Using AdWords to Effectively Test Campaigns

AdWords offers online marketers the perfect platform for testing many aspects of our marketing campaigns. The ability to run different ad variables in unison or identical ads to different landing pages can offer us a huge amount of insight into what works well, and what doesn’t when we run our marketing campaigns.

Unfortunately, the 19^th Century British Prime Minister, Benjamin Disraeli couldn’t have been more portentous when he coined the phrase, "There are three kinds of lies: Lies, damned lies, and statistics." And we know well, how many statistics Google offer us.

The real problem that testing presents for us though isn’t in misleading statistics, it’s much earlier in the process. The most common problem we come across in online testing is poorly designed or inappropriate tests.

Testing is a hugely powerful tool that can have a massive impact on the success of our online campaigns so it really pays to make sure that the tests we carry out are well considered and capable of asking the questions we need answers to.

Clearly Define and Limit the Test Parameters

One common problem with testing is that we try and test too many variables at the same time. Take for example the AdWords ad copy. Often we’ll test two entirely different versions of an ad. Doing this limits our ability to understand and interpret the results to maximum effect.

Having changed both titles and descriptions to both ads it’s impossible to know which individual elements made a positive or negative difference and by what degree. In this scenario we could run many tests that would provide us with much more comprehensive data on each individual element to build up a picture of the perfect ad.

Let’s look at a single title and think about the variables we could test. For the purposes of this example we’ll use the title, “Earn more with AdSense”

Test 1: Change the order of the words (with AdSense earn more)
Test 2: Change the case of the words (Earn More With AdSense)
Test 3: Turn into a question (Want to earn more with AdSense?)
Test 4: Turn into a Capital Case Question (Want To Earn More With AdSense?)
Test 6: Change to an answer (How to earn more with AdSense)
Test 7: Change into a Capital Case Answer (How To Earn More With AdSense)

These are just a few possible tests to the title of one AdWords ad. There are many more including things like punctuation, word placement, etc. but this should provide a feel for the extent we can test one, individual element.

Consider the Testing Window Carefully

It’s often mistakenly believed that the longer a test is continued the more accurate the result will be. The problem with this logic is there are two distinct factors that occur during testing, internal ones that we can control (Ad copy, landing page, etc) and external ones (competitor pricing, market demand, seasonality, etc) that we can’t control.

Ideally, tests should be run in parallel to lessen the potential of an external variable skewing a single result. However, this isn’t always practical, particularly when comparing data as part of an ongoing improvement strategy.

External conditions should be tracked along testing periods to add context to the results. Things that may not appear significant during the testing window can suddenly become very important with the benefit of hindsight.

There will be an optimal period of testing where you can have high enough exposure to the test without having too much risk of external interference but what that is varies dramatically depending on what’s being tested. It’s up to you to work out where the balance is in your business.

Monitoring During the Test Window

You should start monitoring the results as soon as the test starts and continue to do so throughout the testing window. It may be that an obvious trend starts to occur a little way in, prompting you to change something mid test to assess the impact.

Monitoring during the test window can help you quickly adapt to the results as they occur. You can then decide if you should eliminate some elements when there is enough evidence that they are not producing the results you are looking for.

Share Your Results and Challenge Assumptions

Testing with AdWords is a great way to test your theories and market knowledge by using real world scenarios directly affecting your business. It doesn’t live in bubble though and cannot work in isolation.

You should always take your findings back to the business, compare the results sets with other analysis, zoom out from the intense level of detail and contextualise the results. Only by sharing and reviewing what we’ve learned can we start to think about what questions we need to ask next time.