What’s Working in Performance Max, First-Party Data, and Generative AI in 2024

Episode Description

2024 has brought in a new era for marketing: thriving in a privacy-focused environment without sacrificing success.

While that’s on one side, there are continuing developments in ad automation and generative AI on the other side.

Considering both of them, here’s a question: has PPC become tougher or easier?

That’s an interesting discussion. But it’s clear that how we do PPC has surely changed. What worked before hasn’t been working today.

Hear from Mike Ryan, Head of Ecommerce Insights at Smarter Ecommerce in this episode of PPC Town Hall where he touches on all of those topics.

Mike shared some really amazing insights into what he has seen work and not work in the past few years in Performance Max, generative AI, and first-party data. He also discussed a powerful script he built that creates a timeline of Performance Max brand traffic.

We have a lot in store for you in this episode.

Tune in to learn:

- How are EU advertisers preparing for a privacy-driven future

- What happens when Performance Max and Shopping campaigns run in parallel

- How to capitalize on first-party data

- How generative AI has impacted PPC in the last few years

and more

Episode Resources

Check out Mike’s PMax script: https://smarter-ecommerce.com/en/goog…

Episode Takeaways

EU Advertisers and Privacy Regulations:

• The Digital Markets Act requires tech companies to implement Consent Mode v2, affecting data collection and user tracking.

• Non-compliance may disrupt remarketing and conversion tracking, prompting agencies to support clients in meeting regulatory requirements.

   

  Performance Max (PMAX) and Shopping Campaigns Together:

• PMAX can cannibalize brand traffic, impacting the performance of standard Shopping campaigns.

• Standard Shopping campaigns should act as a fallback to capture missed traffic and complement PMAX’s lower-funnel focus.

   

  Leveraging First-Party Data:

• First-party data is more than just lists; proprietary data, particularly at the product level, should guide campaigns.

• Structuring accounts around first-party data helps differentiate brands and ensures automated campaigns align with business goals.

   

  Generative AI’s Impact on PPC:

• Generative AI tools like ChatGPT streamline research, creative tasks, and scripting.

• Tailored GPTs improve campaign alignment with brand goals and deliver better recommendations.

Additional Takeaways:

• Balancing Automation with Insights:
• Google’s new controls on search terms and negative keywords need to be understood and used to refine campaigns.
• Diagnostic tools offer insights into brand vs. non-brand traffic and data trends.
• Future of PPC Automation:
• Search themes and exclusions may replace traditional keywords as PPC campaigns become more automated.
• Unofficial Google GitHub products provide innovative ways to differentiate in PPC management.

Episode Transcript

MIKE RYAN: I just think first party data is so much more than audiences. It’s so much more than just a remarketing list. And it bugs me that first party data has become just synonymous with basically a remarketing list or email lists and stuff like that. When actually your business, you have all kinds of proprietary data and in e commerce, particularly at the product level, which you can then feed into the platform, a common hack that you see out there or Is like this idea, it gets it back to brand and non brand that you can that you can apply brand exclusions or brand negative keywords.

If you know the right way to, that you can do this in a performance max campaign. But the problem is that it will nuke the brand traffic in the shopping part as well. And so then you would create like a standard shopping campaign that would be a fallback.

FREDERICK VALLAEYS: Hello and welcome to another episode of PPC Town Hall.

My name is Fred Vallaeys, I’m your host, and I’m also CEO and co founder at Optmyzr, a PPC management tool. Today we’re bringing back one of our favorite guests, someone who’s very active in the PPC space, shares a lot of great Deep insights lots of data and the reason he can do it is because he works for another great Agency tool vendor who work a lot in the shopping ad space smarter e commerce.

So my guest today is Mike Ryan and we’re going to ask him some great questions about what’s going on in europe What’s going on with pmax? What’s going on with first party third party privacy? Much more so stay tuned and let’s get rolling with another episode of ppctownhall. All right, mike. Welcome back to the show Great to see you again

MIKE RYAN: Yeah, absolutely fred.

Thanks for having me back. And I know you were you were in austria not that long ago actually Yeah, it’s starting to be a while ago. We and we missed we missed each other, but so it’s good to catch up again

FREDERICK VALLAEYS: Yeah, exactly. I was there in summer. So yeah, my wife is austrian as as is your wife, right?

MIKE RYAN: Yes, yes,

FREDERICK VALLAEYS: funny thing to have in common,

MIKE RYAN: right?

FREDERICK VALLAEYS: Exactly. Funny thing. But I guess she pulled you to Austria. I pulled mine to the United States, even though I’m originally Belgian, but yeah. So but yeah, great to have you back on the show and you know, Do a little bit of a better job than I did explaining what Smarter E Commerce is, because I, I, I think I didn’t do it full justice there.

MIKE RYAN: Oh, no worries at all. So Smarter E Commerce also known as Smec. We’ve been around for, we celebrated our Sweet 16 recently, so we’ve been around for quite a while, and thanks. And yeah, we have this pure E Commerce focus. We used to be in some other kind of verticals as well, but we’ve really, Specialized in e commerce.

So, you know, you, you said like agency tool and that actually, we kind of are at the intersection of those. Cause we are a solution provider. We have, we’ve been developing software for years. But we also put a lot of professional services and consulting on top of that. So we kind of call that like SAS plus like the software as a service, plus this human expertise, which we find is just, Actually increasing in importance all the time.

