Easy and efficient – testing changes with AdWords Campaign Experiments

Posted by Matthias Wobrock on 23 Feb 2011

Testing and marketing go hand in hand – if you sell a product or service, you want to know how to best reach your customers to generate as many sales as possible. This is where testing can help you. You might want to test different layouts of your brochure before sending a final version to all your subscribers, or optimize the text and images on your campaign landing page using Google Website Optimizer.  If it is not your site, but your AdWords campaign that you want to test, you have a new best friend in your AdWords tool box: the AdWords Campaign Experiments (in short: ACE).


While testing brochure designs or variations of a TV spot require you to carry out user studies and form experiment groups, testing online is free of charge, fast and efficient. Your website visitors will not even know that you are testing different versions of your campaign. Previously, the only way to test two different settings on an AdWords campaign, ad group or keyword was to use one version, and then the other, for a certain period of time each, and then compare the results – unless you wanted to go through a complex setup with two slightly distinct campaigns running on different schedules. Testing one version after the other is called sequential testing and can result in extremely flawed data. You can never be certain if your change is performing better or worse because of the actual changes made or external factors you cannot fully monitor or control – these could be seasonal effects, news articles, or even the weather in the destination you are advertising in. With ACE, Google will split traffic between your variations in real-time, and positive or negative external effects that might influence your campaign’s performance will affect both versions you are testing. Your final results will therefore still yield actionable data.

ACE gives you the functionality to experiment with any combination of bid, keyword, ad group or placement changes on the Google search and content network. It will help you make important campaign decisions based on the right information. To quote the words of famous web analyst Avinash Kaushik: “I wish I could put into words how much I love ACE. It is truly a blessing for anyone that does paid search marketing. So many companies large and small truly suck at doing AdWords properly […]. This sucking can be solved immediately and awesomely by using ACE.”

What will the impact of adding certain new keywords be? What if I changed the bids on my top keywords? How would my traffic numbers change when I used a different match type for some of the terms in this ad group? These are just a few examples of where ACE can shine. It helps you to better understand the impact of changes and lowers the risk of testing new advertising strategies. If your test didn’t bring the desired results, you can just end your experiment, and start a new one.

You can find the ACE feature under the campaign settings tab in the AdWords reporting interface, and most information on it in the announcement of ACE on the AdWords blog, Google’s official ACE page and in the ACE video tutorials.

There is one powerful experiment setup that is not documented that well. Ad group experiments allow you to not only test keyword bids or different keywords, but manifold other elements you might want to experiment with – for instance, your original ad group versus a range of more granular ad groups with similar keywords, or different sets of copy and banners. You will have to use ad group tests for this. Create copies of your existing ad groups, change every element you want to test on your copied groups, and start a campaign experiment with your original groups set to “control” and the new, copied ones set to “experiment”. I recommend that you don’t focus too much on CPCs and costs per conversion using this method (unless your new versions contain changes that might alter them directly) and more on your actual conversion rate, as click costs will naturally be lower on your control groups which have already been running for a while and have a positive account history paired with good quality scores.

Lastly, give your experiments time to run and try to not stop your experiments too early. You might experience ups and downs in the beginning that will eventually even out. The more data your experiment gathers, the more precise your result will be. Put some sophisticated math behind the changes you make to your AdWords campaigns, start a testing culture within your account or agency and you can be sure that you will be making just the right changes to your campaigns based on accurate data.