A/B Testing with Google Optimize

Home / Optimization / A/B Testing with Google Optimize
google optimize
This image shows testing different content for conversion rate optimization

Google Optimize was released in March of 2017, making it easier and more affordable for small businesses to do A/B testing of their website content.

The Scientific Method

Applying the scientific method to web design was born out of necessity for big web businesses with a lot of traffic. When you receive millions of visits to your website each day, just making a tiny change to the website that causes a one percent change in conversions can result in the gain or loss of millions of dollars. When millions of dollars are at stake, you test every change on a portion of the population before it is released to everyone.

The scientific method is a procedure that has been used in the natural sciences since the 17th century for gaining knowledge through the testing of hypothesis. The method consists of an iterative and cyclic process where one makes observations, thinks of interesting questions, formulates a hypothesis, develops testable predictions, tests the predictions in a controlled experiment commonly referred to as an A/B test,  and then rejects or accepts the hypothesis.

An example for improving a conversion funnel on a website might look like this:

  1. “People are not clicking on my button.”
  2. “Does color have anything to do with it?”
  3. “I hypothesize that more people would click on my button if it was green.”
  4. Using statistical methods, I’ll set up a controlled experiment so that the only independent variable is the color of the button.

The hard part

The concept is simple, but making it happen is hard. To do a controlled experiment, you need to randomly put each visitor to the website into one of two buckets. The control bucket or the experiment bucket, and the visitor must remain in that bucket throughout the experiment. Then you need to show the visitor one of two different websites, depending on which bucket they’re in. If the visitor comes back later during the same experiment, you need to show them the same version of the website that you showed them initially.

This can be done with cookies, those things that websites use to store information on your computer, and web programming with conditional statements, but it’s a significant amount of work.

Additionally, you need some statistical analysis skills. The sample size and experiment duration and other confounding factors must be considered in order to determine if the experiment results in any statistically significant findings.

Vendors make it easier

Two vendors rescued webmasters with their solutions. Optimizely, the value-priced market leader, made it easy to do A/B testing by simply adding a snippet of javascript code to your website. VWO (Visual Website Optimizer), the cheaper alternative with set prices, provided a similar solution. Both vendors provide WYSIWYG controls so that even non-programmers can easily code HTML, and both distill the complicated statistics into easy to understand results.

The only problem with these vendors was the price. The cheapest package started at $300/month, which was hard to justify for a small business that didn’t get enough traffic to complete a single experiment in under a month. Additionally, the return on investment was way lower due to the small amount of money involved in improving conversion rates on a low traffic site. Finally, most small businesses don’t have the human resources to invest in doing continuous conversion rate optimization.

Google to the rescue

Google, one of the pioneers of the freemium model, has entered the market. Extending their freemium web analytics package, they have released a free tier of Google Optimize, it’s conversion rate optimization solution.

If you’re using Google Analytics, the two packages work together offering capabilities beyond simple A/B testing. If Google Optimize identifies that one of your tests converts better, but only for a specific segment of the population, there’s no reason to just pick one version. Instead, you can keep showing a different version to each visitor, depending on what segment they’re in.

For example, if the blue button converts male visitors better, and the pink button converts female visitors better, but the blue button converts visitors better overall, there’s no reason to just use the blue button after the experiment is over. Instead, you can show a blue button to male visitors, and a pink button to female visitors.

We expect the free tier of Google Optimize to allow small to medium-sized businesses to do A/B testing for conversion rate optimization.

About Fierce Geek Inc.

Fierce Geek Inc. is a Bay Area web design company focusing on search engine and conversion rate optimized websites for small to medium-sized businesses. When designing or redesigning a website, we apply research backed hypotheses on the initial design. We then offer a very affordable monthly plan where we will continue to optimize the website using the scientific method. If your web traffic isn’t converting, give us a call.

Leave a Reply