Responsive Search Ads: How to Create Better Ads in a World With Only RSAs

Episode Description

The end of #ExpandedTextAds is here.

So how should you prepare your accounts going forward? Should you still create some last-minute ETAs? And how can you create better #ResponsiveSearchAds?

Get these questions answered and many more, plus learn actionable tips you need to structure and test RSAs in this PPC Town Hall featuring Julie Bacchini and Mark Irvine.

In this episode you will learn:

• How to structure your campaigns with RSAs?

• How to create the best performing RSAs?

• Tips for using Ad Variations

  and so much more.

Episode Takeaways

How to Structure Your Campaigns with RSAs

• Consider consolidating ad groups around themes rather than specific keywords, reflecting the shift towards broader targeting capabilities of RSAs.
• Transition from single keyword ad groups (SKAGs) to more theme-based ad groups (STAGs) to better utilize RSA’s dynamic content capabilities.

How to Create the Best Performing RSAs

• Start with a smaller number of headlines and descriptions to allow the RSA algorithm to optimize effectively before scaling up.
• Pinning elements in RSAs can help maintain control over ad content, ensuring critical information is always displayed.

Tips for Using Ad Variations

• Use ad variations to test slight modifications of text within RSAs to fine-tune messaging without overwhelming the system with too many options.
• Regularly review and refine RSA components based on performance data to optimize the delivery of relevant and effective ad content.

Additional Takeaways

• Embrace the flexibility of RSAs but maintain strategic control by limiting variations to those that are meaningful and aligned with your campaign goals.
• Continuously monitor and adjust the RSAs based on performance insights and keep communication open with stakeholders about the implications of automation in ad creation.

Episode Transcript

Frederick Vallaeys: Hello and welcome to another episode of PPC Town Hall. My name is Fred Vallaeys. I’m your host. I’m also co-founder and CEO at Optmyzr, a PPC management software suite. So PPC Town Hall is where we get to talk to experts in the industry about the most pressing issues and topics, where we go about over strategies, but we also try to get tactical so that you’ll come away from these episodes with advice and things you can actually apply to your accounts.

So one of the burning topics In the industry nowadays is performance max. But we’re not talking about that. We’re going to talk about RSAs, the other burning issue. So RSAs are replacing ETAs. When you watch this episode, when we’re launching it, you have something like two days left of ETAs. If you’re watching the episode in replay afterwards, ETAs may already be gone.

So we got some great guests joining us today to talk about the end of ETAs and the start of an RSA ad world.

Ah, yes, rest in peace, ETAs, and to discuss that topic, we have two fantastic guests. Let’s start with the first timer on PPC Town Hall, Mark Irvine, we finally got you to come on the show.

Mark Irvine: I’m sure you thought you could avoid me for a while longer, but Fred, it’s so good to be here.

Frederick Vallaeys: Yeah. So Mark, tell us a bit about yourself.

You’ve been in the industry for quite a while, right? So you.

Mark Irvine: Yeah, as youthful as this face looks I’ve got some 12 years behind me on this. I’m currently the director of paid media at Search Lab Digital. But I’ve got a long history of being in the industry, speaking at a number of events. I know both of, You guys from following you online being at events with you guys for a long time So I feel like this is long overdue, but i’ve had the pleasure of speaking alongside you guys at events in the past Being on similar webinars with you.

Previous history includes word stream And then of course i’ve been on a list with both of you as a top PPC influential Influencer for a number of years in the past. So I’m really thrilled to be here.

Frederick Vallaeys: Yeah. So Mark is being modest. He was actually number one on that list. I don’t know what ratings Julie and I scored that year, but Mark was the number one.

So you got to listen to him. And I did have the pleasure of actually seeing you in person again, a couple of months ago in LA at the B2B conference. So. That had been quite a while, but it’s always nice to see people in person, but these are fun too. When we do it virtually or Julie this has been, I don’t know how many times for you on this show, but the host of PPC chat.

So we know you’re super busy. Tell us a bit about yourself for people who may not have watched you before on this, on these episodes.

Julie Bacchini: Sure. I have been in, you know, PPC since literally the beginning. So every time I say that it makes me feel a little bit old, but it also is true. So I was thinking as, as we were, you know, getting ready for this, that I remember the, I remember when ETAs were you know, introduce and what the big, you know, earth shattering time that was.

So I’ve been doing this for a really long time and I look forward to sharing sort of that longevity, you know, perspective.

Frederick Vallaeys: Are you also, you’re not the mother of PPC? PPC mom. I’m PPC

Julie Bacchini: mom.

Frederick Vallaeys: It’s

Julie Bacchini: true. When you, when you need a little encouragement, when, you know, you’re feeling like. Things are difficult.

I just want everybody to be happy and live their best lives. So, you know, if you need like a little PPC mom, my DMs are open on Twitter, you can always reach out.

Frederick Vallaeys: Fantastic. Thank you, Julie, for being PPC mom. All right, let’s jump into the topic here. And lots of people may need PPC mom, but again, just to set the stage here really quickly.

Julie’s gone through all of these changes, but Google at different points in time has had different text ad formats. The one that we’re currently working with is called ETA, Expanded Text Ad. Microsoft, I think, calls them Extads. The, those are being replaced now by Responsive Search Ads, RSA for short.

