5 Things You Should Know About AdWords Campaign Experiments

AdWords campaign experiments are a really great, controlled way to test how changes you make to a campaign affect performance. You can test just about anything — bid changes, different ad copy, a new campaign bidding strategy, and the list goes on.

First, I’ll briefly cover how to set up an experiment and review performance metrics, then I’ll offer a few tips that you should know before you dive into campaign experiments.

Setting Up an Experiment in AdWords

To run a campaign experiment, you first have to create a Draft version of the campaign in which you’d like to make and test changes. The fastest way to do this is to choose the campaign you want in the lefthand control panel of the interface, then choose ‘Drafts > Create New’ next to your date range button in the upper righthand corner:


You’ll then be prompted to name your draft campaign, which can reflect the type of test you’re running, such as Target CPA Test:



You’ll then be taken to the Draft version of the campaign where you can make needed changes for the experiment. Once you’re finished making changes, you can choose the blue Apply button in the upper righthand corner. At that point, you can either choose to apply the changes to the original campaign or create an experiment:

If you choose to run an experiment, you’ll need to name it, select a start and end date, and decide what percentage of traffic you want to go to the original campaign and what percentage you want to go to the experiment campaign:

And that’s it for the setup!

Experiments are extremely quick and easy to set up, so if you’ve been letting ideas of a complicated set up hold you back from running experiments, now you know it’s an easy process!

Reviewing AdWords Experiment Performance

To see how your experiments are performing, just select the ‘All experiments’ option in the lefthand control panel:


After that, just click on the campaign you want to see data for, and you’ll see Original vs. Experiment stats at the top:

In order to know if the differences in metrics are statistically significant or not, you’ll need to look at the icons next to each metric. Google provides this handy chart for explaining what the icons mean:

Based on this chart, I can see in my example above, that impressions for the Experiment campaign are statistically significantly lower than for the Original campaign. This is certainly something to keep an eye on, but considering this client is focused on lead gen I should let this experiment continue to run until enough data is gathered to determine performance differences in Conversions, CPA and Conversion Rate.

5 Things You Should Know About AdWords Experiments

Now that we’ve covered setup and performance review basics, here are five things that you’ll want to know before you get started with campaign experiments.

Campaign Negative Lists Don’t Carry Over

This one is extremely important.

If you are using campaign and/or ad group level negative keywords, those will get copied over since your experiment campaign is derived from a draft campaign of the original. However, if you’re using campaign negative keyword lists you will need to go in and add your experiment campaign to the list. If you don’t, then your performance stats will pretty much be a wash since negatives are covered for your original campaign but not the experiment.

Changes Made to the Original Campaign Aren’t Reflected in the Experiment Version

If possible, it’s best not to make changes to the original or experiment campaign versions while the experiment is running to help keep performance data clean, as all campaign elements would remain the same and identical outside of the piece you are testing. However, if adjustments are unavoidable you need to be sure to manually mirror the changes you make in the control campaign in the experiment campaign as well.

Not All Campaign Features are Supported

According to this AdWords help post, ad customizers that use “Target campaign” or “Target ad group,” as well as the target search page location, target outranking share, and target return on ad spend (ROAS) automated bidding strategies aren’t supported in campaign experiments.

Only One Experiment at a Time Per Campaign

You can schedule up to five experiments for a campaign, but you can only run one at a time. This makes sense in terms of keeping the experiment data clean; if you have multiple tests running, it will be difficult to pinpoint which change caused statistically significant differences in performance metrics.

AdWords Editor Adjustments and Views are Limited

In AdWords Editor, you aren’t able to edit bid strategies in campaigns with experiments. You also can’t create, pause or remove an experiment, nor are you able to see control vs. experiment statistics. For these types of adjustments, and to view performance, you’ll have to resort to the AdWords interface.

AdWords experiments are a really useful tool to test how changes affect campaign performance. There are some nuances, but once you are aware of those the setup and launch are easy processes. Nothing is holding you back from setting up meaningful, statistically relevant experiments in your accounts!

Have you tried out AdWords experiments? What tips do you have? We’d love to hear them in the comments!