Optimize For Conversions Using The AdWords Rotation Option

AdWords recently introduced a new feature to their Ad Rotation Settings called “Optimize for Conversions” (February, 2011).

AdWords Advanced Settings Ad Rotation – Optimize for Conversions

To find this go to: Campaign Tab | Settings | Advanced Settings | Ad Rotation (expand by clicking the “edit” link)

Many advertisers may find this to be a marked improvement over the old “Optimize for Clicks” option. However, if you are already pursuing an ad optimization strategy, you might still find some advantages in your solution. This article covers both situations.

This article is intended to be a brief introduction to the feature and a discussion of the real world tradeoffs we might expect to see, as well as some practical recommendations for the basic mechanics of ad optimization.

What Does Ad Rotation Do?

Put simply, this feature will measure our ads on effective conversion rate (CTR x Conversion Rate), and gradually aim the vast majority of our traffic to the top-converting ad in the Ad Group. Non-performing ads will obviously get less traffic. (There has been some confusion out there about which Conversion Rate is used: 1) CTR x Conversion Rate, i.e.: from impression to conversion, or 2) Conversion Rate only, i.e.: from click to conversion. Thanks to jim Prosser from Google for clarifying. This features uses 1) CTR x Conversion Rate. Additionally, when there is not enough data, it defaults to CTR only, just like the old Optimize for Clicks option.)

Furthermore, if we create a new Ad, it may get a fair shake, but it will usually have to be appreciably better than the existing winner to wrestle the traffic away.

For some advertisers, this may be all good news, and possibly better than whatever the current ad optimization strategy is. For others, this feature fits in between the old “optimize for clicks” feature and our current strategy.

Real World Tradeoffs

Similar to the “Optimize for Clicks” feature, “Optimize for Conversions” is going to measure one metric, and send the majority of the traffic to the ad that maximizes that metric. Again, this sounds like great news, but there are some real world tradeoffs worth mentioning.

For starters, AdWords tends to be really aggressive with how it shares traffic to the winner. While very similar ads may get similar traffic share, it is usually the case that AdWords has chosen a winner that sees the vast majority of the traffic, as in 80-95% of it, even when it appears like the data just is not there.

Confusingly, the same ad in another AdGroup might be determined a loser and get virtually no traffic, for no obvious reasons. While it is possible the algorithm is just that smart, this one often looks like a case of simply being quick to judge, or at least not being consistent in judgments, and then being really aggressive about acting on that judgment. (Obviously, these are characterizations based on data I have seen over the years – and certainly your experience may vary.)

In this sense, we are setting AdWords loose. It might not make the best decision, but it will definitely act on it.

One area of confusion in the industry has been about whether the Ad Rotation options consider only the ad, or the ad + the keyword, like the Quality Score system. Again, thanks to Jim Prosser from Google for clarifying, the Ad Rotation options use only the Ad data. This feature does not consider the keyword.

Still putting aside whether AdWords makes a decision that we can rationalize with the data we see, when it comes to splitting traffic, it is going to split the traffic a varied amount. In one version of a well designed test, we split the traffic evenly and consistently so we can actually measure the performance of the ads directly with each other without attempting to scale the performance of one versus the other based on how much traffic each one received, or how that varied over the duration of the test.

That means by using either of the optimize options we would hamper our ability to measure performance in a controlled environment.

You may not find the optimizing for conversions is the best idea. Allow me some license to make this point. Let us imagine an ad that reads something like “Do not click here unless you have a gun to your head and are going to buy my stuff.”

You might imagine that most people will not click, but the ones that do will probably buy, (perhaps right before calling the police.)

Bottom line; there are some ads that may have a high conversion rate, but that might not be the best overall ad for your business. If you are already running your own ad testing strategy, you are probably familiar with this dilemma.

It is not CTR alone, and it is not conversions alone, but some combination of metrics, that maximizes my money earned at the end of the day. Whatever metrics you choose, you may need to set aside AdWords’ optimize features in order to do your own rigorours ad testing.

Then again, they are both welcome additions to the family, and a far cry better than “nothing.” At the end of the day I think we can all the agree the one strategy we should absolutely not pursue would be to choose the “Rotate Evenly” option, but then fail to do any optimizations of our own.

At any rate, this option is still fairly new as of this writing, I am sure there will be plenty of information and expert advice out there soon enough.

Editor’s Note: This article was updated after its original publication, following clarification from Google Adwords.