And RSA is an ad format where instead of fully qualifying the ad and saying here’s the headlines and the descriptions that need to show together You’re giving google 15 up to 15 different variations of the headline Up to four different variations of the description And then on the fly the system will put it together as it sees fit to try to provide the highest quality ad for all those unique queries Because remember what?

15 percent of queries are always unique to Google, right? That’s their favorite stat, but so show the best ad for them or those scenarios. So that’s, what’s happening here. This is my first question for both of you. And maybe I’ll start with you, Julie, but if you’re watching this live, the live premiere, you still have two days to make ETAs.

Is that something that you would recommend still doing, or are we just moving on completely to RSAs at this point?

Julie Bacchini: As sad as it makes me to say this, I think. Okay. You probably just need to be moving on to RSAs. I feel like what’s going to happen. I’ve already seen this in my own accounts. I’m sure Fred, you have such huge data set, you know, much more than any of us individual practitioners do.

But even in what I’ve seen in my own accounts, even before ETAs are being, you know, sunsetted or retired or whatever euphemism we want to use. They’ve been getting less and less impression share, you know, the percent that they’re served. So I have a couple of, like, really strong performing ETAs that, that still get a decent amount.

They’re served a decent amount, but I would say if I was going to generalize, By and large, RSAs are already being served in favor of, you know, over the ETAs. So, can you take time and create more? Yeah, you can. I’m not sure that that’s the best use of your time though, at this, you know, at this juncture with where, where things are going.

Frederick Vallaeys: Yeah, that makes a lot of sense. Mark, what about you?

Mark Irvine: Yeah, I, I’d agree with that. I mean, I think that one of the biggest challenges that Google has always had with ETAs, standard text ads, all of them is that people had been slow to create many of them. And that’s kind of what gave us the advent. of RSAs from Google’s perspective anyways.

So if you’re kind of sitting here, you’ve known about this for, you’ve known about this migration for a long time, and you already have two days left to create ETAs, nothing’s stopping you from taking that initiative now, but I feel like you’re best just leapfrogging and getting into this. I have heard of some people who are out there Writing in their Black Friday, their Christmas, their holiday ETAs just to have them in there so they can turn them on later, but that seems like effort that could be better utilized here in July.

Frederick Vallaeys: Or you could use a really nice software like Optmyzr. And automatically generate the the Black Friday ads for the next 30 years and automatically put into different years of each of them. Yeah, but that’s a little crazy, right? Who knows what we’ll be doing in 30 years. Julie, you I will

Julie Bacchini: be retired.

I can tell you that right now.

Frederick Vallaeys: PPC grandma.

Julie Bacchini: I might I mean, you know, it’s possible. I don’t know. But yeah,

Mark Irvine: we’re making big assumptions about there being 30 more years, too. But that’s a different conversation.

Julie Bacchini: That is true.

Frederick Vallaeys: That is true. Yes, at times we live in Julie referenced. The fact that RSAs already take so much more impression share, so we did, in fact, run a study and these are some slides that I presented at SMX advanced.

So the branding is on there for that, but yeah, basically, whether it’s due to the quality score components or whether it’s due to Google, just simply favoring the RSA ad format. We did look at ad groups. This is across 1. 7 million ads, 13, 000 plus accounts. And what we saw is that ad groups that contain RSAs have around 2.

1 times as many impressions as ad groups that do not. So a very significant reason to to really look at RSAs as the future. Right now. So today I think we’re not talking so much anymore about is it ETA versus RSA? Which one should we, should we be doing? I mean, RSA is the only way forward, but you know, this kind of tells you the story that even if you still have the ETAs, they may just not help you all that much. But now, do you think there’s still a role for ETAs, legacy ETAs to keep playing? I mean, some people have them around, right? You can still pause them, unpause them. Is that something that you guys have thought about strategies around? Or are they just kind of going to sit in the account as a failsafe?

Mark, why don’t you start? Yeah,

Mark Irvine: I’ll start here. So it was one of those conversations when RSA first rolled out. Google, in fact, made that recommendation that you should test your RSAs alongside your ETAs. And in my testing over several years of this I’ve seen that particularly when that RSA starts out.

This is a machine learning based system, and machine learning starts out random. Having that ETA that a marketer wrote gives that RSA a better testing advantage point than testing against random. And so I still see that, hey, if you’ve got ETAs that ETA doesn’t become worthless. It still actually can help be the training wheels for your RSAs in your account particularly as you’re launching new ad groups, particularly as you’re building new things.

I think that the RSAs you have in your account, if they’re performing well still belong in there until like RSAs truly are outperforming them in every way.

Frederick Vallaeys: Yeah. Julie, any thoughts on that?

Julie Bacchini: Yeah, I mean, I feel like it’s And again, your mileage may vary as far as what’s going on inside of your account, but I have some accounts where the ETAs are consistently outperforming the RSAs, so I’m continually trying to, like, improve the, the RSAs that are in those, those ad groups, because I know that at some point, Google being Google, if, if history is any indicator in how they have Treated things that they’ve deprecated, you know, there’s going to come a time where the ETA is like, technically, you know, they can still serve, but I believe that they’re going to just slowly, even ones that are, that are performing well, even from, you know, numbers that Google would be looking at.