And you know, we’re quite busy with performance max these days. So we have a software called enhanced performance max. Yeah. And that’s. Short intro to the company.

FREDERICK VALLAEYS: And then let’s talk a little bit about what’s top of mind for you in PPC and for smarter e commerce. So you mentioned there’s a new PMAX tool, but what kind of interesting data are you seeing these days?

And for people who don’t follow Mike on LinkedIn, Do make sure you subscribe to him or follow him because he just put out quite a bit of really good research from the company. It gets, it gets complicated. There’s always a lot of good discussion around the data, but like, what’s one of the ones you’re working on right now?

MIKE RYAN: So right now there’s a couple of things we’ve got a script coming out soon. Actually, I’m not sure when this will air, so maybe it’s out by then. But it’s looking at this brand non brand topic and performance max and helping to visualize that as a time series. So I think that’s going to be I hope people will like that.

It’s been really popular with our teams internally. We’ve also done some kind of pretty nerdy stuff around historicizing or storing changelog data to better understand, What’s occurring underneath the hood with performance max, because like there are data points that are hard to work with like related to budget, for example, or changes to return on ad spend targets.

And so, like, that’s something that we published late last year. We were showing basically how how actual return on ad spends. How it changes based on different changes to the target return on ad spend. And that’s actually just one data point in a, in a much larger set that we built.

FREDERICK VALLAEYS: So, so that’s really fascinating, right? Because I think one of the bigger questions that I keep hearing and that there’s not a really solid answer for is how long does the system, Google’s machine learning system, have to learn after you make a change? And so sometimes I hear Google say, Don’t make any change of any kind because it kicks the system back into learning.

And so then you’re going to have to wait two weeks. And so even something like a budget change, a change to the target, like do it very sparingly, but then there’s the flip side, which says like, listen, if you need to have a different target return on ad spend because something in your business is shifting, or you need a different budget because you have a clearance sale or some other seasonal event, like you can’t make the machine learning force you to not make the right business decisions.

Right. And my take is often that. So when you do make a change to the target return on ad spend, it generally opens up new pockets of traffic that may not have been part of the mix before. And that’s the stuff that the machine may have to learn about. But anything that you’ve historically already advertised on those search queries, like it’s not like the machine is dumb and all of a sudden forgot everything about that.

Right. So, but what’s your take on like the how active should you be in an account without hurting it?

MIKE RYAN: Yeah it’s a great question and I think you described it well. I definitely see it the same way. You know, you might start participating in different kinds of auctions and reaching different kinds of users as well, depending on these changes.

And you know, it’s a lot more dynamic In the past, there would be a more kind of linear relationship between a change in your, your efficiency target and then like your cost per click, for example. But now it’s much more fluid in something like performance max because this can impact more like the potential inventory mix, the placement mix what kind of audiences you’re reaching and so on.

But we, we do find that it’s, it’s a pretty robust system that Google has built there. And, and typically You know, I think frequency of change is another topic, but magnitude of change. What we see is that there can be some slippage when you’re making really big changes at once. You know, the algorithm is going to follow, but I think it’s good to break those into more stepwise changes.

Yeah. What would

FREDERICK VALLAEYS: be some, are you talking about like a big jump in a target, a big jump in Budget, or are you talking about big changes? Like, Hey, we just overhauled the audiences or we just overhauled the creative components.

MIKE RYAN: Well in the first place, I meant like a big change in, in the targets. Also a big change in budget can be like budget is a, If you don’t have a ROAS target in place, then budget is a very important way of kind of pacing the campaigns.

These are, these two things, they’re basically all about just pacing the campaign. And yeah, of course it’s an awkward position because there are not as many promotional controls left as there used to be or promotional levers. And so when, when budget and return on ad spend are your, they’re like your, your gas pedal basically.

But it’s like, Hey, this is supposed to be a strategic thing. Like this is supposed to be something that’s modeling profit in my business. There’s not a gas pedal. So it, it is a bit, it’s a bit awkward. Yeah.

FREDERICK VALLAEYS: I like that analogy, right? If you think of it as a gas pedal, it’s not like you’re on the I mean, you go from your house to the office.

You got to change speed, right? Whether you’re on the city roads or on the freeway. And I mean, it’d be ridiculous if it was like, Hey, set your one pedal level and go to work at the same speed. I mean, you’d be knocking down children on the way to school and then you’d be like a road boulder on the freeway.

So that doesn’t make sense. And it’s also interesting because I think a lot of advertisers still think of like target ROAS as like the ultimate goal, but it’s about profit, right? Profit or maximizing revenue or whatever matters to the business, like clearing out some inventory that you need to get rid of to make space for the new stuff.

Totally. Hey, but So you kind of alluded to it. So there’s unintended consequences, right? If you make big shifts in budgets or targets. So let’s take an example here. Say that you get really aggressive all of a sudden on target return on ad spend and you dial it way up. So the unintended consequence could be that you’re shifting from.

Kind of a lot of generic queries to maybe more brand focus because brand is going to be cheaper. It’s generally going to perform at a higher ROAS. So by changing the target return on ad spend, you’re basically changing who your ad is shown to. And it sounds like that’s a script that you have. So tell us a bit more about the script and brand versus non brand and, and especially said the, the change over time, like how does that help people?

Manage better accounts.

