
“Father Of Advertising”
Table of content:
- Expanded text ads aren’t really going anywhere.
- What responsive search ads really are.
- They can have significantly more impression volume.
- They can have significantly lower conversion rates.
- Lies, damned lies, and statistics.
- Does A/B testing still make sense for responsive ads?
- How many ads should you test per ad group?
- How to test: 3 types of RSAs to start with.
- What to test: working on promises, in themes.
- Location Insertion. Use it.
- The hidden opportunity.
Who’s on the saddle matters more than the saddle itself.
Advertisers.
We tend to worry about the wrong things.
But first.
I must apologize to everyone to whom I copied and pasted the same response about how to deal with Responsive Search Ads. After doing that a few times, it’s been clear to me that I should put down my thoughts publicly for anyone interested. If that’s you, thank you for your trust.
In short:
What you put in little boxes matters much more than how many little boxes there are to put in things.
That’s it.
You can skip the rest of this article to something more productive, as this is all I’m going to repeat.
And some will disagree but…
1.
Expanded text ads aren’t really going anywhere.


Just ads.
I can buy cheap tickets to Hell, presented to me in …
Expanded Text Ads.
That’s what searchers still see.
In fact, what searchers see are just ads.
What happens in the background is none of their concern. Either the ad is interesting enough to get a click, or it’s not.
It doesn’t matter that you’ve used 15 assets instead of 3 or that a robot said your ad strength is poor. If the ad gets clicked on, it will show.
And yes, you can still control exactly what the ad contains.
And yes, you can do that by pinning all assets.
And no, that does not make you a control freak.
And no, you shouldn’t worry about ad strength.
Exhibit AFake ETA vs Real ETA.
Here’s a responsive ad that was turned into an ETA by pinning all assets. We’ll call it the “Fake ETA”, as only 1 combination can result of all the pinning. It is competing just fine against a real ETA.

It’s ok to pin all assests.
The expert eye will have a few more questions about this test, which I will answer in the comments.
It’s just one small example. You can, and should, convert your best performing ETAs into “Fake ETAs”.
Get something that works for your best campaigns, and take things from there without worrying that you’ll no longer be able to create ETAs.
Exhibit BBetter ad strength doesn’t mean
better performance.
A short case study by iProspect shows that Ad Strength doesn’t really mean much. Even if Google says otherwise.

Ignore ad strength.
Ad strength is just another invention by Googlers, designed to get advertisers to do what they want them to do. Just like the optimization score. Both, completely empty metrics to me.
What matters still, and always will, is whether people choose to click on your ads more often than other ads. If they do, Google will happily show it, pinned or unpinned, short or expressive, expanded or responsive.
It’s about what’s in the ad, not its form.
2.
What responsive search ads really are.

It’s simply a form of ETA testing on steroids. Controlled by Google. With Google’s best interest first, yours second.
And that’s ok.
You’re simply creating a myriad of Expanded Text Ads, and letting Google decide which it likes best.
And that’s ok too.
It’s always been this way anyway. Google does and has and will, always show the ad they believe will generate more money for them, before they show the one that will profit you more. It’s been this way even when you explicitly choose to “rotate indefinitely”.
That’s ok because you do get rewarded for it, and if your targeting is right, it’s just one small part of your conversion funnel, of which you have full control.
So you’re letting Google test on your behalf, with the assets you provide. You have 100% control over which assets you provide. In exchange, you get a testing engine that will test those assets for you.
Given enough time, Google will show one or a couple of variations more often than the others.

Click to view larger image.

After 5,000 impressions in the last 90 days, you will find that your RSA is simply just two or three ETAs accounting for more than 80% of the traffic.
Two or three ETAs competing against each other. Just like it has always been.
One important thing to note is that you can have up to 15 headlines in an RSA, and up to only 4 descriptions.
So many headlines, so few descriptions.
A responsive search ad is mainly a headline testing engine, that will help you find the headline combinations that Google likes best — quicker.
We’ve been trying to simplify things in Tenscores, by showing the hierarchy of ad combinations after enough impressions have been gathered and leaders found (almost ready).
3.
They can have significantly more volume.

RSAs can get more volume than ETAs.
According to Google, 7% more conversions on average. Others have found they can get 300% more impressions, and at times even as high as 10 times more.
Cannibals!
Whether they cannibalize impressions away from their ETA sisters, or whether they magically attract new impressions, this is important information to know and understand.
One theory is that because of all the potential combinations, they get a better auction-time Ad Rank, which puts them in front of more search queries more often and thus gets them more impression volume which hopefully translates into more conversions as well.
This is all good when it works.
Until it doesn’t.
4.
They can have significantly lower conversion rates.

Beware, the increased volume may or may not translate into more conversions. In fact, it can get really bad if you’re not watching.
Exhibit CMore impressions at the cost
of conversion rate.

