Google Optimize is a free platform to design and run website experiments including A/B tests and multivariate tests (MVT). It’s built on top of Google Analytics and a native member of Google’s marketing ecosystem.
One of the classes I took at Bentley as a marketing analytics graduate student was the experimental design class instructed by Paul Berger. It was one of my favorite classes and definitely sparked my interest in this topic.
Since then, I have been genuinely reading every online article (or blog post) I came across about digital marketing experiments. (Unfortunately, many of them have serious issues either due to the design of the experiment or the interpretation of the results, but I will write about this in detail later.)
So in this blog post, I want to talk about Google Optimize, which is a tool to run website experiments including A/B tests and multivariate tests (MVT). And I wanted to write about it because I believe it’ll eventually become our go-to tool to design web experiments — mainly because of the fact that it’s free and it’s part of Google’s marketing ecosystem.
Google Optimize is a platform that enables one to run website experiments. Optimize is currently free, but not its enterprise version (Optimize 360).
You can use this link to sign up for free.
Google Optimize was built on top of Google Analytics. So if you are already using Google Analytics, the only thing you need to do to deploy Google Optimize is adding a single line of code. You can find more information here.
Google Optimize uses the Bayesian method rather than the frequentist method, which is the method they often teach at Stats classes. It’s important to be aware of this fact because if you’re familiar with the frequentist method and looking for the p-values at the end of your experiment, you won’t be able to find them in Google Optimize results. Instead, Google Optimize will give you the probability of each variant being better than your baseline along with the probability of a variant being the best. Learn more.
Any thoughts? or questions?