MIKE RYAN: Yeah. I mean, so the this, the script basically it’s, it’s working with the search term insights report, which is different than your classic search terms report and there’s some things in there that are frustrating, frustratingly missing, like there’s there’s no costs for example.

So then you can’t know the return on ad spend. You can’t. But the big thing here

FREDERICK VALLAEYS: is I think they don’t actually show you search terms. Terms, right? It’s search themes.

MIKE RYAN: No. So there, so basically there are search categories at the top level. And if you just view that report, like in the UI, in the standard view, you’re going to see those, but you can drill these down.

It’s nested. So you can drill that down to search subcategories and you can then drill all the way down to search terms. But you won’t, does that

FREDERICK VALLAEYS: work now in scripts?

MIKE RYAN: It works in the UI, it works in, yeah, it’s available now in API and, and with scripts. Yes. Nice.

FREDERICK VALLAEYS: So they finally rolled that out. Yes.

I know that was a big advertiser request for a while there.

MIKE RYAN: Yep. I, I personally nagged our wonderful ads liaison, Ginny Marvin about a million times. But one of the reasons I wanted to do that is so that I could do exactly this, understand what’s happening with the brand, non brand share because these search categories, it’s.

Sort of a strange phrase for that in a way. It’s their, their Google’s kind of doing Ngram clustering for you, which they’re just finding, you know, they’re, they’re, they’re, they’re clustering together related phrases, related or kind of phrase like variants. Don’t wanna misuse some terms here, but typos are covered too.

But basically you can then you can then look for your, your brand term in that or other terms, but let’s take a look at that use case and you can find out like, okay, what amount of conversion value out of this PMAX campaign is, is, is important. Is coming from brand traffic and what amount is non brand and then, you know, the tricky thing is that the, the report, it has selectable time ranges, but this is always just like a snapshot.

So you can look at a snapshot of the last seven days or a snapshot of the last 30 days. Or however, whatever time range you put in, but it’s not it’s not actually a time series. It’s not, you can’t, there’s no sequential element in it. So that’s what the script does is it stitches those little snapshots together into a time series so that you can see how that’s developing over time.

And this works basically as a diagnostic tool. Similar to like what Mike Rhodes has done with his performance max script that shows your, your placements or estimates your, your placements because Google doesn’t officially report this information. These things to serve as diagnostic tools so you can make a change on your return on ad spend target, for example, and you can see exactly like you mentioned, if you became way more aggressive on your target, did it start drilling into brand traffic, then you would see that branded conversion.

Our share of conversion value or share of conversions rising and you could then diagnostically know that that’s happening or vice versa. Let’s imagine that you, that you said a much less aggressive target is that then going to allow performance max to, to kind of expand its ad inventory more and, and, you know, get into more I don’t know, YouTube or whatever kind of placements, all these, these other Google owned and operated properties.

So these are like, these are just diagnostic tools. It’s, you can sort of only triangulate what’s going on in performance max. Because they’ve, The, the metaphor, I’ve made this before, but it’s like the, it’s like this this, this fable of blind men and an elephant and each one of them is touching a different part of the elephant and completely misunderstands what the animal is.

But if you, if, if these people would speak to each other, then they could kind of size up the whole animal together, if that makes sense. Yeah, it’s, it’s tough to sometimes like grasp exactly what’s happening in those PMAX campaigns. So that’s very cool that you’re building a script and the Mike Rhodes script, anyone who hasn’t checked it out, do look it up.

FREDERICK VALLAEYS: There’s a lot of buzz around that. I think he’s on version like 34 or something.

MIKE RYAN: I know I have 50, 52 drop today. I just saw that. 52

FREDERICK VALLAEYS: drop today. Okay, great. So so he is charging, I think for that script, at least for the advanced capabilities. Are you going to make yours free or how, what, what’s the game plan there?

MIKE RYAN: Yeah, this, this is, this is free. I think there’s just going to be, you know, an email wall or something like that. But Are you a

FREDERICK VALLAEYS: scripts guy? Are you, did you write it yourself or did you use some generative to help you out here?

MIKE RYAN: So I, I don’t, I’m not writing these myself. I work with basically one of our developers.

We’ve been working together for years. Like we used to have this free data visualization tool called Orbiter. And we’ve worked on other fun projects before, but Yeah, I think he might have actually been working with chat GPT a bit. I do, I use chat GPT as well, like for Microsoft Excel formulas and stuff like this, but I’m not scripting.

FREDERICK VALLAEYS: Okay, cool. So the, the other thing that you do. Which is great for the industries. You have your own podcast and I know there was a recent episode, which was really great. You had someone from smarter e commerce come on and talk a little bit about first party data and the deprecation of third party cookies, which always seems to be threatened by Google, but then delay upon delay upon delay.

So we’ll believe it when we see it for real. But, but certainly there’s this push towards privacy, right? Like that’s no. Secret in Europe. You’re in Europe. They are pushing that harder than anywhere else. So I’m always curious for the audiences in the United States and elsewhere. What are you seeing in Europe that we will eventually have to worry about?

MIKE RYAN: Yeah, great question. So for sure, the, the, the European union is very, Arguably kind of protectionist about this topic. I think some of it comes from a sort of sibling rivalry with the U S and I don’t know, but it feels, it feels like there’s more there, but anyway to keep it on topic,

FREDERICK VALLAEYS: that’s a whole different episode, right?

