Why You Should Be Using Google Ads Experiments To Scale Your Campaigns [Case Study Included]

As PPC experts, data is our best friend.

Google Ads experiments are an easy way you can build data within your account to make more strategic, performance-driven optimizations. The experiment framework provides insider auction knowledge you can use to better understand how campaign performance will be impacted based on the variable you’re testing. More data in your hands means better insight to scale campaign performance more effectively.

You also maintain more control, saving you time and marketing dollars. Since you choose the campaign, variable, experiment duration, and budget split, you’re using a controlled environment to test a new strategy. This way, if an experiment does not produce the results you were hoping for, you haven’t disrupted the main campaign’s learning phase. 

Experiments offer you the data and time/cost savings to move the needle faster on your campaign performance. 

To help you get started, we’ve provided a summary below on setup instructions, recommended variables to test, a success story, and our 80/20 rule to experimenting long-term.  

What are Google Ads experiments?

Google Ads experiments are a type of testing environment you set up to mirror a current search, display, or video campaign in your account. Experiments allow you to test variables in the live auction for a set timeframe, before deciding whether to make the optimization to your live campaign.

When you build an experiment, Google Ads automatically replicates your campaign you’re choosing to use for the experiment, splits your budget between the test and the original, and only changes the one variable you are looking to test. This could be a new bid strategy, a new keyword match type, ad variation, etc. (we’ve provided a summary of our favourite experiment types below). The experiment will then run in the same auction as your active campaign so you can split test the optimization amongst the same audience – all other campaign parameters remain the same.

How to set up Google Ads experiments

First and foremost, you have to come up with a hypothesis. Yes, we’re going to be scientific with this one! 

By having a proposed outcome in mind, you can better track the progress throughout the experiment and align the experiment set up with your goal. Here are a few examples:

• By adjusting my bid strategy to a Maximize Conversion with tCPA I am going to increase conversions while maintaining the same CPA.   
• By testing out “60% Off” vs. “$15 Off” in RSAs I am going to see an increase in CTR and CVR. 
• My testimonial video ad will achieve more conversions within my YouTube remarketing campaign vs. my close-up product video ad.  

There are different experiment types you can select, which have different testing capabilities to them. Below is a walkthrough of how to set up an experiment using these various experiment types:

1. Navigate to the “Experiments” tab within Google Ads 
2. Select the type of experiment you want to create: Optimize Text Ads, Video Experiment, or Custom Experiment 

• Optimize text ads = RSA experiment
• Video experiment = YouTube Ad experiment
• Custom experiment = Custom Display/Search experiment 

3. Select the campaign on which you wish to run the experiment.
4. Create a name for your experiment that indicates what variable you’re testing.
5. Specify the variable you want to test depending on the type of experiment you selected above, the experiment goals and experiment split: 

Note: we suggest using a 50/50 traffic and budget split, and a cookie-based experiment split as shown in the screenshot above to limit the same user seeing both versions of your campaign, giving you more accurate data. 

6. Set your run dates and launch. You’ll be able to review experiment performance in your main campaign analysis table, as well as under the Experiments tab. 

What are common variables to test in an experiment?

Best practices when setting up experiments

• Have a clear hypothesis in mind going into the experiment so you can measure performance strategically. 
• Apply a 50/50 budget and traffic split to ensure you’re giving enough weight to your experiment.
• Give yourself at least 3-4 weeks to run the experiment so you gather enough intel.
• Only test one variable at a time in any given experiment to maintain accuracy within the experiment. 
• It’s ok if the outcome isn’t what you hoped for – now you have more data at your disposal to continue testing.

[Case study] Google Ads experiments – a brand campaign success story

A client of ours is an industry leader and captures a large share of the overall Canadian market in their industry. Competitors bid aggressively on their brand name to try and steal market share, and as a result, we have to be aggressive with our brand campaign to maintain top-of-page visibility. 

We noticed increased competition within the auction for their brand campaign over the course of three months and as a result, our cost-per-acquisition (CPA) was steadily increasing. This became a concern as historically the brand campaign drove lead volume at a much lower cost.

We chose to run a Google Ads custom search experiment with the goal of decreasing the CPA so that we could generate more conversions at our same spend level. 

Up until the experiment we were using a Target CPA bid strategy, however with the CPA increasing and top-of-page impression share decreasing, we wanted to test out a new bid strategy. Our hypothesis was that by changing our bid strategy to Target Impression Share, Google could better prioritize our positioning in the auction to generate more branded search traffic and conversions, bringing down our overall CPA.  

We set up the experiment to run for 30 days and saw clear results indicating CPA performance would improve by switching the original campaign to a Target Impression Share bid strategy. We implemented the bid strategy change in the active campaign once the experiment was over and after another 30 days we saw substantial increases in Brand campaign performance.

As a seasonal client, we compared year-over-year results for the same 30 day period to see the impact our new bid strategy had on the brand campaign: 

This was just one experiment and one variable tested, yet the growth in lead volume was significant. By running the experiment first, we were able to strategically optimize the original campaign without risking a bid strategy learning phase that could impact peak season performance for this client. 

Develop a long-term testing model using the 80/20 rule

We recommend thinking long-term about experimentation and setting aside 20% of your total monthly budget to continually test each month. This 80/20 rule allows you to maintain consistency with 80% of your budget, while using 20% of your budget to uncover new opportunities that can be applied to your ongoing efforts to make the 80% even more profitable long-term. 

Keep in mind Google Ads Experiments are just one of the many testing frameworks you can leverage to scale campaign performance. There are plenty of other tools out there that you can integrate to gain a more holistic testing model across your marketing funnel. Creating a holistic testing model offers you further insight into your customer journey across all your landing pages and advertising channels.  

From Google Optimize, to Hotjar Screen Recordings, to Facebook’s A/B testing capabilities, there is much more to explore. We’ll save that blog post for another day!