Google Optimize - A/B Testing Tool

Google Optimize: Google’s A/B testing tool

Today’s post is dedicated to a tool that every marketer and digital agency should take into consideration when optimizing the performance and usability of a website. What can you expect from Google’s tool? How is it configured correctly? How to run an experiment?

For all the above questions, we will be answering throughout the article.

Table of Contents

First of all: What is Google Optimize?

Google optimize is a tool that Google released back in 2016 in beta form and then launched worldwide.

This tool basically allows you to do experiments on your website without having to touch anything of it thanks to some tags that you implement on your page to connect it with Optimize.

Thanks to these tests, you can discover from design errors to options in it that make you multiply x2, x3 or xn the number of sales,or lead acquisition, on your website.

In addition, it can be joined with Analytics which makes the measurement of the different results much more effective.

That said, let’s start with the implementation of this tool.

Creating an Account and Container in Google Optimize

The first thing is to choose an account name which can be the name of the company or the name of the website.

Later Google will bring by default three boxes checked, you can leave it as it is or deactivate the ones that do not interest you. In our case, we deactivate the first two because we do not want to share our customers’ data with Google, even if anonymously.

Finally, add a container to the account. The name of this container can be changed at any time.

At this point the Google Optimize interface should already appear inviting you to create your first “experience”.

Linking Google Analytics

Once you’re done setting up the container, you need to join your Google optimize account with Analytics. This option is, in my opinion, better than launching into creating a first-time experiment as recommended by Google.

Once you display the option to “link with google analytics”, you have to click on the button that says “link property” and when the tab opens, choose the property you want to link with Optimize.

Do not worry if you choose the property wrong because later you can change it.

Install Google Optimization Snippet

What you have to do in this step is to install the Optimize code on your website.

To do this, you have to click on “install optimize” and you will be displayed a pop-up window with the code.

At this point you’ll need to choose how you want to install the Optimize code. It can be replaced by replacing the current analytics code with this new one or through Tag Manager.

If you choose Analytics, you’ll need to go to your website and replace the current analytics tracking code with the new one and save your changes.

With Google Tag Manager

Doing the code implementation with tag manager is easier if you already have a little experience with this tag manager. In any case, I will explain it step by step in case you want to try.

If you have not yet created a tag manager account, I leave you this article: What is tag manager

Once you have established the custom analytics variable, you will have to create a new Google Optimize tag that fortunately for everyone, is already predefined by Tag Manager.

Once we have chosen the tag, we need the ID of the Google Optimize container, it is the one that begins with GTM and was created at the beginning. This ID will have to be pasted where it says “Optimize container ID”.

Then, where it says “Google Analytics Settings” we will have to choose the variable we create for analytics.

Finally, in the “Activation” section, you will have to choose “All pages” if you want the container to be activated on all the pages and thus be able to test on all of them. If, on the other hand, you only want to do it on certain pages, you will have to create a specific activator.

Installing the Google Optimize special code

When the tool was launched, it had a small flaw that was that after installing the code, when loading the web it did like a small jump that was weird, I would say bad, since for a few thousandths of a second, a version of our experiment appeared that should not (as a summary), which worsened the user experience. To solve this, Google proposes us to use a small patch that improves this user experience.

The snippet must be placed before the new analytics tracking code or, if we have implemented Optimize through Tag Manager,before the Tag Manager tracking code we have on the web. That is, we have to paste the patch on it.

In case you have any doubts about this, I leave you a Google article where they explain everything much more detailed:

Creating the first Experience in Optimize

Once the above codes necessary for the correct functioning of the tool have been included, we can start creating the experiments.

**Before continuing, I want to make a point: if you have just installed the codes and you miss an error message in the tool, do not worry, it may be a matter of time before Google Optimize recognizes the code and allows you to create the experiments. Let it “rest” one day.

First steps to create the experience:

  1. Name it
  2. Choose the page of our website where we want to do the experiment
  3. Choose the type of experience

When choosing the experiment we have 4 options,the most used is the A /B test that serves to do tests within the same landing to which you can change the design through optimize.

The next option is the multivariate test that allows you to create different landing models with different combinations to see which is the one that gets the best user response. For example, you have 4 featured images and 2 titles (h1) that you like and you don’t know which one might be better to achieve your sales goal. With this multivariate test you can create the combinations of title + image that are necessary (8) to do the test.

The third experiment would be a redirection test. In this case you have two completely different landings and you want to try which of the two works better.

Finally we have the personalized experiment that differs from the rest because it can be a permanent changee. In this case, what happens is that you create an experiment by changing, for example, a part of the design of your website and segment the people to whom you want to show the experiment, if you find an audience that is responding well to this experiment, you can “select” it so that this change always occurs in visitors with that profile.

Practical example of A/B experiment in Google Optimize

In one of our websites we did a small experiment in the form of an A/B test to know if by changing the main title of the post and some image, we managed to reduce the bounce rate of the web (the landing was the chosen home was the home).

To do this, we let 40% of visitors see our website with the normal design and 60% see it with the modified design. This weighting is done at the beginning of creating the experiment and is called a “variant”.

We also segment by place,choosing in our case Madrid.

As a goal we choose one of those that come by default which is the aforementioned “rebound”. Here I leave you an image with the rest of the default objectives that we can choose. Some of them only appear if you have connected the optimize account with analytics.

Finally you choose if you want to receive notifications of the experiment by mail, if you want the experiment to be shown at 100% of the total visitors of your landing and if you want the event to jump when loading the page or when there is another circumstance.

Done all this, we would only have to launch the event.

**Note that after launching the experience in Google Optimize you will not be able to change some of the rules that you have entered. Among those that you can change is the variant and the weighted percentage of visitors to whom the changes have to appear.

Conclusions of using Google’s tool for live testing

As a general conclusion, I want to emphasize that this Google tool can make the difference between a landing of the bunch and a landing fully optimized to convert.

While it is true that this Optimize can fall short if the web we want to work on is very large and has a lot of data, for 90% of sites it will be more than enough and, possibly, we never need to use its full potential.

One piece of advice I would give you is that when you are going to do the A/B or multivariate tests, try to make the changes made between the original landing and the test landing (in the case of a/b), or the different combinations in the multivariate experiment, are very noticeable. This is a CR premise Or. If we really want to see results, the designs or changes for the tests have to be important to conclude more accurately if one design really is not as good as the other.

I hope you have served this post explaining Google Optimize. Do not hesitate to leave your questions in the comments.