So that’s the

MIKE RYAN: the why of, of all this. But yeah, so basically they, they recently passed some, some new Legislation or regulation. It’s the Digital Markets Act. And what this does, it designates these big tech companies as gatekeepers and which is it’s a special designation and it’s designed that they can be more stringent against these large companies.

Platforms like Google and Meta, Apple, for example well, not punishing every other company out there. So it’s really kind of a targeted measure against these very large companies. And it, it just there’s, there’s several implications of this. There’s going to be changes to the search results here in Europe.

But relevant for this is it mandates the use of consent mode and specifically consent mode version two is what we’re rolling out here. And so this is basically just a way You know, you, these cookie banners, this is the consent topic. You know, people need to opt into tracking and, and opt into personalized advertising and all these other marketing use cases.

And so basically you’ll have a consent management platform that’s kind of handling those banners for you. And this is just a way of allowing your consent management platform to communicate. The status of consented or unconsented to Google, and then they can take different actions from there. Like if the tag is going to fire and stuff like that in Google Tag Manager.

FREDERICK VALLAEYS: So it sounds complicated, Mike. So from an implementation perspective for some of your clients, like are they going through implementation of this? Is this complicated? Or is this basically you go through a couple of Google setup screens and click a few buttons and then you’re done with this?

MIKE RYAN: There’s a bit more, there’s a bit more to it.

I guess I would describe it as, as moderately complex. And you know, that’s, we are, it’s like a, a project scope that we offer to help our clients make sure that they’re, that they’re all set with that. So we are definitely that team is very busy right now with those scopes because I guess what I forgot to mention was that there was a six month deadline on this.

So I think, I believe the day is March 6th, if I, if I remember less than a month away. So everyone is scrambling to make sure that they’re, that everything is squared away because the implications of this is that if you don’t have this implemented, then in the first place like your remarketing will stop working.

But there’s actually a whole suite of potential consequences. And Google is signaling that later in the year, like you’d even have Major problems with your conversion tracking and stuff like that too,

FREDERICK VALLAEYS: right? So measurement is going to break and then some of your first party audiences And that’s interesting too, right?

Because remarketing like that is It’s first party data that you have a relationship with that customer And to be able to act on that you would have to go through consent mode to to make sure that that continues to work now a little side track here, but If remarketing stopped working and you were running PMAX campaigns, maybe we’ll finally see how much money Google is really driving for you, right?

Without just taking your SEO traffic.

MIKE RYAN: Yeah, I, I had the same, the same thought, actually, it was one of it’s one of the first things that crossed my mind and and I’m, I’m, I’m sorry, but I know, I know that

FREDERICK VALLAEYS: all of a sudden, like all of these advertisers are like, wait, what, we get great results without remarketing or like, we’re, we’re, we’re spending this much money on these performance max campaigns.

Like, let’s shut it down. I

MIKE RYAN: guess they’re, they’re, they’re like volume and would collapse or roast. Like they’d be like, it stopped working. And it’s because it was, cause it was propped up by remarketing. No I mean, I’m sure we’re going to have, the reality is we have a lot of clients and I, and I bet despite our best efforts and their best efforts, I bet you someone’s going to miss the deadline and I want to be the first one to hear about it so I can check that.

Let’s go. I’d love

FREDERICK VALLAEYS: to see that data. That could be Hollywood.

MIKE RYAN: So, but in regards to like blending shopping and And PMAX, there are use cases for that. A common hack that you see out there or is like this idea, it gets it back to brand and non brand that you can that you can apply brand exclusions or brand negative keywords.

If you know the right way to do that, you can do this in a performance max campaign, but the problem is that it will nuke the brand traffic. In the shopping part as well. And so then you, you would create like a standard shopping campaign that would be a fallback. But this gets to an important mechanism, which is like standard shopping.

So Pmax takes priority over standard shopping. When the same item is being advertised. But also and you see this with the brand non brand effect, there’s also auctions and queries that PMAX might miss. So I think it’s good advice to have a standard shopping campaign as a fallback in there because it can pick up some traffic that PMAX is missing.

And then. There are more advanced use cases beyond that, but I don’t know how much detail we want to get into it. That’s another episode, maybe. No, I mean, you definitely have good blog posts on that, so check those out. But that’s interesting because in the beginning, I think PMAX was sort of built by Google as this.

FREDERICK VALLAEYS: Hey, you still have your search campaigns or your display campaigns, but set up PMAX because it’s going to be Sort of that incremental volume of the things you might be missing but now you’re saying the world has shifted to PMAX is the standard campaign for many advertisers and You want to set up these traditional campaigns around it for anything where PMAX isn’t playing and give me an example of that So where might be PMAX?

Completely miss it. Well, just a quick, I mean, I think it’s a, it’s a great point. You mentioned that definitely PMAX was pitched as complimentary towards standard campaign types. And the idea was that you know, the one with a better ad rank would serve. Typically this was the, this is the main. Priority mechanism in there.

MIKE RYAN: What we saw over time was that actually PMAX seemed to just be really cannibalizing these campaigns. And I genuinely think Google didn’t think that would happen or believe that would happen at first, but now we’re just seeing like Google display should migrate to PMAX dynamic search and migrate to PMAX.

