I really do love running experiments in Google Ads. I get super annoyed when I make a change, say adding new ad variants, and then the new variations get no traction. Google often won’t give them a chance to see how they would do if they got even distribution with the existing variants. Enter Experiments. They are a great way to see statistically valid results and feel confident about your test. If you haven’t created one recently, you should just to enjoy some of the new features that Google has implemented to make creating and implementing them even easier than ever. I will highlight some of those features today, making it super easy for you to get started with your own experiment.
What I love…Ease of Set-Up
Google Ads has made it easier than ever to set up experiments. Before, you used to have to create drafts and then apply the draft and, and, and. It wasn’t always easy and we were often finding errors in the set-up. There are two key ways experiments are easier to set up now.
1. Auto-Suggested: I’ve seen at least two places where Google has suggested an experiment and will do all the work for you to set it up in just a few clicks. One was in the campaign settings for campaigns still on manual cpc to see if I’d like to test another bidding algorithm.
The other has been right in the experiments section itself.
In that case, once you click on the “Create Experiment” button, all you have to do is name the Experiment, set start and end dates and hit “Create Experiments” – it’s that easy!
2. Manual Creation in the Experiments Section: If you navigate directly to the Experiments section, you can launch an experiment in minutes. There are three types of experiments:
Text Ads: These are great if you want to test one component of an ad across multiple ad groups and/or campaigns. You can quickly do this via Find and Replace, Update URL or Update Text rules. There are numerous examples of what you could test but below is one to give you an idea.
Video Ads: For these, it’s best to test different videos but use all the same campaign settings. Users in each experiment arm won’t overlap with users in another experiment arm. Note: For measuring results, the experiment arm with the lowest cost per conversion will be the winner, and 100 conversions per experiment arm are needed to determine the results. You can have up to 4 experiment arms.
Custom Experiments: Test anything and everything! These are my favorite but they are only available for Search and Display at the moment (I can’t wait for PMAX!!!). The set-up is so much better than it used to be! Pick your campaign, add a “suffix” to differentiate it from the control campaign, make your changes (which will save as a draft) and then click schedule to determine the final touches such as your goals. 
My favorite setting is the “Enable Sync” one, which allows changes that you make to the control campaign to automatically be made to the experiment. This is great for your regular maintenance tasks such as adding negatives.
What I love…Ease of Reading Results
Once you have an experiment going, you can easily see the results once you click on it in the Experiments section. Here is an example of an experiment that is currently running for one of my clients.
You can clearly see that the control campaign is going to win, but we can tell the difference in key metrics across the board and also, the confidence level of the data. Google Ads says this is only at the 95% level, but in many cases, I don’t believe you have to get to the full 100% confidence. I had wanted to let this experiment run a full month but I’ll actually end it early. Regardless, Google shows you the results in a manner that makes it so easy to see how the experiment is doing. Additionally, I can easily click to apply the experiment (which rolls it out and pauses the control) or end it, with just one click (see options in the bottom right of top section in the above image).
What I love…Creating More Experiments
There are times recently when I’ve been unsure of how to improve a campaign. When that happens, I recommend to the client that we just test it to see what happens. Some clients have been more open to this than others especially because you win some and you can lose some (like we did above). However, I don’t like letting campaigns just run so we’re constantly thinking of new things to test and never settling for the status quo. Not sure what to test? Here are a few ideas:
- different landing pages
- campaign settings
- geographies
- match types
- bid strategies
- CTAs
- RSAs
- DSAs
- audiences
Some of these have been of much debate lately, especially with all the changes in automation. Not sure about using Broad Match keywords? Why not test it? We did, but that’s a future blog post. Stay tuned. 😉
