How to Split Test Google Shopping Ads



Would you like to increase your profitability through testing? Split tests or AB testing are tools for you to use to find out how certain changes influence your conversion rates.

Google tools are already standards for marketing and web development. And many of them are free and accessible from all over the world. One of them is the Google Ads Experiment feature.

In this article, we will focus on showing you how to split test Google shopping ads effectively. You can do this yourself, ask your team to do it, or look for professional assistance.

Let’s find out how to go about split testing Google Shopping Ads. We’ll start by giving you some general info on AB testing and five important ways to test your campaigns.

What is split testing?

Split testing, also known as A/B testing or bucket testing, is a type of research method used to prove a hypothesis. Simply stated, you get two audience groups which you divide into two segments. One is the control group for which you change nothing. The second is the experimental group. You use them to test different website layouts, content, emails, ads, etc.

Of course, AB testing takes time and money, which may mean it’s not a viable option for everybody. The logistics also imply that it’s a complicated process for which you need resources. But it’s one of the most effective and safe ways to carry out changes.

AB testing is a very common tool used when launching new medicine or cosmetics. That’s why it takes so long for new formulas to launch on the market. But you can apply it to many other areas, like marketing.

AB tests give insight into customer preferences regarding your products and users. They can help you improve conversion rates and figure out what other changes must be made to keep improving your rates.

How to plan split tests for your Google shopping ads

Before starting a split test, there are a few things to consider. You may feel inclined to skip the planning stage, although it’s by no means a waste of time. The test itself will take a while, so you should spend enough time planning things to be sure you get the results you’re looking for.

If you’ve already worked on ideas, it’s great to prioritize them. Figure out how important the aspect you’re testing is. Then, think about how well it will work compared to your control version and how much time it will take to achieve it. Think about which idea should be your number one priority after a short analysis.

Next, figure out a timeframe for your test. When can you expect significant results? Can you test your idea in due time? With pre-test analysis, you can estimate how long your AB test takes. And you can take measures when implementing your change release schedule.

Most likely, your AB tests take weeks and any changes you make can impact the results. You can test more segments at the same time as long as they don’t influence each other.

Plan for your tests and those of your colleagues. You can be very disappointed when, halfway through your ads test, you have to make changes.

Tools used for AB testing

Let’s see what you can use to complete your AB tests. Most of the tools we’ll mention are meant for AB testing on your website. Some can even aid in changing your website. You’ll usually need the following:

  • A meeting to gather your team, brainstorm and write down ideas
  • Tools to help you make experimental ads
  • Tools to split website traffic into two groups
  • Calculator for pre-test and after-test results.

Let’s see now you can benefit from these tools and how your business or website can benefit, too. Let’s see how to interpret your split test results.

How to interpret split test results

You have to equip yourself with a lot of patience. Every result will have some degree of uncertainty. You’ll calculate statistical significance to know how sure you can be that your result is due to chance or not. It’s all a matter of calculations and percentages.

Use an AB test calculator to calculate your results. To check how uncertain the results can be, you can also run AA tests. You’ll most likely be able to see differences on identical variants after a few days, helping you keep your test on track when you want to end it sooner than usual.

How to split test product information

You can split test product information as a means to optimize your product feed. Product information testing refers to images, titles, and extensions. You can do it with feed management tools or manually, using Google Merchant Center. After you decide what to test, you’ll need some groups of products.

Then, you’ll use one of the available methods. Let’s see what the two most popular ones have in-store:

1. Cluster analysis

This type of analysis takes into consideration certain performance factors. You divide your groups based on historical performance metrics like revenue, clicks, costs, and conversion. You can keep track in a spreadsheet or programming language to make things easier for yourself.

2. Random split

This analysis method is done based on product ID. When you use numeric values for product IDs, it’s easier to assign a test group to even numbers and another test group to uneven numbers, for example.

The essential part is to have an equal number of products for each group. That way, your test metrics will be close. Once you have the split, you can make the changes in the ads for your test group. Be sure you can report on any of your ad or product IDs and their corresponding groups. You can easily analyze them and find the most successful ones.

How to split A/B examination campaign configurations

You have to do more than just split based on product IDs when you want to check campaign configurations like ads or targeting. In an ideal test, you offer the sale product to your control group and your test group. Performance can vary based on products.

Test your settings using one of the following three split methods. Remember to focus on creating equal groups and to do the data cluster analysis beforehand. Focus on differences when you perform your analysis.

1. The Customer Match split system

Customer Match allows you to focus on first-party audiences in Google Ads. The method consists of uploading a list of targeted emails from your database and Google will match them to accounts.

You can make a cookie split in your preferred CRM when using Customer Match. The next step is to create two campaigns by using two audiences. Keep your original control campaign and change what you want to test in your second campaign.

When using Customer Match, you have the option of running a third campaign for potential clients that are not in your database. You’ll retain any potential conversion while running your test.

The customer match split is a random test usually found in A/B tests. It offers reliable data simply because your test subjects are randomly split.

Of course, the Customer Match method has its drawbacks as well. One is that it’s difficult to implement. Not everyone has the resources to split in their CRM randomly or to implement customer matches. And it’s not complete because it only applies to the customers in your database. That means your results aren’t necessarily valid for new customers.

2. The geo split test

You use the geo split method to find campaign uplifts. By implementing this method, you’ll see if there’s value in advertising on branded keywords, for instance.

For geo tests on Google Ads, you divide your market into smaller regions (geos). Each geo is assigned to a test or control group. The users in a geo group see the changes you made to your campaign.

The test groups see the usual (control) campaign. You can make the split based on region or country. Just make sure they are well correlated. It’s strongly advised to use cluster analysis to determine your two groups.

It’s easy to set up geo groups. You just need two very well-correlated groups. If you have a little Google Ads experience, you will have no trouble setting it up. You split the same campaign, so the result has little influence.

Consider seasonality in this instance, especially when you’re selling sensitive gear. Also, it can be difficult to do appropriate cluster analysis on location targeting when using smaller geo-locations.

3. The campaign split

This method implies splitting your accounts into two groups with high correlation. The groups need to have a similar number of key metrics like costs or conversions.

The test group gets the changed experience while the control group sees your current campaign. When you track both campaigns, you’ll see the differences in performance, if there are any. And you’ll be able to make additional changes.

The biggest advantage is that you can easily implement the campaign split. All you have to do is change the settings for half of your campaigns.

Unfortunately, it’s not without drawbacks. This split method is not very reliable, especially because seasonality can affect different campaigns. And preparing for it can prove difficult. Cluster analysis is harder to do when you have smaller sets of data at your disposal. Of course, you can always look for professional assistance. If you’re located in California, the best Los Angeles SEO companies are an option.

Common mistakes with A/B testing

Keep in mind that AB testing is nuanced and depends on many factors. You’ve already read about the practical aspects of split testing. Think of the most common mistakes that come up:

  • You make changes that are too small.
  • You test with a small sample of users.
  • You follow AB test best practices blindly.
  • You initiated a test without a thoroughly-planned hypothesis.
  • You end your test too soon.
  • You test too many changes at the same time.
  • You stop your AB test after a first fail.
  • You don’t share your results with the rest of your team.
  • You ignore the results you don’t like.


The quality of a split test result depends on your primary assessments and setups.

Follow the five approaches meant to aid you with the optimization of your Google Ads campaigns and see what results you get.

When you want to succeed and focus on your data, there’s a high chance you’ll find a way to get the results you’re looking for in no time.