We already saw some others and yeah, the trend just continues. But examples where PMAX might miss. I mean,

So like a PMAX campaign, for example, is going to serve on, on YouTube and, and on like it’ll serve display ads and stuff like that. But I think it’s a misconception that Performance MAX is this full funnel campaign type. Just because it is serving on these channels that are capable of, of serving mid and upper funnel traffic doesn’t mean that PMAX is actually doing that.

You know, I’ve been. I, I got some really interesting answers from Jenny Marvin in this direction that PMAX is really a lower funnel campaign type. So there are just other things that you can do with a standard YouTube campaign and there are these more. Mid and upper funnel activities that you can take and other campaign types And where pmax is really not going to go there.

It’s not going to cover that ground. It’s going to be Just more focused on that conversion so I think that’s that’s one areas is because people Like if you if you think that you’re covering the funnel because you have pmax turn on you’re not you need those mid and upper funnel campaigns additionally if you’re if you’ve got a lot of campaigns blanketing the lower funnel Then then in that case, there’s likely to be cannibalism and you need to straighten that out

FREDERICK VALLAEYS: So, you and I both of our companies are sponsors of ppc survey.

MIKE RYAN: Yeah

FREDERICK VALLAEYS: There’s a new set of results for 2024 depending on when you watch this It may be coming out sooner. It may already be available, but go to ppc survey And one fascinating like Preview that we saw in looking at the data was the question, has PPC become easier or harder than it used to? And so the, I think from Google’s perspective, what they’re certainly pushing for is, well, P max should simplify.

And I think that’s the truth for millions of advertisers who would have been incapable previously of getting good results on Google ads. But the survey obviously went out to PPC practitioners and many of you watching. So thanks for filling that out. But what we see overwhelmingly in that segment I think they say it’s getting harder and I would agree, like there are just more moving pieces and there’s more obfuscation of data and it takes more scripting and tools and like deep diving to really understand what’s happening and to take back the control that we were used to.

So I think what you just explained is a brilliant illustration of the fact that. It’s, it is harder than it used to be. Would you agree with

MIKE RYAN: that? Oh, I do. And, and I, I mean, and it’s something it’s funny because I think that it’s a conflict in performance max that it probably has the greatest value proposition toward like a small, a smaller business.

And yet it, it works best when there’s more data. And so it’s tricky because the people who have that data are then feeling stimmied by the lack of controls and then the people who. Benefit from that simplicity then don’t have the data that it can thrive. So there’s a catch 22 there. And you know, your book was called leveling the playing field.

And I, I, I came to learn this other phrase. I think it’s from like Australia or something like that, but it’s like cutting the tall puppies. And I actually like this a little bit better because leveling the playing field. It just feels like it’s making it fair and stuff like that, but I feel like the

FREDERICK VALLAEYS: book is actually called Unlevel because the point of the book and I know you’ve read it right, but it’s basically about giving back that edge to the people watching the show here and the practitioners who do this professionally.

And what, what Google did is they leveled the playing field. Yeah. They made it easy for everyone. And we’re looking for ways like, how do you tweak the robots? How often do you do it? Like how big are the changes? What kind of scripts do you install to unlevel it back and tip it back in your favor?

MIKE RYAN: Yeah, exactly.

And, and to that point, like this idea of like cutting the, the tall poppies, I think it’s like you know, it means kind of just. Making it harder to be outstanding specifically, and that’s your point. Like, how do you differentiate when apparently 90 percent of the market is using this technology? My numbers are a bit lower, but tenuity reports 90 percent of the market and yeah, how do you differentiate?

And so there are different, you know, tactically, we believe a lot in data activation to do that as well, but yeah, it’s, and I

FREDERICK VALLAEYS: think. I mean, it’s you know, the next version of that book, right? So now it’s not just us as humans fighting against Google, cutting off the tall poppies and wanting everyone to be basically socialist and equally successful or equally unsuccessful.

But now we got GPT coming in and generative and. I mean, so before I think there was a notion that as humans, we, we still add certain value to Google’s automations and we’ll have other things to do, but now like even white collar type jobs, like, well, GPT is doing a pretty amazing job writing my emails for me.

And it’s doing a pretty amazing job generating images and it’s getting better all the time. So what is it really that we’re going to add value? I think it’s, it’s evolving so quickly, very fascinating to me, but I’d love to hear your take on that. Like, do you see like a shift within Smec, for example, of what people work on given the rapid evolution of technology?

MIKE RYAN: Yeah, I still think, I think we’re, we’re still quite early in, in the kind generative AI. And, and I also, I wonder if there won’t be some kind of a ceiling coming. I think like, I don’t know because it’s been on the market for over a year now and I remember my impression at the start was that this is going to be changing very fast and getting more powerful very fast and just wait till GPT 4 comes out and and stuff like that and yeah absolutely GPT 4 is really cool compared to GPT 3 or 3.

5 but it It’s still far from that level of disruption in my opinion. I think

FREDERICK VALLAEYS: well, I was gonna say so I keep playing with Claude and now Gemini, which is the next version from Google instead of Bart and The co pilot for her Microsoft and yet I keep going back to GPT because it’s like just orders of magnitude better and it’s something it didn’t do very well I think A month ago, now there’s GPTs and there’s new plugins and there’s the, the GPT four model became significantly cheaper.

So if the way that I’m deploying it through spreadsheet plugins and API tools that we built around it, it just becoming so much more accessible. And it’s, it’s really cool. And then also like hearing that they’re already building GPT five and Sam Altman is seeking 7 trillion, not billion, but 7 trillion, which is more than the value of.