I think they’re still going to be pushing traffic to the RSA. So I wouldn’t turn any strong performers off, you know, just because we’re hitting that point where you can’t make any new ones, but I think, you know, You should have in the back of your mind that there is probably going to be a time where.

Even a one that’s performing strongly now probably isn’t going to, and then your only option is going to be to replace it with an RSA. So you should be to Mark’s point about the learning, you know, the machines we’re not used to Google ads need and learning periods to like, this is a whole other level and a whole new.

Dynamic that we’re not used to on the search side of things. Like social has been like that always, but search now has this whole learning period. So I think. You know, giving yourself, if you have strong performing ETAs, you can give yourself a little bit of a buffer for those RSAs to kind of find their footing.

And you can have those ETAs still, you know, still performing for as long as they’re, they’re going to get reasonable, you know, percentage served.

Frederick Vallaeys: So the advice is do not remove them, do not delete them, because they may serve a purpose at some point in the future, especially those really strong performing ones.

We do have some Interesting statistics on this as well, right? So one concept that people have been talking about is the fake E. T. A. Which is basically using an R. S. A. With every position pinning exactly one text so that you’re recreating your strongest E. T. A. S. And so what we’ve seen is that advertisers who spent many, many years optimizing expanded text stats, they probably have something that’s going to do better for a while than what the RSA could do.

And so for those advertisers, there’s and you can see this in the numbers, right, but the CTR, is if you pin all ad positions, that’s what the yellow is showing you, then the CTR is actually higher than if you pin no positions, which is the blue, or if you only pin some positions, which is the red. And the same thing for conversion rate.

If you have really good performing ads that you’ve optimized over the years, pinning some things to those positions that you know are going to perform better are going to boost your conversion rate. Right. So that’s one point D, but, but it kind of goes into the point here, and this is I don’t know where, okay, so I don’t know where that slide is, but we had a different slide that made the point and let me remove this, that made the point that E, so you look at CTR and conversion rate as you are deciding factors of better performance, but because of all these additional impressions that go through RSAs that whole point could be moot, right?

I mean, you could have a better. conversion rate. But if you get fewer impressions towards those ads, then you still lose out on the net new conversions. And that’s really important to keep in mind. Any thoughts on that? Agreed. All right, we’ll jump to the next topic. So the next question then, now that we’re moving on from ETAs, What’s the right number of RSAs to have in an ad group?

Mark, we’ll start with you.

Mark Irvine: Yeah, I think that a little bit of this is how are you treating those RSAs? If you’re treating that RSA as an ETA, then I think that you can have all three. So you can have up to three active RSAs in an ad group. If you’re treating one of those RSAs as an ETA, Then you can be a little bit more liberal with how many of these RSAs you’re playing with.

You could still in fact have three RSAs and two of them effectively be an ETA, right? What I’ve seen people who have gotten a little bit beyond just that ETA testing do is they’re testing, I really like the idea of A B testing two RSAs in that group. With the idea that You have one that’s kind of your safe ETA.

You probably have very few assets. You’re speaking to your large audience. This is your best assets from your ETAs writing very safe RSAs. And then you have a little bit more crazy of an RSA that’s testing against it. That’s speaking more towards your fringe audience. You have a whole lot more creativity in your headlines, descriptions.

Knowing the fact that it’s going to, that learning on that is going to be very sporadic. It’s going to take a long learning period. It’s going to go write some crazy ads that you, a human being, wouldn’t write versus the other safe one might do. It performed better on. I’ve really liked that from a testing standpoint because you’re basically using machine learning, do the kinds of learning that you wouldn’t have done with ETAs.

And so that’s kind of how I’ve been approaching it. So I, I’m an advocate for that two RSA kind of model.

Frederick Vallaeys: Yeah. Two RSAs. That’s a lot of what I’ve heard as well. Although Google recently says a little bit more one RSA. Sometimes two RSAs, Julie, what are your thoughts?

Julie Bacchini: I agree with what Mark said and I’ll, I’ll build off of that and say, I have been testing cause I’m really interested in the whole pin versus unpinned performance.

And so it depends on your situation, you know, some, some brands you have to, like, you have to pin stuff from a, just from a compliance standpoint. There’s, there’s no choice. There are a lot of industries where you. don’t have to do that. And so you can also play around a little bit and have one RSA that is Pinned with, you know, you could do some pinning or you could pin most things, right?

And then you could test that against an RSA that you just like, let the machine do its thing. And see what ends up, what ends up happening. It’s kind of fascinating to, to do it and see what the machine, what the machine chooses and how Which one gets served more it’s, it gives you, I feel like a little bit of a window, as much of a window as we get anyway, into how, how the machine thinks about things and how it prioritizes things.

So in this coming, you know, being old school and being used to, you know, like designing my own tests for how I want to test stuff. And now really not being able to do that in the same way at all. I’m trying to find little ways where I can gain insight that might give me, you know, a little bit of an idea of like, Oh, this seems to work better than that.

And then is that going to hold true across accounts across industries? So I think trying to think about your testing in a different way and what learnings can you derive from one account that you might be able to then pull out and be like, huh, that was pretty interesting. I wonder if the same thing would happen in this account, which is in a totally different industry.

Cause I. The different things happen. It’s a lot more different in that way, I think, now than it has been historically when you’ve been testing things.