A little less conversation, a little more action, please.
One example of how bad it can get.
Need another?
Exhibit DMore conversions at higher
acquisition costs.

Can you guess why the RSA has such a poor conversion rate? (Same landing page)
This last one might seem less obvious.
An RSA with 45 conversions at $60 each, while the ETA gets 29 conversions at $19 each.
Were the additional 16 conversions worth the additional cost? Your business will decide.
5.
Lies, damned lies and statistics.


The point I’m trying to make is that it’s not the format that matters, it’s the content. And that should be the only focus.
What if RSAs generally have a better CTR and more volume than ETAs? When you look at averages, that might be the case indeed.
Until it’s not.
Exhibit EAn ETA can get more volume
if it wants to.

Would it be the same if the ETA was a pinned RSA?
Some will say that RSAs are better across the board and will show you data to prove it. Some will say that ETAs are better across the board and will show data to prove it. And they’ll be right, according to the data, but beware of Simpson’s Paradox.
What’s missing in statistics on ads, including the one I shared above on ad strength, is their individual context. The people they are targeting, the intent those people have, their needs, their desires, their dreams… and the message that’s put in front of them, which will work or not work depending exactly on what it contains.
A responsive ad will work or not work, based on who it speaks to and what you put in it, on a case-by-case basis.
You can find an RSA that will outperform a winning ETA if you work on it, fine-tune it continuously until it’s just right. And you can do the exact same thing with an ETA.
6.
Does A/B testing still make sense for RSAs?

Of course. Why would it not?
Google has started suggesting advertisers only have one RSA per ad group, nothing else.

To stop testing is to stop improving your ads.

The most important word in the vocabulary of advertising is TEST. If you pretest your product with consumers, and pretest your advertising, you will do well in the marketplace.
Twenty-four out of twenty-five new products never get out of test markets. Manufacturers who don’t test-market their products incur the colossal cost (and disgrace) of having their products fail on a national scale, instead of dying inconspicuously and economically in test markets.
Test your promise. Test your media. Test your headlines and your illustrations. Test the size of your advertisements. Test your frequency. Test your level of expenditure. Test your commercials. Never stop testing, and your advertising will never stop improving.
David Ogilvy, “Confessions Of An Advertising Man”, page 86.
Responsive or not, not a single ad can contain all the possible messages, promises, offers that you can test on your audiences. Not one.
We test ideas, angles, promises, offers. And we test them in different themes, that let us learn about what makes our potential customers tick. The learning is for us, not an engine.
One RSA per ad group, does not, and will never, cut it. I hope all the exhibits above are enough to show you that things can go in various ways. If you’re only testing one RSA per ad group, then you’ll never know what you’re missing out on.
And the idea that Google shows the right combination to a specific user on the fly as the search happens using predictive factors that only the algorithm knows, doesn’t have much teeth for me. As we’ve seen, eventually, only a handful of combinations get most of the exposure.

Exactly.
7.
How many ads to test per ad group.

Just two.
My opinion has changed over the years on this. When Google suggested 3 ads per ad group, I was fine with it.
Not anymore.
A/B testing is not as straightforward, and as easy as we make it look. I’m guilty of this. False positives are rampant.
There are many things that can make an A/B test go wrong, and most test results are, in fact, wrong. That’s why, often, when applied they don’t actually produce the long-term performance they initially suggested.
It becomes exponentially easier to get a test wrong the more variations there are. The more variations, the more data you’ll need, the longer you need to run the test, and the chances that you get it wrong are higher.
What to look for in an A/B test are big changes in specific KPIs. If I’m trying to improve Quality Score, doubling or tripling my CTR is what I’m looking for, and I wait to see the QS effectively change to confirm a test a winner.
If I’m trying to reduce my acquisition costs, I’m looking to half the number, with better landing page conversion rates, while keeping tabs on the quality of conversions and continuously re-assessing if the test is valid.
Despite that, I still get my tests wrong all the time.
And that’s why, for the vast majority of small business owners that I talk to, and who follow this blog, two ads are the number. Not more, not less.
The exception used to be that if the ad group has a lot of traffic and it can gather an enormous amount of data in a relatively short period of time, then 4 or even more ads were fine. But with responsive search ads, one ad is already too many ads for any scenario that I can imagine.
It’s harmless to test more than two, nothing is going to break if you do. In fact, Google will choose to show one more often than the others. It’s only when you get serious about making your message as effective as it can be, that you test two angles at a time for the best chance of not ending up with false results.
This should be a topic for an entire post. In the meantime, I will direct you to the work of Martin Goodson: a very insightful PDF (bottom of that post) and a video. There are more.
8.
How to test: 3 types of RSAs to start with.

