Improve Your AdWords Campaign Performance Using Advanced Campaign Experiments

Digital advertisers are always trying to find new ways to achieve better performance. When working with large accounts with thousands of keywords and hundreds of campaigns, you won’t want to risk a drop in performance by testing new strategies on the whole account. Therefore, you test them on a small portion of your traffic, before rolling the changes out if the test is successful.

To help with this, AdWords released AdWords Campaign Experiments (ACE). This blog post will introduce some of the possibilities offered by this tool, as well as provide a practical guide to setting up an experiment.

Introduction to AdWords Campaign Experiments

The AdWords Campaign Experiments (ACE) feature allows you to test changes on a limited section of your activity, whether that be ad groups, keywords, landing pages or ads. In this way, the ACE feature offers a controlled environment in which to run your test. It’s important to note that ACE is not available at campaign level.

Campaign experiments work very simply: choose what you would like to test and select the percentage of your auctions to undergo the changes from a list of possible distributions. Based on this percentage, Google will decide whether your control or experiment split will take part in a given auction.


Improve Your AdWords Campaign Performance Using Advanced Campaign Experiments


How to set up an experiment

First, choose the campaign in which to run your experiment. In the settings tab under advanced settings, open up the experiments section to define your experiment settings. As you see in the picture below you can specify:

  • Your test name
  • The portion of your actions which will participate in the test
  • Start date (if you don’t select one, you will have to start the test manually)
  • End date or default end after 30 days


Improve Your AdWords Campaign Performance Using Advanced Campaign Experiments


Once you have defined the settings, you can add the experimental changes to your keywords, bids, ads, or ad groups and assign them either to the experimental or the control group.


Improve Your AdWords Campaign Performance Using Advanced Campaign Experiments


If you run your test on existing keywords, ads, and ad groups, they will appear by default in both your control and experiment groups. If you want them to appear either in the control or in the experiment group, select the control or experiment test tube which appears in the status column. When adding new keywords, ads, or ad groups to your experiment, select “Add as experiment only” and mark them as experimental.

Improve Your AdWords Campaign Performance Using Advanced Campaign Experiments

Up to 1,000 changes are possible. Depending on the kind of experiment you want to run, this may be important. So, if you want to add new experimental keywords, beware that every keyword will be counted as one. If you want to test a large number of keywords, you can also define an entire ad group as experimental so that it will be counted as one change, regardless of the number of keywords it contains.

It is also very important to know that the settings applied at keyword or ad level will be the same as those of the ad group they are in. If you mark a keyword as “experiment-only” (blue test tube) in a control-only ad group (grey test tube), it will receive no impressions.


Improve Your AdWords Campaign Performance Using Advanced Campaign Experiments


Once you have added the desired experimental changes, assigned your keywords, bids, ads or ad groups to the control or test groups and set a start date, the test will run for the selected time.

Setting up an A/B landing page test

The ACE is particularly suitable to run A/B tests for ads. You can test text changes to see if new ad copy receives more clicks than your current one, or test a new landing page for your ads. The easiest way is to track clicks is by adding a ValueTrack tag type={aceid} to the tracking template or final URL of your ads or keywords. To do this, Google AdWords must be connected to a Google Analytics account. An ID-number will show up in the referral URL in Google Analytics so that you can see whether the click came from the control or the test group.

When testing new ad copy or landing pages, be mindful that tests can impact your keywords’ quality score. If your test ads or landing pages perform worse than your current ones, they will negatively affect the QS of your keywords. It’s important to remember that the existing ads in your control group will have historical data and your new ones won’t. To create a clean “experiment environment” when testing, try the following:

Pause the original ad group so that the historical data won’t affect the test. Next, duplicate the ad group twice to create one control and one experimental ad group, in which you can test new ad copy or landing pages. Google will allocate your actions to the control or test group based on the distribution you have selected. Once you have run your test, you can reactivate the original ad group. One problem you may encounter is a slightly higher CPC, since the duplicated ads have no history.

Limitations of ACE

One problem with these experiments is that they don’t work with flexible and advanced bid strategies applied to the keyword, ad group, or campaign levels, even if it’s the same strategy. If you have applied a bid strategy such as enhanced CPC or automatic bidding, you must remove it to launch the test.
If you get the error “Experimenting is not supported for your current bidding option”, you have bid strategies applied at some level. Simply use the filter “bid strategy source” at keyword and ad group level to see which bid strategies are applied. Remove them by selecting “Use campaign bid strategy,” which must be set to the default “manual CPC”.
This problem is a shortcoming of the test feature, since it forces you to disable bid strategies which might be appropriate for your campaign goals.

Keep in mind that the test feature is still a beta feature. It may not yet cover all of the aspects that you need or desire to test. Remember also that changes at campaign level are not possible and
that the feature doesn’t currently work with flexible or advanced bid strategies so that you will have to remove them before starting an experiment and use manual CPC instead.
In general, we recommend running tests on small portions of your auctions, since the test may negatively affect keyword performance. For the same reason, we advise you not to run experiments during important seasons for your business.

We hope this article has been useful to you. We’d love to hear about your experiences with AdWords Campaign Experiments!


Luana Quaresima

Luana Quaresima joined the crealytics Account Management team in 2015. She is a digital marketing enthusiast with a 3+ years’ experience in e-commerce, SEO, PPC, content and social media marketing across different verticals such as retail, fashion and tourism.

  • Experiments is an excellent feature to test out certain hypotheses, quickly. Though, managing multiple experiments can be cumbersome as you have to keep tabs on all the live experiments.

    Here’s a script to monitor performance of multiple experiments easily. With this script, you can easily find out performance of multiple experiments with respect to original campaign on various important metrics like Clicks, Conversions, Conversion Rate, CPA, etc.