Frederick Vallaeys: And I find that fascinating. So tell us a little bit more about how exactly you go, but like, what’s, what’s the scenario where you have this learning and then apply it across many accounts, many campaigns.

Did you even like measure things across campaigns or would you tend to find insights from single ad groups? And then how do you do all of that?

Julie Bacchini: I mean, I feel like for me, the kind of stuff that I’m looking at when I’m doing what I described would be different types of headlines, different types of, so it, I think we’re talking sort of about like thematically.

So when you’re writing different types of headlines, you know, you have, you might be writing something that’s very brand heavy, right? So you might have a couple headlines that are very brand heavy, and you might have a couple that are very like pain point focused. You know, and then you have something else that’s like talking about a, a, a prime benefit of, you know, what it is that you’re, you’re trying to sell.

And then you can write the descriptions differently too, so that you’re focusing on different aspects, you know, are you speaking more to like the, the pain points, are you speaking more to like, here’s what we do and why we’re awesome. And then, you know, again, seeing the differential and how, what the machine chooses and then how you’re, you know, cause you’ve got your best guess, right?

Cause you’ve had your experience with ETAs up until this point, where you would have put those things together and you’d be like, all right, well, what if we test this offer and what if we put that together with this description? Cause this description seems to resonate, you know, and, and, and perform better.

But what if we then match it with this other component that, you know, was doing well in another ad? And I think you can still do that inside of RSAs, but it feels really different and you’re not doing it as definitively maybe, or as directly as you were able to do with an ETA. But I think you can still play with those things and you can still have different versions going against each other.

And then you can try to see if you can find a trend, you know, where it’s like, Oh, well, when these two get put together, it seems to perform better, but you need longer time horizons for all of this, because when you were serving an ETA, maybe you had two ads running against each other. You know, in two ETAs running against each other, you’re the time horizon that you would need to be able to, to make some of those conclusions or say, Hmm, I think I’m noticing something here.

I want to dial in on that. And I want to test, you know, this piece a little bit more specifically. It takes longer, I think, just because of the nature of the multiple pieces being able to be reshuffled in the RSAs to get to a point where you might feel like, Hey, I might be onto something here. I think I’m noticing a trend like this one, this.

Particular description, no matter what it’s paired with. Seems to really be strong, you know, for just for, you know, for, for one example. So I think it’s like, everything is, I would say everything’s out the window from how we used to do things, but a lot is out the window or it’s, it’s on its ear right now because it’s just, it’s not the same, but you can find the same stuff.

You just have to figure out how to. Do it and be patient enough and get your stakeholders to be patient.

Mark Irvine: And I mean, like with that, it’s not just about time, but it’s also about like volume too. Like obviously someone who is running a national or international campaign with thousands, hundreds of thousands of impressions will get that data get lots of data about all of their different variations versus your local S and B might have an ad group that, that You’ll be lucky if you get a couple hundred impressions over the course of a month, and you likely won’t be able to test the same level of assets, the same level of variation in that SMB account as you do in Optmyzr’s own advertising account, because everyone’s out there searching for, searching for that.

You’ve got much better room for optimization there than a lot of super small businesses do.

Frederick Vallaeys: Yeah. So that’s interesting. And one thing we were showing on the screen here for people listening to the podcast, but what Julie was alluding to was when you do this ad testing, you can basically take multiple RSAs and you can build them around themes.

We also see some advertisers who have their evergreen RSA and then they have a promotional RSA, but you can think of that promotional one as just a different theme, right? Is is that promotion as a team going to perform better than the other one? And it sounds like and mark is saying too like these are just longer time frame tests You just need to have high volume to do this now.

There’s a different way to do Had testing in an RSA world. And we should talk about this a bit as well, right? But let’s talk about ad variants and where those fit into your strategy. But because I think that might actually be one of the solution towards doing a quicker test of a smaller portion where you don’t test all 15 headlines, but you just say, well, let’s test this variation of the headline against that variation of the headline.

Mark, any thoughts on ad variants and how you, you all use it?

Mark Irvine: Yeah, so like that’s the whole thing when it comes to RSAs. And particularly when you’re talking to your stakeholders, they hear that like, oh, up to 15 headlines, up to four descriptions, and left to their own devices, what are marketers going to do?

They’re going to create 15 headlines and four descriptions, right? That’s how Google has told them to do it. Take that. You do the combinatorians, the permutations, the combinatorics of it. That’s over 40, 000 different variations. I don’t know who has time to go through and test 40, 000 variations. It’s probably honestly a disservice to provide that for most businesses.

As you do the math of it and you break that down, if you gave it 10 headlines instead of 15 headlines, it sounds like you’re missing out on a third. That actually brings the total number of variations down to 10, 000. So you actually cut it down by 70 percent. If you bring it down to 7, you bring that down to about 6, 000 different variations.

And so the long and short of this is like, as you’re getting started and you might be sensitive towards that learning period, you might be sensitive towards Google being able to optimize this on your own behalf. You’ll be sensitive to all the different bad variations Google will create. My genuine recommendation for small immune businesses or for people who are hesitant to start this out is actually start smaller.

Look at six headlines, look at seven headlines, test those out, get, build confidence, have the machine learn, and then add in that eighth, add in that ninth. Slowly scale out your RSAs rather than going the deep end.

Frederick Vallaeys: Interesting. And the way you work with numbers, I have a feeling you did that in a previous career at WordStream.

