The Most Underused Google Analytics Features - Experiments and A/B Testing
Google Analytics is a treasure trove of hidden features. Sadly, so many of these wonderful features live their lives completely overlooked, hidden in the depths of Google Analytics navigation. One of these features is Google Experiments found in the Behavior section. Experiments allow you to run A/B test with ease.
Earlier this year I was tasked with increasing conversions through a specific landing page for a client. One of my main tools for doing this was Google Experiments. By simply changing the Call To Action (CTA) I was able to increase conversions by 157%! And better yet, I have hard data to show the winning variation instead of subjective opinion.
How an A/B Test in Google Experiments Works, The Short Version
Let's say you want to test copy or CTA placement, or if your page does better with a Banner image or without one, whatever your little heart desires. Create copies of the page while leaving the original in place. You then place a little code that Google Analytics generates for you into the header of the original page and the copies (your variations). Google will then randomly send people who land on the original page to one of the variations or the control (the original). Now you sit back, wait a couple weeks, and see which variation wins.
How To Set up Experiments - The Long Version
Pick A Goal
What are you testing for? If you're an eCommerce store your goal is likely an eCommerce sale. A content site might have a goal of increasing email signups. B2B sites might define their goal as lead form submissions. This goal is going to be tracked to determine which variation comes out on top for the experiment.
Set Up Variations
Leave your original page in place and create as many variations as you want to test. I typically use /v1, /v2, /v3 as my variation pages to keep it simple. Make sure your variations are not in your sitemap and also have a noindex tag so they are not crawled by Google.
Set Up Google Analytics
One thing that I love about Google Analytics is that they have pretty solid documentation. So why recreate the wheel for setting up the Google Analytics side of things when you can hear it directly from the source? Check out Google's instructions here.
Start The Experiment
Experiments typically last for a minimum of 2 weeks and can go much longer depending on sample size and if one variation can be declared a winner that is statistically significant. If it is a narrow increase, your experiment will likely run for much longer than if you see a dramatic increase in conversions.
A Winner Is Named - Now What
You have a winner! Your third variation crushed it and now you can take these lessons to further optimize other pages on your site. So you're all done, right? Not exactly. Remember, we created multiple copies of the same page so now we need to undo our work. The winning variation's changes should be implemented to the permanent page. This means updating copy, creative, or design. Now delete your variation pages and make sure you place 301 redirects from your variation to your original in case a user bookmarked one of your variations or linked to it externally.
Google Experiments can help you optimize your site in a few simple steps. This sections of Google Analytics is often underused but pays huge dividends. Take the guesswork out of your design and test!
What successes have you seen with Google Experiments? What questions do you have in setting up your own? Leave a comment.