I think Microsoft and Apple combined, that is how much money he wants to raise just to build microchips, teach these future models. And one scary thing, you may have heard me say this before, but the AI is scary. Not in that it’s going to like kill all the humans, but the power that’s required to train the next model.

If you think about these microchips and the 7 trillion that Sam Altman is seeking there is going to be a point very soon when all the world’s power is going to be necessary to maintain and train these models. What do we do then? Right. Like, do we heat our houses or do we train GPT 17? And that’s one thing that.

We’re definitely gonna have to all figure out

MIKE RYAN: hopefully the models will be able to figure out like cold fusion power or something. So we have unlimited energy, but then things are looking more positive.

FREDERICK VALLAEYS: Yeah, but these are the bottle just drifted and did cold fusion the wrong way. A small mistake just blew up the planet.

MIKE RYAN: Yeah, exactly. Yeah, but I, I, yeah, I don’t know how, cool. I don’t know if there’s a ceiling here because I also like, I don’t know what the goal is of having that 7 trillion. Is that so that there’ll be, that there’ll be, you know, generalized artificial intelligence? Like, is that so that there’ll be this kind of computer God?

Is that what, what is that? Or is it going to be only incrementally better? Because I think that there’s a diminishing returns here. And there’s also real challenges with the training data. The, the web is already. polluted compared to the way it was before. And

FREDERICK VALLAEYS: it is. And that’s where I think the GPT capability and, and this is sort of like what I’m preaching these days a little bit is build your own GPTs because you’re exactly right.

Like GPT four itself is based on all the internet’s data and there’s a lot of junk in there. And so it’s going to have biases that you may not. Desire to have, but if you work with clients as an agency now you can build a GPT that has the brand style and the guidelines and the historical data about what this advertiser has tried, what kind of campaigns they’ve run, what their mission statement is.

And if you start having conversations, it’s going to follow that pattern. And so these are confusingly called GPTs, but it’s a capability with open AI. to basically train your own model. And I think that’s evolving quite rapidly and it’s super useful in my opinion. So I’m, yeah, I think that’s going to be very helpful.

MIKE RYAN: Yeah, I, I think so too. And, you know, just that you can, you know, feed it PDFs and, like, and whatever, all kinds of, you can really, you know focus on what it’s, what it’s trained on and what it’s good at. There’s also going to be a lot happening with open source models as well, and that’ll be another thing to watch.

Cause you know, there are people who are running pretty impressive stuff like. On their laptops and yeah, but I don’t, I don’t want to get too, I don’t know if we’re still on topic or, I mean, I’m just enjoying the conversation. So,

FREDERICK VALLAEYS: yeah, no, I, I enjoy talking about GPT and so like bring it back to the PPC topic a little bit.

So in Optmyzr, we obviously have a lot of gen AI capabilities to help suggest new attacks. We have a whole conversational bot that can have a conversation with you about optimization opportunities in an account. It can narrate reports for you. So instead of just a visual, that’s funny because we went from like, Hey, I’m an account manager.

I can talk to you as my client about what’s happening in your account. But you’re like, no, show me a chart. Okay. So we have a reporting engine. We show you pretty charts and like very easy to understand data. And then it’s like, well, but can you explain what that data in that bar chart means to me? So now we’re using generative AI to bring back that thing that we used to do before we had the chart.

Okay. But it’s really cool because we don’t have to put human time in it. And the system’s quite good at understanding. So so yeah, I mean, I think that’s where it applies to PPC and, and, and I think sort of whoever can be on the, it’s again, it’s about unleveling, right? Like what is the advantage that you can gain in that system where so much of the automation is forced upon you?

And being installed by Google, but like what if you have first party data and you can somehow bring that into a generative system that then does something on top of Google systems to put in better ads and give it like that better starting point because Google can only do so much. They don’t understand fundamentally.

A lot of your business. And that’s where the humans have come in, but the humans need to spend a lot of time. And like when it came to scripts, for example, I know how to write scripts. I’ve written them myself. But like two weeks ago I needed a script for Microsoft ads. Sure. I can figure it out. But like.

Oh, is it called Bing ad scripts or is it Microsoft ad scripts? Like what’s the nomenclature that they use? I’m like just have GPT write it because it knows and I’ll fix the little errors that it makes, but it was such a time saver. And I think fundamentally where the whole world is going now is like anyone that we hire into the company, like if they give me the answer to like, I can’t do that.

Like, that’s not an acceptable answer anymore. It’s like, I haven’t tried doing that. Like GPT just opens, like removes these barriers of like, Oh, I don’t know how to do graphic design. Well, that’s fine. Like mock something up in GPT. It’s not going to be perfect. Like Dolly’s not going to be perfect, but you can now have something that you give to an actual designer to take it like that final mile and do it.

But you don’t have to like. verbally explain to them and then have them misinterpret it. Like, that stuff I’m going to do with my GPT. I’m going to do 15 prompts until it kind of understands what I meant. And it’s going to help me understand that maybe I’m not explaining it well enough for it to understand.

So like, if I mis explain it to my human graphic designing team, like, how do I expect them to do a great job? But with them, it takes a day of turnaround every time I mis express what I wanted. That’s it. Right. And so that’s where I’m so excited about like all of these barriers that knocks down and how it makes us faster and enables us to unlevel the playing field.

