Give yourself a pat on the back if you have developed an ecommerce business with a steady flow of orders. Perhaps you have had success with search engine optimization, pay-per-click advertising, and social media. When you are ready to accelerate your business, however, A/B and multivariate tests can help. By testing the behavior of your visitors under different circumstances, you gain a deeper understanding of adjustments you can make to your site that will improve conversions.
Google Website Optimizer is a free service to conduct A/B tests. Once you get through the initial configuration stage — which can be challenging — you can discover how your shoppers respond to the changes you make on your site.
A/B vs. Multivariate Testing
A/B testing — also called “split testing” — refers to showing website visitors two versions of a page and seeing which one performs better. You might decide to test many variations of the same page. This is still considered A/B testing.
Multivariate testing is more complex. It involves testing multiple variables simultaneously by identifying those variables and evaluating different combinations to that produce the best results. We recently addressed the general need for multivariate testing, at “How to Test Multiple Variables on an Ecommerce Site.”
In my next article, I’ll explain specifically how to set up multivariate tests using Google Website Optimizer. Today, however, I’ll focus on A/B testing.
Setting-Up Website Optimizer
Access Google Website Optimizer on its own page, or via the Google AdWords interface, under the “Tools and Analysis” menu.
To utilize Website Optimizer for A/B testing, insert three pieces of code on your site for each test. Talk with your ecommerce platform provider about modifying your web pages, making sure you can edit the templates on your site with the appropriate code. Some ecommerce platforms can accommodate Google Website Optimizer; others may require customization.
<head> </head> section of your pages.
Identify the control page. Add the control and tracking scripts to your control page. This will indicate to Google that the page with these code snippets is the original version that you are testing against.
Identify the variation pages. Add the tracking script to your variation pages. Google will see the pages with the tracking snippet only as the variation versions of your test. Keep in mind that you can try many variations. For simplicity on this exercise, I’ll have just one variation.
Identify the “success” page. The conversion code gets pasted into the success page on your site. If you are already doing AdWords or shopping engine tracking, you are likely familiar with this. For an ecommerce site, you need to identify the page that the shoppers see after placing their orders — make sure that the conversion code is included.
Once you have mastered the code insertion, you have one more hurdle: You need traffic to evaluate the results of your test — around 100 conversions per variation will provide meaningful results of your test. Your Google Analytics landing page report will help you determine when you have your 100 conversion results. This could take a week or longer, depending on the site.
What to Test
The items you can test are limited only by your imagination. Here are some basic examples.
- Size or placement of call to action button.
- Message in your page heading.
- Color of text.
- Placement of photos on your product page.
You can test many different sections of your page. You can literally create an A/B test for any element on your website, including pricing, descriptive text, and trust symbols. The power of A/B testing and the Google Website Optimizer environment is that you can get a clear understanding of which factors create the best results and take a scientific approach to fine tuning your pages.
Creating Your Test Pages
Here is an example of creating an actual A/B test.
Determine a variable to test, such as a call-to-action header test to determine whether adding a stronger headline to landing page will help our site’s conversions.
Identify the control page. In this case, I’ll use a popular product page on our site. Copy the URL for the page into a temporary text document on your desktop.
Create a copy of the page and add the test element. As with our control page, copy the variation page’s URL into your temporary text document.
Identify the URL of the success — or conversion — page. If you are not sure, place a test order and determine the URL for the final page of the process. Add the success URL to your list of addresses.
Implementing Your Test
Now it is time to set up your test at Google Website Optimizer.
Click “Create a new experiment.”
Click “A/B Experiment, the simplest way to start testing fast.”
Since you have already done your prep work of identifying your control, variation, and conversion pages, you can click “I’ve completed the steps above and I’m ready to start setting up my experiment.” Then click “Create.”
Fill out all the fields on “A/B Experiment Set-Up: Name your experiment and identify pages.” This should just be a matter of entering your experiment name and copying and pasting your three URLs. Then, click “Continue.”
Installing and Validating Your Tracking Code
Copy each snippet of code into the correct page on your site.
Once our code is in place, click “Validate Pages.” Troubleshooting validation issues can be a challenge. If you have a template-based ecommerce platform, you might need to confer with your platform support team. Otherwise, assuming your validation is successful, you are ready to test.
Click “Start Experiment.” Now you are off and running, which means that it is time to … wait.
You will need some time to amass a good set of test data. For the first few days, you should check back at Google Website Optimizer to see how many page views and conversions you are recording. As Google Website Optimizer records data it will help you determine the winner of your test.
The control and each variation will show the following results.
Estimated Conversion Rate. The conversion rate for your test and variations.
Chance to Beat Original. The likelihood that the test will beat the original. The closer that number is to 0 percent or 100 percent, the more definitive your test will be.
Observed Improvement. Compares the conversion rate of the original to the conversion rate of the variation.
Conversion/Visitors. As with many analytical tools, this metric compares the number of conversion to total visitors to the page.
A clear winner will display a green bar and indicate that you have identified the best variation for your website. With a conclusive winner, you can make adjustments to your site and get ready to repeat the process with another test.
A/B testing is a state of mind. With a solid test environment and some well-planned tests, you can bring about some significant improvements to your ecommerce business.