Ever since the dawn of Adwords, advertisers have had the option to run more than one ad variation per ad group. This is one of the features that makes Adwords so attractive – the ability to test different ad copy and landing pages against a set of keywords and learn which performs the best.
Until this week, though, there was always a problem with the definition of “performance.” For Google, and for a handful of advertisers, “performance” is defined as “click-through rate” – the ad that generates the most traffic. But for most advertisers, “performance” is defined as “conversion rate” – the ad that generates the most desired actions, commonly known as conversions.
Until this week, Adwords offered 2 options for serving multiple ads: Optimize or Rotate. Rotate is simple to define: your ads will rotate evenly among impressions, with each variation getting approximately the same number of impressions. So if you have 2 ads, each will display on about 500 out of 1,000 impressions.
By default, the Optimize setting is turned on – changing it requires editing your Campaign Settings. And “Optimize” sounds great: after all, everyone wants to optimize their campaigns, right? Ha, wrong. Optimize (until this week) rotated ads based on click-through rate: the as with the highest CTR would, over time, be displayed on a larger proportion of impressions. It’s not uncommon to see as much as 80-90% of impressions going to one ad with “optimize,” meaning the ad with the lowest CTR barely gets shown. It’s also not uncommon for the ad with the best CTR to be the ad with the worst conversion rate – so you end up spending a lot of money for not very many conversions. Not good.
But what if you were an advertiser who wanted to drive traffic, with conversion optimization as a secondary goal? What if your ad test is just starting out? What if you’re a new advertiser and you don’t even realize you have a choice?
Good testing principles indicate that all test variants should be shown to test samples that are relatively equal in size and demographic. For instance, if an ad only shows to females age 18-34, there’s a good chance the results won’t translate to men age 45-54. So you want to divide up your sample 50/50, and make it random. But if your Adwords ads are set to “Optimize,” that’s most likely not going to happen.
Never fear, though – Adwords to the rescue! This week, Google launched a third option: Optimize for Conversions. Finally, after years and years of advertisers asking for a way to serve the best-converting ad more often, Google came through! Right?
Sort of. As with many Adwords features, there are a few caveats. First, note in the documentation this important caveat: “If there isn’t sufficient conversion data to determine which ad will provide the most conversions, ads will rotate using ‘Optimize for clicks’ data.” Yikes. It’s pretty obvious that most ads will amass a statistically significant number of clicks long before they reach a statistically significant number of conversions. So really, any new test is doomed to start Optimizing for CTR – thus messing up your conversion test results from the start.
Also, the word on the street (or at least on Twitter) is that Optimize for Conversions will optimize based on conversion rate, not number of conversions. So you may have an ad that gets a great conversion rate, but not many clicks; or vice versa. Either way, the system could be making the wrong decision about what’s working for you.
So for now, I’m sticking with “Rotate,” even though Google warns me every time that it’ll ruin my results.