October 8, 2021
The Ideal QA Process in Digital Advertising
As someone responsible for your business’s or client’s Google Ads account, you probably have a list with different ideas for optimizations. Often after implementing these optimizations, you might have doubts about whether your changes in the account will improve performance or not. Sometimes the change you are making is just a bulk change in the account, and you’ll evaluate performance by comparing period over period data sets. While in most cases this works perfectly fine, there are instances when your changes could lead to account disruption or bad results altogether. With Google Ads Draft and Experiments, you can avoid this by testing your ideas before actually implementing them on a larger scale.
Let’s say you have read some positive reviews about Google Ads smart bidding. You are eager to activate enhanced CPC across all your campaigns and drive significant improvements to account efficiency, but you don’t want to make such an impactful change to your account in one big swoop. Often the safest way to try and test new strategies in Google Ads is with Draft & Experiments.
So where to start first? Pick your high-volume campaign, and create a Draft and then an Experiment campaign with 50/50 split.
Since the Experiment campaign is entirely new and has no historical data, you should run it for a couple of weeks before enabling the eCPC setting. Consider the Experiment campaign as an entirely new campaign in your account – it has no historical data, no quality score. That means that directly comparing your existing campaign and experiment campaign in the first week after launch can give an inaccurate view. For some account changes it may not be that important, but for smart bidding, you should consider collecting some historical data before starting the test. It will ensure that your results are accurate, and you will have a lot more confidence when trying to determine the winner.
While testing is in progress, Google Ads will provide all the necessary information about Experiment performance in the interface. You can find a quick comparison dashboard that will show the difference between your main campaign and test group. Long gone are the days when you had to use an online A/B testing significance calculator. Google Ads will display all this information right in the same view. If any of the metrics reach significance, Google will mark it with the blue asterisk symbol. It even shows the level of significance and confidence interval.
When you decide to end the test, you can apply the Experiment campaign structure to your original campaign by just clicking the Apply button, or end it and leave the original campaign unchanged.
To launch your experiment successfully, make sure to follow the tips below:
· Before setting up a test, you should think about your hypothesis and specific goal you want to achieve with your experiment. It will help you figure out the main KPIs that you will analyze when determining the winner.
· Set the experiment end date to be after your actual anticipated test end date. This will ensure that your test does not end before it was able to gather enough data. You can always pause your test manually, and this is a recommended way of doing it. You wouldn’t want to end a test before it was able to reach significance.
· Avoid making changes to both the original and experiment campaigns. You don’t want to intervene in the testing process because it will skew the results. There are some exceptions to this rule; for example, when you want to safely test your vendor’s bidding capabilities and run a third-party bidding algorithm against Google’s bidding strategies.
· Experiments work at the campaign level. If you have a case where you want to test a new feature across your whole account, we would recommend running experiments for just a few campaigns to start. Otherwise, you will have to set up Google Ads experiments for every campaign in your account.
· If you are using a third-party bid management platform, your experiment campaign will appear as a new standard campaign. Name your experiments wisely so that it will be much easier to find them later.
· Experiment campaigns will not carry any historical data from your original campaigns. That means they will start gathering all historical data from scratch. Keep this in mind, especially when analyzing Quality Score and overall performance during the first couple of weeks of the test.
· An experiment will not increase your campaign budget; instead, it will use the existing amount that was set in your original campaign. This amount will be split based on the percentage of campaign traffic (either search-based or cookie-based) you send to the experiment campaign versus the original campaign.
There are plenty of different tactics you can use experiments for, including:
· Device settings and bid adjustments
· Landing page test
· Location targeting and bid adjustments
· Ad extensions
· Negative keywords
· Structural campaign changes
· Remarketing audiences
· Display interests
· Display placements
· Display topics
While the experiment tool allows you to test most of the possible scenarios, there are some limitations that you must keep in mind before starting. For instance, experiments are available only for Search and Display campaigns, and you are only allowed one active experiment per campaign. (You can schedule up to five experiments at a time, but activate only one). Below are some other features that are not available in Drafts & Experiments at this time:
· Ad schedule reports
· Category & Search terms
· Auction Insights
· Display Placements report
· Scheduled email reports
· Ad customizers that use “Target campaign” or “Target ad group”
· Keyword bid simulator
· Some automated bid strategies:
· Target search page location
· Target outranking share
Considering the amount of optimizations accounts are receiving on a daily basis, it is worth using Drafts & Experiments to test more significant changes. Even with current limitations, the tool provides a quick and easy way to launch tests for Search or Display campaigns. By using this powerful testing feature, you can strategically approach every major change in the account in the safest possible way. Regular testing is the only way to drive constant optimizations and account improvements, so make sure to keep Drafts & Experiments in your account manager toolset.
Looking for help figuring out what to test next in your account? Contact our team for an account analysis.