Responsive ads are just a new tool to learn how to use to reach our objective. Apply the options below to specific cases, and use them when appropriate, closely watching how they serve your audiences and your goals.
There are many ways to use pinned items, the following are the ones I would suggest trying.
Option #1The Loose RSA
This is the RSA as it is intended. 15 assets, no pins.
It’s the first option to test because it can work.
The format itself isn’t what matters. It’s what you put in, that does.
Take all your best headlines, add a few more. Take your best descriptions, add a few more. Pit it against your current ETA in the ad group, watch and see if you get better performance.
It can be structured around a theme, or not at all. Throw it all at the wall and see what sticks.
It is ok to try, and most of us already have, because as we’ve seen before, there eventually comes up a winning combination, from which we continue the work.
Important: I wouldn’t rely too much on this form to test all my messaging. I’ve seen combinations never get more than a couple of impressions. Google does choose what they’re going to test from our assets, and some combinations might never get the chance to prove their worth.
Option #2The Fake ETA
I like cheese, so I’ll call it FETA.
Pin all items and recreate an ETA from an RSA.
If you have assets you already know work, use them.
This is for those, for a reason or another, like full control, who don’t need the extra potential volume and are just happy with the continued use of ETA.
If you’re in an industry with very expensive click costs or many regulations, this might be the only option for you.
If you pit this version against the loose RSA, chances are the loose RSA will cannibalize all impressions, and this one won’t show much.
Unless it has a much higher CTR.
And so the best way to test a FETA against an RSA is to put completely different messaging in the RSA, in search of new combinations that might work better than the one you already have in the FETA.
If you prefer to play with a bit of wiggle room, use the one below.
Option #3The Partial RSA
Recommended!
In this one, you will pin your assets by position.
- Pin 3 different headlines in position 1.
- Pin 3 different headlines in position 2.
- Pin 3 different headlines in position 3.
This is probably the safest of them all. A good middle ground. You give Google some material to test, while still controlling how your message is displayed.
You’re not giving full responsibility to Google to come up with sentences that make sense. You’re giving Google your own ideas, and letting it do the testing for you, in correct language.
If you don’t want Google to show weird combinations, this is the option. You can have a structure:
- Headline 1: 3 different ways to talk about the keyword
- Headline 2: 3 different ways to offer what you promise
- Headline 3: Your brand or call to action
It’s an example. You can create your own template from a winning ad you already have, and use it throughout your account as a starting point, then have everything tested against that initial template.
All other options derive from this one. You can choose to pin just one headline and leave the others loose. You might want to have your brand name always appear in Headline 3.
It’s up to you.
9.
What to test: working on a promise, in themes.

This one’s for Vijay who asked:
” Is there a chance to do something like multivariate testing with the 24 Ad variants you proposed in your pdf book (may be User-centric titles in a RSA and the Product-centric titles in another RSA or some other kind of split), and quickly know which variant or the set of variants shall perform well? Would love to hear your thoughts on this.”
Vijay Veerubhotla, XOXODay.com
The templates he is referring to are here. I particularly appreciate the question, as it allows me to update those ideas, given what I’ve learned since I wrote it, and given the opportunity of RSA.

