This is a guest post from Will Cozart, Labs Analyst at
Advertisers were somewhat skeptical of the new Responsive Search Ads (RSAs) when they were first announced at Google Marketing Live 2018, as the format represented one more step into an automated search marketing world. So, what are RSAs (see this article with the full 411), how do they work, and what kind of performance have we seen now that the ad type is fairly mature?
Responsive Search Ads: a Quick Refresher
Google’s RSAs simplify the ad copy creation process by allowing machine learning to manipulate a single ad, containing multiple headlines and description assets, to serve the best possible combination for each user in each individual auction. More details on implementation and specs in the link above.
What We’ve Seen
In early testing comparing RSA performance to existing expanded text ads (ETAs), 3Q account teams didn’t see very encouraging results. While our holistic agency testing confirmed these results, evidenced by one of the test accounts below (an 18% lower CTR for RSAs? Yikes!), we realized that it didn’t paint the full picture of the ad type’s impact on an account’s performance. So, we changed up the testing methodology.
We decided to look not only at the performance of the different ads next to each other, but the overall metrics of campaigns with both RSAs and ETAs in every ad group, to those with ETAs only, all else being equal. We tested with RSAs containing only existing ETA assets to negate any impact from the copy itself, and used Google’s Drafts and Experiments tool to split auctions between the control and experiment campaigns.
The findings were now much more promising, and in line with Google’s claim of “up to 10% more clicks and conversions when using with Expanded Text Ads.” This test included 15 accounts in different verticals, and included nearly 40 different campaigns. It’s also important to note that all accounts were tested using the “optimize” ad rotation setting.
You’ll notice RSAs tended to provide the most benefit on the Non-Brand side. As the ad type is able to adapt to individual auction dynamics, advertisers are eligible for additional impressions they might not have had by using ETAs exclusively. Perhaps a combination of assets available only through the RSA format gives your ad just enough additional relevance to win an impression not possible with the existing ETAs.
On the Brand side, advertisers tend to capture a much higher share of available impressions already, so our accounts didn’t see much of a performance improvement from RSAs when looking at volume deltas for the median campaign. Some of the accounts saw extraordinary results with RSAs on the Brand side, skewing the average numbers.
Overall, our campaigns saw traffic lifts from RSAs that came in close to Google’s claims, and even improved efficiency slightly. By allowing Google to manipulate our RSA assets and choose the “best” ad to serve to each user, in each auction, in conjunction with the “optimize” ad rotation setting, our search accounts are bringing in more traffic to our clients’ sites. RSAs take very little time to get up and running, so why turn down these gains?
How to Test and Optimize
Curious about RSAs and want to try them for yourself? Have you tested them in the past, but analyzed performance at the ad level and decided ETAs still reign supreme? Have no fear! It’s never too late to get started and go after those incremental performance gains for your account.
Start by adding RSAs using your existing ad copy into each ad group of a campaign (or multiple, or all of them), and look at the performance of the campaign in question in sum. This can be done a couple of ways:
- Use the Drafts and Experiments tool (preferred) to evenly split the auctions. You’ll also get the nifty performance comparisons and statistical significance calculations right there in the UI.
- Look at ad group performance before/after adding RSAs (if you don’t expect any major outside shifts in performance, like promotions, competition, or budget changes).
While we’ve established that adding in an RSA per ad group, alongside ETAs, is a best practice for search campaigns, your work doesn’t stop there. Check out your RSAs’ asset performance to see which ones are serving most, and least, often. Consider replacing low-serving assets with new, meaningfully different ones.
I’m not talking about swapping out a single word, but the entire line of copy. Don’t be afraid to go in a new direction—you might be surprised! RSAs are powered by machine learning, so they might need a bit of time to ramp up, but we typically saw performance normalize around one week after launching. Good luck testing!