MIKE RYAN: Yeah, definitely. And I, I mean, I’ve used it in a very similar use case. I’ve used it to create like a sketch that then I can hand to our graphic designer. And you know, ultimately that product that he’s going to make is, Is, is, is much better in quality or it’s more what we need in the end. But just having that sketch is so cool.

I, I know I want to circle back to something that you said though, that I really like, cause I agree completely with what you said Google just doesn’t understand that much about your business in the end, because I think where Google is so strong is the audience. And, and now it’s this funny thing, especially with the pandemic.

To go back to where we started to with the privacy legislation. This legislation ends up actually kind of reinforcing Google’s walled gardens. I always see this big conflict because on the one hand, you’ve got these anti competitive legislation or regulation trying to stop walled garden behavior.

But then the privacy regulation is very reinforcing of that. But so we have a, quite a dependence and increasing dependence on on google for some kinds of data. And then there’s the first party data that we can add. But I just think first party data is so much more than audiences. It’s so much more than just a remarketing list.

And it bugs me that first party data has become just synonymous with basically a remarketing lists or email lists and stuff. Like that, when, Actually your business, you have all kinds of proprietary data. And in e commerce, particularly at the product level, which you can then feed into the platform.

So I just see, I see a lot, there’s still a lot of potential. And in some ways it’s things that we talked about for years, but it, time is a flat circle or, or there are these other market forces that are putting more pressure on there and bringing it back to the forefront, but yeah. Yeah. Yeah, I completely agree with you, Mike.

FREDERICK VALLAEYS: And so one thing that also really frustrates me is that people, when they talk about first party data, they are so limited in their view and they, like you said, they see it as a remarketing list, but there’s so much in your business. And I think we as a, as an industry just need to help people understand what is it.

In first party data, what can you do with it? And how do you, how do you action it? Right. Because I also think if you go to a business person or an advertiser and you ask him, like, tell me some things about your business, they’re going to give you a long list that might say, Oh yeah, that’s like, that’s all first party data.

That’s really useful. But now how do you turn that into something that you can act on? And that’s again, about these limited levers that Google gives you. Because even when they say like, Give us, attach an audience to PMAX, like what does that really do? Like they don’t tell you, it’s like, well, it helps the system learn.

What does that mean? Like, what is it learning? Like, when is it using it? How is it combining things? And so at some level I’m like, if Google can’t explain that and maybe it’s not that they don’t want to explain it. It’s just that it goes into the machine learning and the machine doesn’t explain and when I was at google, so I was in the on the quality score team And advertisers would constantly say like tell me why my quality score is bad and i’d be like I wish I could but we don’t have a system within google That turns that seven out of 10 quality score number into something More specific about what you could do about it, right?

So and that’s the downside of machine learning and it is getting better right now You can sort of peel back the layers of the onion But like what are debates that are being used in the neural network like?

MIKE RYAN: I

FREDERICK VALLAEYS: mean, and even if you explain that, like, what does it mean? What do you do with that? Right. So we have to go back and sort of figure out what are the levers.

How do we do something with first party data? And if that means we have to set up multiple campaigns with different settings, because that’s the only way to enact that data, then that’s what we do.

MIKE RYAN: Yeah. Yeah, totally. I think that’s Your account structure is one of the last ways left that you can reflect your, your business goals in the end.

And because if you just imagine that you’ve got a single campaign, talking e commerce here which we’ve seen, you know, we see that happening and we’ve seen Google recommend that too. You’re, you’re just, You’re telling the algorithm that everything matters equally to you. So, you know, it’s, it’s, it’s, these we’re talking about earlier in the budget and the ROAS and imperfect as those things are, it’s a way of communicating difference.

Into the system. And another thing that’s really important is the feed of course, and that’s something that gives me a little stomachache on the generative side that I know Google is getting better at generating feeds. And that this is going to be more prominent in merchant center next. And for the longest time so actually, yeah, I think supplemental feeds are back in merchant center next night.

Got to double check, but supplemental feeds were, were absent. And these things concern me because that’s another one of your, your last ways of really injecting information into Google ads. And so that I, that, that’ll be that’ll be an uncomfortable day. I think when those things are gone, you know, my concern is that one day this results in a scenario where.

This is basically just like a like a drip feed bottle that a hamster has in their, in their hamster cage, you know, and and it’s just kind of metering out conversions or something like that without without, without a lot else to it. And at that point, you know, what do you do with the channel besides just spend more, I don’t know but we’re not there yet and I hope that it won’t get that far.

I think Google has been doing to their credit. a much better job in the PMAX era of listening to advertisers than in the smart shopping era. That’s something that I appreciate a lot.

FREDERICK VALLAEYS: Great. So that rolls us into some audience questions. And the first one is from Adrian Bold via LinkedIn, and it is a PMAX question.

So I think it’s really relevant here. And I think we’ve covered it a little bit, but so the question He sometimes sees search campaigns not converting as well as anticipated, despite the search queries and the ad relevance being appropriate. Yet in the same account there might be a PMAX campaign and that seems to be doing really well in terms of conversions, cost per acquisition, ROAS.

To what extent do you think that PMAX could be claiming the credit? And basically then making those other campaigns that seem highly relevant, not perform quite as well as the PMAX campaign.

MIKE RYAN: Mm hmm. There are, it’s complicated and I think we can’t quite know the answer, but there are, there are different levels in here.