You were, what, director of Yeah,

Mark Irvine: I was a senior data scientist over at WordStream for a long time. So it’s one of those things like, hey, I always made a joke that marketers took statistics 101 their freshman year in college and then never looked back at the math of it. And so as they, as we now embrace like a machine learning.

Riven marketing campaign because you’re seeing this not on Google, not in Bing, but like across the, across the spectrum here, they’re pushing this a whole lot. And now this platform that was written by engineers for engineers by engineers translated out to marketers, Google’s trying to be very. Simplistic in how it explains sophisticated, sophisticated data scientists out to a marketer for the 30 minutes that they’re going to be in Google ads.

So what they do is they say like, Hey, put a lot of descriptions in here, put a lot of headlines in here, make them varied. And they give very big rule of thumbs to make this digestible to people because they would rather you put in 15 than put in. Three. That’s the they know that given enough time, Google will get enough to test 15 get able to test 40 40, 000 variables here.

But you as a marketer, if you’re listening to this call, obviously are a little bit more involved. You’re probably better off testing out small and growing that intentionally in time. So there’s still going to be a little bit of restraint that marketers are going to want to have here as they build out their RSS.

Frederick Vallaeys: Yeah, so I have a restraint and then even with 10, 000 variations. I mean, honestly, the data Google gives us in the combinations report is not good. I mean, you basically get impression levels. So, you know, okay, this ad got 30 percent of my impressions. And then these 9, 000 basically served less than a percent.

That’s not helpful, right? I mean, if you don’t care about impressions, we care about conversions and that number is just not there. And that’s actually where I think ad variations can be a bit helpful because it at least on the statistical. I suppose a statistically significant basis tells you what is better and what’s worse.

But I’d love to hear from Julie too, what she thinks about hat testing.

Julie Bacchini: I have two, I have two thoughts on this. One, I just want to say, because we’re sort of implying this, but we should say it specifically. And that is Google’s going to pressure you when you’re in there building ads that, you know, more, more, more, more, more, more, more you don’t have to buy into that pressure.

And it does it not just by suggesting that that’s what you should do, but it tells you your ad strength is poor. It tells you like, it’s, it’s, it’s, There’s a lot of pressure inside of Google when you’re, Google Ads, when you’re building these ads to not do exactly what you guys were just talking about doing and, and, and finding your own scale for what works and what makes sense for you.

And I think the point that Mark made about volume is worth repeating because it’s incredibly important. And, We are consistently being told by Google that this is a one size fits all and like everything will, will be applied and work equally well across all sizes of business, all industries and everything.

And that is a bunch of crap. So it’s just not, you know, they, of course they’re putting it out that, that way, but it, it’s not. And so I think. Realizing that and knowing that you know, your job as a PPC professional is to help to be able to contextualize for your stakeholders where they fall in that continuum and what pieces of this sort of standardized advice that Google ads is giving relative to RSAs might be effective for them.

And which pieces it’s like, you know, that’s probably not going to work as well for us as it is in, in the aggregate that Google is, is referring to. So I think part of where we add value in this process as things get more and more automated is being able to. Help put each advertiser that we’re working with, put them in the right bucket and do it at the right pace and all of that.

Because, you know, if you just listen to Google, everybody should be on, on the interstate, you know, going at the same pace. And I just, I don’t think that it works equally well in that way for, for all sizes of advertisers, for all industries, you know, for lead gen versus e comm and all that. There’s so many places where the, the road should really split off and Google doesn’t present it in that way at

Frederick Vallaeys: all.

That’s a great point. And then, so even in terms of like splitting off in the road signs that Google gives you, ad strength is a big one, right? But let’s talk about that for a minute. So, so the findings that we have, and we talked to a lot of people and even Google says this, but that’s strength. Is an indicator based off of the history of all the ads they’ve seen in the past.

It tells you, Oh, we don’t have enough variety or you don’t have enough headlines because we generally see that 15 headlines gives you more incremental conversions and we’ve measured that to be true, right? So, but this is what the machine learning system is meant to do. It says compared to the average of every single advertiser we’ve seen over the past year or two, that’s a lot of advertisers, but based on that.

We think you should be doing X, right? But, but like Julie saying, well, no, you’re your own business. You, if you’re Nike and you say, Hey, we should try this new slogan. Just do it. Well, the machine learning system from Google would say that’s dumb. That’s going to have bad ad strength, right? Because it’s never seen anyone do that.

But that doesn’t mean it’s a bad marketing decision. And so I think ad strength, if you’re not sophisticated, it’s a great guideline. It kind of gets you within the lanes of what Google expects most people to do. But if you’re sophisticated and you listen to this episode. You think about marketing quite a bit, then use that with a huge grain of salt.

I mean, do what you think is the right thing and then use actual metrics to decide if that’s working or not. Ad strength, by the way, if you put in the greatest ad campaign ever, it’ll not change its opinion on what’s actually happening from the data. It’s just looking back at history. Anyway, Mark, thoughts on that?

Mark Irvine: Yeah, I mean, like, part of this is, like, how does Google communicate a sophisticated, subjective notion of are people going to respond to this ad based off of a checklist? AdStrength really is just Google going through and checking does Your combination of ads hit some kind of criteria. It wants to Google wants to get people out of the box of like writing one ad with one head with three headlines and two descriptions.

