Expanded text ads (ETAs) aren’t going anywhere just yet, but Google’s latest experiment to make responsive search ads (RSAs) the default option is a sign of where things are heading. Responsive search ads are indicative of the broader shift to machine learning and automation, where bids, ads, landing pages, etc. are powered by algorithms designed to predict better outcomes than we can manually.
The promise of RSAs is that when you feed the system a bunch of unique headlines and descriptions, it can test and learn to predict the right combination for each search. Google also talks about the convenience of RSAs, the fact that they can show more text than ETAs and their opportunity to compete in more auctions when there are high quality score combinations of copy that match more queries.
With the introduction of RSAs, Google began encouraging advertisers to move away from “overly fixating” on click-through and conversion rates and instead focus on the incremental lift in clicks and conversions from RSAs. Tinuiti’s director of research Andy Taylor wrote about this on Search Engine Land last year and concluded that click-through and conversion rates in fact don’t matter. Why? “Any conversion rate can be acceptable, so long as the price paid for the clicks results in an effective cost per conversion,” he wrote.
By design, comparing the performance of RSAs to ETAs is tricky because they are two different animals. If you’re evaluating RSAs on incrementality, their conversion rates might be lower than ETAs but the efficiency of those conversions might be better — lower cost per conversion, higher margin and/or lifetime value — and come from impressions your ETAs weren’t eligible for. But measuring this is far from straightforward because the reporting on RSAs is limited and there’s no way to easily tie a query to an ad much less an RSA combination. Frederick Vallaeys of Optmyzr explains how to measure the impact of RSAs with query mix analysis (one more reason to lament the loss of longtail search query data) using a script, but Google doesn’t provide incremental measurement for RSAs natively in the UI.
After we learned about Google’s RSA test, I asked marketers what they’re seeing from this ad format as we head into the fourth quarter.
RSAs are not an easy button
Brad Geddes, co-founder of ad testing platform Adalysis, said more advertisers are using RSAs, but overall, he says, looking at any metric such as “CTR, conversion rate, CPI, ROAS, etc., and ETAs win much more often than RSAs.”
Geddes said advertisers that follow ad copy best practices do see RSAs performing closer to ETAs. “The only time it’s close is when the advertiser is testing RSAs that have essentially the same lines in the RSA (and never more than 6 headlines, often with 1 pinned) as they do in the text ad.”
Geddes says the way RSAs are framed as being a great simplifier is also a problem. “One of the big benefits that is touted for RSAs is time savings. This has led many advertisers to just use the same RSA in many ad groups since Google said it’ll figure out how to serve the ads. This is also where some of the performance issues of RSAs come into play.”
Account structure still matters
Account structure and ad group organization makes a difference, too. “When the account is not well set up or the same ads are used in many ad groups (or worse the entire campaign) so that the ads are not as relevant as they should be within an ad group, then RSAs often outperform ETAs,” Geddes added. Google limits the number of enabled RSAs in an ad group to three — two is not recommended. “It seems Google has some real ad serving and machine learning issues when an ad group has multiple RSAs in it.”
“The RSA news is disappointing,” said Atlanta-based digital marketing consultant Josh B., “but the biggest key I can provide is to ensure that your headlines and descriptions are still relevant to the ad group. I know that’s search 101, but I don’t trust Google to ensure that the best ads will be served and I really ensure that I turn off ‘auto-applied’ ads when I create a new account.”
Doug Thomas, search specialist at Trailblaze Marketing, says he’s just starting to come around to RSAs. “The way to approach RSAs is a shift in mindset from testing individual copy to testing copy strategy,” he said. In testing, “At the ad group level, our largest-by-impressions campaign gained about 12% IS [impression share] and our largest ad group in that campaign gained about 18% IS when we implemented RSAs with that simpler method of multiple copy strategies in one RSA,” said Thomas.
Pay attention to pinning
Andrea Cruz, digital marketing manager at KoMarketing, and others stressed the usefulness of pinning headlines. “Generally I am not a fan of them but one tip that I always share is to make sure the CTA is pinned in one of the headlines. Because you probably don’t want to have an ad that doesn’t state what you are going to get or what the action to take on the page is I have found helps with conversion rate.”
Azeem Digital says he maxes out every headline and description when testing RSAs and pins headlines when he wants to ensure the brand always shows, for example. “I’ve had some surprising results with headlines that I didn’t expect to do well outperforming the rest,” he said.
Advertisers should be testing RSAs to understand when and how to use them or find copy to optimize their ETA ads, agrees Geddes. “If Google ever gives us proper RSA stats and reporting, then their usage will also increase. As Google isn’t being transparent with conversion metrics based upon how the RSAs are being rendered, it is also hard for some people to trust them.”
Geddes expects that when ETAs do go away, “a lot of accounts will see a decline in performance…and others will just recreate the ETA experience through pinning ad lines.” He said many large clients have asked them to create a feature that essentially allows them to recreate ETAs with pinning. “That’s how much many of our larger advertisers prefer ETAs to RSAs,” he said.
The post RSAs: Are they living up to the promise? It depends appeared first on Search Engine Land.