AdWords Ad Copy Testing Strategies

AdWords Ad Copy Testing Strategies

When thinking about AdWords ad copy testing strategies, how often should I test different ad copy ideas and what are typical ideas to test?

When I approach AdWords ad copy testing, I am testing with intention. Don't just have two versions of the ad "to see which does better". This doesn't teach you anything, and you are experimenting to LEARN.

Formulate a question.

"Do Free Shipping call-outs help CTR?"

Then write a pair of ads that test this theory. Run the ads side by side with no optimization enabled till it reaches significance, or your arbitrary "it's a wash" threshold. Never test more than one question at a time, unless you're a statistician. It's sexy to say you're doing "multi-variate testing", but you're probably just making your test take longer to run, and making the math harder to get right at the end. Plenty of people say that you should "always be testing at least two ads", but if you don't have a question you want to answer, then don't fix what ain't broken.

How often you test should be based on how often you have new hypotheses. That being said, you're likely to want to do a few rounds of testing when you start a new campaign, possibly in just the first few weeks. After that, testing will likely become less frequent, as you've already iterated toward a more effective Ad. From then on, you're only really working to avoid letting your Ads get stale, or when a good question pops into mind.

Remember, however, that CTR for CTR's sake is not the goal. Conversions (and therefore revenue) are. If you're selling beach sandals, but your Ad talks about summer vacation, then you might get a great CTR, but the lack of relevance will result in a very poor conversion rate. In a more extreme example, imagine if your Ad were entirely irrelevant to your landing page, but otherwise attractive to clicks--"Free Cruise!" would draw clicks, but few would convert, since you're not actually offering them a cruise. You're offering them sandals.

Therefore, what you're looking for is actually the Impression Conversion Rate. How many of your Ad's impressions are likely to convert, on average? This is going to be a very small number, obviously (I'll often multiply it by 1000 to make the reports easier to read), but it takes into account the combined effects of both the CTR and Conversion Rates, resulting from your Ad. If you're using an A/B test calculator, then all you have to do is swap in the impression count for where you'd normally enter the clicks--from there, the math is the same, so determining significance is the same.

In the long run, you'll want to do some testing on new Ad Groups to get in the right ballpark, but fiddling with Ad copy just for the sake of "testing" is rarely productive. If you don't have an idea, a question, a hypothesis to test, then invest your time somewhere else--somewhere with higher return. As with all A/B testing, the majority of changes simply don't matter. That does mean that when you uncover one that does move the needle, it's all that much more exciting, though.

Have questions? Drop me a line via the Contact form--I love to talk shop!