So it wants to make sure that everyone’s like embracing this and embracing some of its best practices. But we all know that best practices are often meant to be broken. You’re going to get a good ad strength. If you put your keywords you put DKI into your headlines here. Is that going to be the best ad that you’ve ever read in your life now?

And that’s, that’s just kind of what this is, is it’s a checklist that Google’s, Google’s created a guideline for people who are new to this to get started. But that doesn’t mean

Frederick Vallaeys: that we had a former colleague of yours, Neva Hopkins. She, she gave us the screenshot, but she did an experiment and she was like, well, I’m going to add all 15 headlines and I’m going to make them pretty unique and varied.

And and the description, same thing, but I’m not going to say anything. There’s no keywords being used in these ad texts. And the only issue the machine had with that was the lack of the keywords, but it was like, oh yeah, 15 random headlines. Fantastic. They’re random. They’re not similar to each other.

So check box. Right. But that’s kind of the problem is it’s kind of a dumb system and you know, you, you, you can all do better than that, Julie, I bet you don’t look at a, at strength too much.

Julie Bacchini: No. And I think that the part that, that irritates me is it’s presented as being so much more sophisticated and smarter than it is.

And I think that’s the part that really craw because. We understand that just because Google says, you know, fill in the blank, right? It doesn’t mean it’s true. It doesn’t mean it’s exactly what you should do, right? Like we have a perspective that allows us to evaluate critically what we’re being told based on what we’re seeing in accounts, based on what we, you know, what our experience is and, and that type of thing.

I think the part that is troubling is that If you don’t have somebody on your team who, or who’s working on your account, who has that level of experience or sophistication, and you were to just go by the documentation and the information that’s shared, you know, from Google and then these prompts and things that you get inside of inside of the UI and frankly, inside of ad Google ads, editor, editor is.

Really harsh. It tells you you have violations when all you’re doing is doing stuff that’s against, you know, what Google’s best practice is. So I think there’s a lot that goes on inside of the Google ads ecosystem that if you were not, if you don’t have the knowledge base, you could be steered in a direction where what you’re doing is good for Google, but it may not be that great for, for the advertiser.

Mark Irvine: It’s really fanatical that over the last couple of years, the Google ads platform has become a sales enablement tool for Google ads more than anything else. It, I of course remember that used to be the case with like your certification exams with some pieces of Google ads in here, but now it’s one of those things like you can’t.

Turn a corner without Google trying to sell you something along the way here.

Frederick Vallaeys: Yeah, everybody wants to make more money, right? So I mean, Hey, I feel it. So it’s a vehicle. Google ads is a vehicle to sell you more stuff. And one of those ways to sell you more stuff, I think, is the automatically generated ad text.

So Google talks about this a bit at GML, how If you don’t have 15 headlines, they would have artificial intelligence systems that will soon be able to tell you, here’s a variation you can use. Now we tested this a little bit as well using OpenAI’s GPT 3 and we found that generally these systems are quite good.

And a couple of years ago, I think I was on stage and I’m sure many other people said it as well, but I said one of the main things that people can still do is In PPC, when the machines handle so much else is to be creative. Well, it turns out machines are actually getting really good at writing a text, even if you give them no other information other than the website itself.

And here you see a couple of examples of tests that we did. How do you both feel about these additional variations that Google is soon going to Potentially auto apply an account.

Julie Bacchini: I think this brings up a lot of questions. This is going to make a lot of people realize that their sites are not talking about what they think they are. I have been recommending for, for a while now that you run DSAs. You know, even if you don’t want to, even if you don’t want that to be a longer term strategy, I highly recommend running some DSAs so that you can gain insight into what Google thinks your site is about. And if you love your SEO team, you can share that with them too.

But I think, I think it’s interesting. I think that. It has the potential to be great if your site is, is well set up and is speaking to the audiences you’re trying to speak to it. Like, I just feel like there’s so much room. There’s so many sites. They’re bad that if, if you’re going to put the, like, not as sophisticated as it’s build machine learning.

Coupled with the AI that’s going to write stuff for you, because we’ve all seen the suggested headlines in there. You know, and then it’s going to be pulling stuff from your site, which it’s like, well, we have thoughts on how that could be better. Like that seems like that’s three places where the wheels could just be like coming off, coming off, coming off.

If everything you have is pretty tight, I can see it doing a good job, but let’s be real. How many advertisers have all of their. Those ducks really in a row so that the machine is pulling from highly organized, optimized, speaking to your audience, you know, content.

Frederick Vallaeys: So we have to have oversight, but we can make great suggestions, but we have to provide the oversight for the machines.

And then honestly, like what you’re saying, if, if. If you find out through this channel that, Hey, your landing page is horrible because the machine can’t figure out what it is you’re trying to sell, that you probably should be spending more time fixing your landing page instead of like dumping more money into Google ads, which is going to, right,

Mark Irvine: PPC is an expensive way to discover that your SEO is also bad.

I would say that like, if I’m trying to do a favor to my SEO team, I would take actually the, I was in the, believe it or not, the keyword planner tool. recently that you can drop your landing page in there and you can get that same kind of feedback of like, how does Google read keywords on that landing page and story might surprise you oftentimes poorly.