One promise, 4 themes, mutliple headlines.
Greed and fear drive many, if not most, decisions.
We’re looking for all the things we can promise the searcher in need, and find different ways to present those promises, in quest of the angle that will be the most captivating.
This framework is the simpler version of how to write ads, as it doesn’t require too much research. If you want a deeper framework, that starts with research about the emotions behind a keyword, read empathy.
A framework helps get away from tunnel vision that often makes us write ads in one direction. Ads that are too similar to make any kind of progress.
My example: “Buy ticket to Hell”.
We’re going to have fun with it, not to be taken too seriously.
We’ll go with 3 promises:
- Cheap flight tickets
- A comfortable journey
- A holiday to remember
Those 3 promises can be tested as 3 separate responsive search ads, or combined in one.
Let’s start with the first promise.
Cheap flights
I use the thesaurus to look up synonyms of “cheap” so I have ideas of other words I might use: low-cost, affordable, economy, low-budget.
Using the diagram, I write some headlines:
Theme 1 — Towards Gain, spotlight on Searcher
- Go To Hell For $331 Only
- Low-cost Ticket To Hell: $331
- Economy Class Ticket To Hell
Theme 2 — Towards Gain, spotlight on Brand
- Lowest Fares For Any Date Range
- We’ll Fly You To Hell & Back, Low-Cost
- Norway’s Largest Airline
Theme 3 — Away from Pain, spotlight on Searcher
- Save Big On Your Flight To Hell
- Get 30% Off Your Ticket To Hell
- Be Gentle On Your Wallet
Theme 4 — Away from Pain, spotlight on Brand
- Norwegian Airline Has Lowest Costs
- Don’t Pay More Than You Should
- No Extra Fees
We don’t have to adhere strictly to the themes, it’s just a way to get our creative juices flowing, let the mind do the rest.
From that first promise, we get 12 initial headlines that we could feed the engine in one RSA and let it figure out what it likes best.
Then the second promise.
A comfortable journey.
I look up synonyms of “comfortable” for other words I might use: cozy, pleasant, comfy, snugly.
I look up what defines a comfortable flight:
- Seat size
- Seat back screens
- Wifi
- Outlets
- Food
I write my headlines:
Theme 1 — Towards Gain, spotlight on Searcher
- Go To Hell In Full Comfort
- A Cosy Flight To Hell
- A Comfy Flight To Hell
Theme 2 — Towards Gain, spotlight on Brand
- Large Seats & Movie Screens
- Free Wifi & Outlets
- Delicious In-Flight Food
Theme 3 — Away from Pain, spotlight on Searcher
- A Journey That Feels Like Home
- Need A Break? Let’s Fly To Hell
- Bored of {Location(City)}? Let’s Fly To Hell
- Hell: Perfect Getaway from {LOCATION(City)}*
Theme 4 — Away from Pain, spotlight on Brand
- We Make You Forget Your Fear Of Flying
- We Make Flying To Hell Feel Like Home
- Norwegian Takes Away Fear Of Flying
There goes material for the second RSA, focused on the second promise.
I test the two RSAs against each other, looking for the angle that is the primary driver of a ticket purchase to Hell.
I use what I learn from this test to write a 3rd RSA. This one could combine the best of my first two RSAs or I might test my 3rd promise, as the contender to the current champion.
I think you get the point.
Once you have a good idea of which promise, and which theme works best, you can find better ways of expressing yourself by using headline formulas (questions, how-to, listicles, news, etc). We’ll discuss this in a future post if you’d like. The above should be enough to start.
10.
Location insertion.
Use it.

To do this, we used to create ad customizers attached to a spreadsheet.
Not anymore.
It’s a feature I found that most are not aware of, even though it’s been there for a while.
I decided to add this section after I got this message:
For the [search] campaign – I found this: Best Homes {LOCATION(City):Atlanta} as one fo the headlines – did you guys add that? I don’t believe I did – and that’s really cool. So that’s like a dynamic keyword if someone types in a specific city it’ll replace ‘Atlanta’ with the actual city name?
From a customer
It’s even cooler than how he described it. The user doesn’t have to type in the city.
Just like keyword insertion, a line of code inserts the user’s location in the ad. And you can choose to display the user’s city, state, or country.

Beautiful.
It’s nice to have. Use it.
11.
The hidden opportunity.

You might remember this quiz, in which switching headlines in an ETA led to a 4X increase in CTR. Responsive Search Ads do this automatically for you. It’s an experiment you no longer need to conduct.
Or this one, where one headline was simply omitted. RSAs also do this automatically, another experiment you don’t have to conduct.
We’ve been used to responsive ads for a long time. I’ve always loved them for display ads. It makes the creation and testing of new ideas a bit easier. That’s what a machine is supposed to do.
For search, it’s a good option, to be able to throw things at Google and let the engine do the work of figuring out what sticks. It can be expensive, just as it can be rewarding.
Of course, we’re not going to pretend that it’s all roses.
Google keeps shoving things down our throats, like killing phrase match, in complete disregard of advertisers’ preferences. Always creating situations that raise our vigilance, alarmed that the changes their force on us might interfere with our objectives, which they often do.
That’s not making our lives easier.
And we also shouldn’t act like we’re powerless when they make their decisions. Google does change course when there is enough backlash. Remember device targeting.
This is another case where they could have done things differently if they wanted to. They still can.
I might be completely delusional but so long as we’re not penalized for pinning all assets, we’re fine. But that line might have already been crossed and I don’t have the full picture.
The silver lining is that we can test more, and more often.
Advertisers.
We have a tendency to worry about the wrong things.
I see it all the time. Customers come to me asking for advice on how to improve Quality Score, in an account where tracking is set up incorrectly or not set up at all. Where keywords are attracting the wrong audiences, and with ads that are never renewed.
I pledge to take this opportunity to take my own ad copy testing to the next level, and to share it with you when I can, so you can worry about the right thing, the searcher.
If you appreciated this and would like to support us, link back to it next time you write, it helps our rankings.
See you on the next one.
Thanks to everyone who sent me a message about RSA, I wouldn’t have written this article if it hadn’t been for you, and I’ve learned a few things in the process that I’d like to write more about. You’ve given an appetite that I didn’t have. Thank you. 🙏 (March 5, 2022) Edit: Added tweets by Julie F. Bacchini and Melissa L. Mackey.