Google announced the global rollout of their AdWords Campaign Experiments (ACE) feature on Monday. ACE is a useful feature that helps you understand the impact of certain changes to your AdWords account.
There are numerous benefits (and some drawbacks) to using ACE.
Previously, when you launched a change within your AdWords account, you had to use a before-and-after approach to your analysis. This meant that you would make the desired change to your AdWords account, and compare your performance before and after the alteration.
With ACE, you can split test two versions of a campaign or ad group simultaneously to see how your changes influence your performance in real-time. The real-time testing aspect of ACE is beneficial because it can help mitigate the influence of outside factors on your tests, such as other sales/marketing initiatives, changes in the competitive landscape, and seasonality fluctuations (by day, week, or month).
Basically, running an experiment with ACE is like split testing your PPC text ads. If you have two ad texts within your ad group, each variation can receive 50 percent of the impressions and you can determine which one garners the best results. ACE works similarly but instead of testing ad text variations, you are split testing variations of your entire campaigns or ad groups.
You can't test everything within your AdWords accounting using ACE. Let's look at five campaign elements that work well when tested with this feature.
1. Ad Group Structure
You can break down ad groups into more finely-tuned themes and test to see how this impacts your performance. You can take an ad group that has 25 keywords (this is your control variation), and break it down to five ad groups (this is your experiment variation). You can also move keyword themes around to see which variation results in the highest click-through rate (CTR) and conversion rate.
2. Keyword Match Types
You may want to experiment with different match types. If you haven't tried phrase or broad match of certain terms, you can add these match types to your ad group and then launch an experiment to see how this impacts your CTR and conversion rate. Instead of inserting these match types and hoping for the best, you can maintain your original ad group (your control variation) and launch a test variation with your new match types.
3. Keyword Expansion and Reduction
If you want to see how new keywords impact the overall performance of an ad group, you can launch these new terms as an experiment. You can see how the ad group performs with and without these new keywords.
On the flip side of the coin, you may want to remove keywords to see how this impacts the other terms with an ad group. Do the remaining terms get a lift in impressions? Does CTR increase or decrease? You can launch two versions of your ad group: one with these keywords intact (your control version) and one version without these keywords (your experiment version).
4. A Note on Google Instant Experiments
Brad Geddes wrote a great article (as usual) on the bg Theory blog about using ACE to test new keyword suggestions from Google Instant results. Here's his process for testing keyword suggestions from Google Instant (good tip for those advertisers who have the budget to devote to a test that may increase spending):
- Choose a few of your root keywords (generally the two-word variety)
- Type these words into Google instant or Web Seer
- Make a list of these words
- Break them into appropriate ad groups
- Enable ACE (AdWords Campaign Experiments)
- Create new ad groups with these words
- Make these brand new ad groups part of your experimental campaign
- Let the experiment collect data
- Measure the results
With this experiment for including shorter term keywords, you should utilize exact match. This way you know exactly which terms are generating the traffic and results.
5. Keyword Bids
You can make aggressive keyword bid adjustments to see how significantly different bids will affect your campaign's results. For example, you keep all of your keyword bids the same in your control ad group, but within your experiment variation you can increase (or decrease) your keyword bids aggressively. With this type of experiment, you can see how elevated bids would impact the overall performance of this ad group.
Remember, you can tell ACE the percentage of traffic that you would like to dedicate to your experiment variation. Also remember to monitor your tests closely. If you launch an experiment and it tanks, your overall performance will suffer.
Some of these limitations will be eventually be address by Google. For now, here's a list of the shortcomings:
- Number of experiments: You can only run one experiment within one campaign at a time.
- Campaign settings: You can't test campaign-level settings. For example, you can't test CPC bidding against CPA bidding. Or, you can't test the ad delivery method settings of Standard against Accelerated to see which one works best.
- No AdWords Editor or API support: You can manage experiments only within the AdWords web interface.
- Reporting and analysis: This is where the feature lacks the most. Within the AdWords web interface you can segment the view of your campaign so that you can see the control variation performance against the experiment variation. This is helpful but it's difficult to dig into these numbers any further than what is displayed within the ad group section of AdWords. If launch a large campaign structure you can't see how all of the control ad groups compare to the experiment ad groups. You will have to look at them on a one-by-one basis or an ad group by ad group basis. This can make the experiment analysis process laborious. This isn't such a large drawback that it should hinder advertisers from using ACE, it's just kind of a pain.
To learn the nuts-and-bolts on setting up an experiment within your account, the AdWords Help section is quite thorough. A series of videos will walk you through the process of getting started with ACE. Now, you need to determine which elements of your AdWords campaign would benefit most from these types of testing -- and get out there and start testing (and improving your PPC performance).