And that’s really good feedback for your SEO team, right? Because then they, they understand that, Oh, Google is reading. This is how Google reads your landing page. This is the kind of stuff that pulls from this. If that’s not the keywords that we want, then wow, where you got it. Correct. That for these RSA assets that Google is auto creating and auto applying.

It’s what it’s from other assets within your ad groups. It’s from your landing page and from your keywords, correct? So it’s those are the three places that Google can find this information to create that content. So ideally, it’s content that you’ve already written. And it’s just that Google is applying it to your ads here.

So this isn’t necessarily like Google going Terminator and creating its own ads that are going to be completely unrelated to what it is that you’re doing. But if a single variable within that is bad, then it is, of course, bad. Cautious that we would want to play here. And of course, Julie raises a good point that as you’re chatting about compliance and legal teams and branding teams, if you’ve ever had that conversation with the client about compliance, then most likely this is going to be a life.

That’s a lot easier if you just opt out of.

Frederick Vallaeys: Exactly pros and cons to this whole thing, obviously what I do want to show here on the slide. So if you focus on the left side, so we asked the machine give me three variations of the following sentence that we put in get started today. No credit card required.

So it’s a tagline for Optmyzr or one of the headlines. And it came back and it said start now, no credit card required. Get started now. No need for a credit card. And these are very subtle variations. As I I’m curious, like, is this even worth testing? Cause, cause one of the ways that I think about it is now that you have had variations, you can literally say, this is the one headline that we’d written, but what if we said it slightly differently, like might it come across more friendly?

More business like, could that actually have an impact on the ad? It is quite helpful in some ways to get the machine to just sort of spin the text as opposed to come up with a whole new thing. As opposed to you having to come up with a whole new thing.

Mark Irvine: Yeah, I want to say, so remember years ago, Google had that first announced, even with ETAs, they would automatically create that ETA.

For you, right? And what I saw from that was, I think that when that happened, most of the time, it was an unobjectionable, slight variation of your ad. And all in all, advertisers benefited some 5, 10 percent of click through rate boosts from that. But what took the story? In the industry was the examples where it did not do that, and I’m sure that that is one of those things.

Like, as we as professionals work with our clients, we find that landing page that has content that is irrelevant. It’s always going to your client is certainly going to let you know when they don’t like an ad. And that’s going to really it. Influence our day to day, our decision making in terms of how we react to this stuff.

But for most advertisers, if things are buttoned up then I think this is a positive for them.

Frederick Vallaeys: Yeah. So I guess the cautionary tale is don’t allow auto apply to happen, but use the machine to give you suggestions, validate to make sense and get 10 percent benefit that it could give you. Okay, great. A couple more questions before we wrap up here, but have either of you changed campaign structures now that RSAs are the the ad format?

Julie Bacchini: I haven’t changed a ton in, in structure as far, you know, it’s, it’s more, you know, it’s more just putting them in existing ad groups. I did want to bring up one point as we were, it’s kind of relates to what we’re the, the previous question we were just talking about too. I think one other shift that.

This encompasses is the last few years we have been sort of more manually working with audiences and part of what we’ve done historically in in PPC in creating ad copies specifically and creating you know, our keywords and having that flow through is we’ve tried to speak. the language of the type of audience that we think we’re most likely to reach with particular keywords.

And I think another interesting aspect of the RSAs and all of the automation is that. What Google wants you to do is sort of put in the language that maybe you would have previously separated out into a couple of different ads or a couple of different ad groups. And it wants to have access to all of that.

And it wants to decide what pieces it’s going to put together. And it’s going to show an individual searcher. And I think that’s also a huge change in the way that we think about things and the way that the actual platform itself functions. And I think that piece hasn’t been talked about a ton. It’s sort of like running in the background, you know, of all this other stuff that we’re talking about.

But I think it is pretty foundational to, to what’s happening and how our thinking is going to shift.

Frederick Vallaeys: Is this the depth of the single keyword ad group? Because now with an RSA, I mean, God knows what combination is going to show up. So what’s the point of having one RSA with a specific keyword? If it could be 400, 000, what is it like 40, 000?

I think

Mark Irvine: it’s 43, 160. If I were to guess,

Frederick Vallaeys: I’m sure that was a very good guess. Yeah, but so a shift from SCAG, single keyword ad groups to maybe more stags, which is what most people are now doing anyway. So single theme ad groups. That’s kind of the structural shift that I see continuing as a result of RSAs.

I haven’t seen anything else really based on this. There is also a little bit of course, measurability. So when it comes to ad variations, they are generally cross campaign. You can select them from single campaigns, but there is again that push towards less. Complexity, I guess, in accounts that we we still believe in adding some level of complexity because this is my big beef with Google and I keep saying this again and again.

So I’m sorry for people who heard it. But to Google, any problem is a bit management problem, right? So if I have a key word or an ad that’s not performing well, or they put me on broad match and you find these queries that are kind of like mediocre. While sure, they’ll still give me a few conversions.

They’ll just bid less for those because the conversion rate is going to be lower. But I’d actually really like to know what these queries are that aren’t performing well, or tell me about these geolocations where my ad is getting bid lower because my conversion rate is not so good. If I knew about these things, maybe I can fix the landing page.

