A/B Testing with Google Content Experiments

eric leuenbergerAny online retailer who is serious about increasing sales knows that ongoing testing of design elements is essential. There are many ways to achieve this, including custom coded, paid and free solutions. In the past, one of the better free options was Google’s Website Optimizer (GWO). However, in August of 2012, Google discontinued GWO and replaced it with Content Experiments. Google says the move brings A/B testing to anyone who wants to use it, with no technical skills needed. They also point to it being completely integrated with Google Analytics, whereas GWO had its own reporting area outside of Analytics.

It is true that the move has made it easier to set up an A/B test, as less tracking code is needed and setup is faster. Still, many advocates of GWO are not convinced that Content Experiments has any real value outside of static landing pages. While GWO worked well with dynamic and static pages alike, Content Experiments requires a different URL landing page for each test variation. This can be difficult to achieve with dynamically generated content, making the, “no technical skills needed,” claim seem like it pertains only to non-dynamic content. This alone makes Content Experiments seem totally impractical for ecommerce driven sites, which are typically comprised of data driven dynamic content.

Despite falling short in certain areas, Content Experiments offers some value. Here is a look at the pros and cons of Content Experiments.


1) Instead of the need to place multiple tracking code snippets in various sections across the page/site, as was the case with GWO, Content Experiments reuses the Google Analytics tracking. Just add a small tag to the original tracking script and it’s ready to go.

2) Content Experiments comes with a setup wizard that walks you through the process and helps you quickly launch new tests.

3) Setting up an experiment is simple and only takes a few minutes, when not developing the actual variant page. (See more in cons.)

4) Google Analytics goals are used as the conversion indicator for each test.

5) URL referral information is not lost on redirection, including dynamic parameters.

6) With Google Analytics Premium, the paid version, users will see fresh experiment data quickly.


1) There is no option to multivariate test.

2) Each experiment needs to have a unique URL, meaning completely separate landing pages for each test. For example, a product page runs from a single template, which dynamically inserts the product ID into the URL to display the content, yet the page itself technically remains the same in nature. In order to develop a variant for the test, one would need to duplicate the original URL, defeating the purpose of dynamic content, and then add to that URL some type of test indicator so Google knows what to display.

Here is an example:

www.theecommerceexpert.com/pageone.php?pid=1 (original / control)

www.theecommerceexpert.com/pageone.php?pid=1&varient=b (variant for testing)

To get this to work on a dynamic site, one would likely need to alter the programming code on the page. Additionally, it could pose a duplicate content issue with Google if it picked up the new URL and indexed it. Google’s answer to this is to make sure the variant is tagged in such a way that robots don’t index it. Technical? Yes.

3) Google Analytics’ goals for Content Experiments means that users will quickly hit the maximum allowed.

4) Users are limited to five test variations at the moment. This means that the variants chosen for testing must be carefully thought through. Large volume sites are afforded the use of many variations.

5) Users of the free Google Analytics have to wait longer to receive results on experiment data, which could result in wasted expenses and lost sales.

What online retailers can do.

With the multitude of split testing tools available, it’s likely that at least one would meet an online retailer’s needs. In truth, one could always make changes to pages without any A/B testing software in place and simply use Google Annotations to indicate the date the test started. This method requires all traffic to see the same test rather than a split, in which one half would see version A and the other half would see version B. The results, however, seen through trending reports over time in Google Analytics, can be very telling and accurate.

When opting for this method of testing, make sure only one item at a time is tested to determine the effectiveness of the change. Testing several elements on a page at the same time often leads to an inability to determine which one(s) had a positive or negative impact.

Online retailers must test everything and must not become complacent with results. “Test thoroughly and test often,” should be your mantra, but keep it simple. The method that is chosen doesn’t really matter, and it doesn’t have to be fancy, as long as it gets the job done. Learn more about Google Content Experiments by logging into Google Analytics and accessing it from the left menu: Content > Experiments.

Happy Testing!

Eric Leuenberger is an ecommerce expert, founder of Ecommerce Amplifier, and owner of Voom Ventures, LLC whose products and services help storeowners and operators increase traffic, maximize ROI, decrease expenses and increase revenue. He can be contacted online at www.TheEcommerceExpert.com or by phone at 1-866-602-2673.