermined amount of traffic to each variation. 

5. Take Advantage of the API
In June, Google opened the Content Experiments API to developers, enabling advanced users to pick and choose which testing functionalities they want to include. Using the API allows you to test without redirects, which provides a quicker and more seamless page load experience for visitors. You can also conduct server-side testing or offline testing (great for interactive kiosks). In addition, developers can use A/B or proprietary testing logic in lieu of Google’s multi-armed bandit approach.

What are your tips for testing website variations with Content Experiments? Please share your thoughts in the comments below.