I can change the offer. I can figure out based on my business knowledge. That it’s, it’s actually a problem that I could control in some other way, besides just bidding, which is a bit of a crutch in many cases. Right. And so, so, and then, and then when you find those scenarios where, Hey, we should have had a different message in Paris.

Sometimes that does require that you split things out or you use ad customizers or you have a different campaign Yeah,

Mark Irvine: I mean, I haven’t had a whole, I sit in Julie’s camp here. I haven’t had a whole new approach towards structuring ad groups campaigns simply because of the shift in ad type here, really, at the end of the day, we’re still playing that semantic game of user searches for X.

I wanted to give them an ad for X and then bring them to the page. How we go about building that is something that like Google doesn’t clearly doesn’t want us to overdefine that answer that this is not a single keyword ad group kind of realm. This is not even at times a place where we know our audiences, Google kind of wants to push more of a how are your personas thinking about this?

What kinds of offers do different people respond to here? Knowing that not everyone is our mass market audience, but like that Parisian has a different sensibility here. So how do I compromise here with machine learning here? Frederick, I think that what you’re kind of saying is that, well, I’ve already done this testing either in the past, or I’ve done this testing with my other Other campaigns outside of Google ads, or I’ve done market research, and I know this stuff.

How can I feed that to Google? And we’re losing a lot of that control. And so I really think of this as, you know, it’s this fight between machine learning, which machine learning is a huge asset to advertisers is a huge benefit, but by its nature, It’s not supposed to be touched by people. You’re not supposed to come in and put a lot of, like, as you get data from machine learning, you’re supposed to let the machine continue to be random and test against itself.

And so I think that there’s still those places where, where you as a person can feed that prior information into Google, be it with giving it fewer assets, with giving it more restraints. You can, Encourage the machine to learn faster. I

Frederick Vallaeys: mean, those sound like a great final thoughts here. So we’ll do some actual final thoughts.

So you’ve been fantastic guests on the topic of RSAs. Any, any final thing that you think people should do or just how people can follow you and stay in touch and learn more, Julie, why don’t we start with you?

Julie Bacchini: Yeah, I think, you know, this, this train is going to continue to move forward. It’s not going back to the station.

So I think that figuring out. How you can work the best within the system as it exists now and as it continues to evolve for the type of account or accounts that you’re, you’re working in, I think is going to be the biggest thing that PPC professionals are going to have to work out for themselves, you know, in over the next year, right.

Which already feels like a long time horizon with the rate that things have been changing lately. So I think, you know, just having. Having the conviction to try some things and maybe not just follow what Google’s pushing you aggressively to do inside of the The platform, you know, it’s okay, you know, you can take your experience and you can, you can try to have it work with, you know, what, what Google wants you to do to try to find that outcome that works best for the accounts that you’re, that you’re working on.

That would be sort of my overall takeaway from like, how do you, how do you process and think about all this, you know, and, and apply it to, to your everyday and what, what it is you’re trying to do.

Frederick Vallaeys: This is not the end of that conversation. So Where can people continue that conversation?

Julie Bacchini: So Twitter is a great place.

You know, I managed PPC chat. We, the hashtag is active all the time. So anytime you have thoughts, questions, seeing something weird, pop it up on, on the PPC chat hashtag. And we have two regular chats each week. We have a tweeting based chat Tuesdays at 12 Eastern, and then we have an audio version. We’ve had that for just about a year now.

We do a Twitter space at 12 Eastern on Thursdays.

Frederick Vallaeys: Great. And we’re big fans of PPC chat, so definitely tune in for those. They’re very helpful. Mark, what about you? How do people get a hold of you if they want to continue that conversation?

Mark Irvine: Yeah, certainly. You can follow my socials at MarkIrvine89 over on the Twitter.

I, like the machines, am always learning as well. And I am going to just Signal boost everything Julie just said. When I was starting the industry 12 years ago, I was on PPC chat trying to figure it out today as we’re all trying to figure out together. PPC chat is still both through the formal chats as well as ongoing.

It’s always a wealth of information, and you’ll have the same kinds of experts asking the same kinds of questions as we’re all figuring out together here. If I had final thoughts on. RSAs specifically, it would be, hey, definitely keep your stakeholders in mind because we’re all sensitive to the people who are paying our paychecks here and how they’re going to react when they see something that they don’t like.

But the biggest piece that I have is We should prepare to fail fast when it comes to machine learning, and that’s often going to be to give it give Google give Microsoft fewer resources to really mess up and mess around. If you will learn quickly. Yeah, exactly. The the the PG version of that learn, learn quickly with this.

And then as it goes along, As machine learning picks up as it finds those variations within a smaller pool, it will scale out faster.

Frederick Vallaeys: Great. Well, Mark, Julie, you’ve been fantastic. Thanks everyone who’s watched today’s episode. If you enjoy these PPC Town Halls, be sure to subscribe on YouTube. These are also available as a podcast on your favorite podcast platform, so subscribe there as well.

Listen to them on the train, in the car, on the beach. If you’re on summer vacation, by the way, my new book is just available on the audio book. So it’s not a good one to pick up for those who like to listen. And then of course, if you want to manage your accounts a little bit more easily, Optmyzrs, PPC management software available for free for a two week trial.

So if you haven’t tried that, give it a shot. We’ll see how we can help you get the best out of your accounts, but thanks again for watching everyone. And we’ll be back in a few more weeks with a new episode. Take care.