Like you know, there can be kind of an attribution shell game occurring with data driven attribution. That’s something that I’ve produced before. Data about where I see that there are these there’s an awful lot of these, these very fractional conversions. And what I mean by that is like, you know, with data driven attribution, you can have half of a conversion, or you can have a quarter of a conversion, or you can have 1 percent of a conversion and any number in between.

And we see that there’s a lot of, of conversions. that are associated at like one or two or 3 percent of a conversion. And it makes you wonder, I call this kissed by attribution. It makes you wonder if the Pumax campaign is just kind of kissing this user journey and for getting credit. There are other things going on.

Like we talked about the. The way that Google ads manages the priority between search campaigns and performance max, since Adrian mentioned search campaigns I mean, typically PMAX is, I would say not doing a lot, a lot of, I mean, it’s, it’s, I would say it’s very, very high overlap, Harry, very high degree of overlap with like your dynamic search.

And it makes sense why dynamic searches is being rolled up into PMAX. And then the other thing that we see is it’s, it If you allow it, it will absolutely suck up brand traffic out of your branded search campaigns, which I find is not okay. But, but otherwise you’ll typically see that your search campaign, the search part of your PMAX will have a very low impression share a very low impression share, and, and that’s typically not a lot of your, of your PMAX campaign.

I don’t know if that answers the question a bit. It’s, we can only look at it with like by casting side lights on the topic. That’s, that’s the whole thing. Yeah. It always depends. And I think that’s a great answer. So another question we got was as far as negative keyword management with a lot of the search query data.

FREDERICK VALLAEYS: Increasingly becoming more hidden is that still a strategy that you deploy managing negative keywords or do you just sort of see that not working anymore as the the precursor to a keyword list future?

MIKE RYAN: Yeah. And again, there are two kind of dimensions here, like in a standard, in a classic search campaign, you can still, of course, run your negative keywords, but.

Supposedly due to privacy reasons. This is people have their kind of conspiracy theories about this, but it seems like Google is hiding a lot of your search term data under like other search terms and they claim that it, you know, these search terms, I guess they could become personally identifiable and that they have to hide these things.

But often there’s a lot of costs bound up with that and you don’t, you know, you. You can’t manage that because the search terms are hidden, but I, and the other side is Pmax, which by default doesn’t have negative keywords. It is possible, but you, you have to, it’s kind of a pain. But I think that keywords are going away and I think that negative keywords then are, are also going away.

I think that within, I don’t know, a couple of years, we’ll be talking about search themes. And we will be talking about exclusions. And so. You know, use those tactics while you can today. That’s fine. They still work. They’re not broken. They might be a bit more limited, but you will still find traffic that you can exclude, bad traffic that you can exclude.

But I think in the future, this is going to function completely differently with these, with, with search teams and exclusions. And that, that’s like what we see, we see that architecture starting to be built up in PMAX already. Like, you know, The people were complaining about the brand topic. So they, they didn’t add negative keywords.

They added a new tool called brand exclusions. And it’s just a different way of tackling that. And like, I think Google does bring back capabilities. Like they brought something like search term reports back. They brought something like negative keywords back, but they do these things on their terms and the way that they’re structured is, is different.

It it’s. You know, it’s a little bit more airtight in the way that’s structured and it does what they want it to do. So it has to be acceptable to them in the first place.

FREDERICK VALLAEYS: Yeah. And I’m also smiling here a little bit because it has to make sense on their terms, but ultimately when you have tens of thousands of employees, like nobody wants to work on like, Oh, just re enable the search terms report or just add that negative.

Keyword functionality that we invented 20 years ago. Like that’s not exciting for an engineer to work on, right? It’s like, Hmm, okay. How can we use like the latest machine learning and AI to do something kind of like it, that maybe helps you. A certain question is always like, are they really listening to advertisers and what we need, or are they just like building cool projects?

And so I think there’s a little bit of both of that. So the best that we can do is like you said, you talk to Jenny Marvin all the time, like give the feedback, use the forums, use the community Voice your concerns and really explain why, right? Like B, B, B, because at some level, when I was inside Google, it was like all of these advertisers just clamoring for the old way of doing things.

Like they, they don’t wanna change. Right? And if that’s seen as the driver of the request, like that’s not gonna fly well. But if you can explain like, how does it really impact my business? And I’ve thought deeply about this and like, this is how it’s not making it work for us and this is why I’m gonna maybe spend more money with.

Microsoft ads. Oh, okay. Now we’re listening. Now we’re, now we’re going to solve your problems.

MIKE RYAN: Yeah. But I also like you, it’s very interesting what you just mentioned, Fred, that imagine you’re an engineer at Google. This could be pretty boring these days, potentially. And there’s some really, really cool stuff going on at Google.

Sorry. I’m just opening like a tab here in parallel. So, cause I won’t remember all these off the top of my head. It’s over at Google’s GitHub repository. And it is like a creative space or a venting ground, or I don’t know what, like a stomping ground for, for these otherwise board engineers. There’s a suite of unofficial Google products over there.

I’m just going to read some titles really quick. Google match markets. If this then add. Lightweight ML Toast, FeedGen, Crystal Value, Soteria, Adios, FeedX. They’re, and these are doing things like modeling lifetime value for you deploying generative AI in your feed doing like feed a B testing helping you with your geo experiments.

There’s all kinds of cool stuff that they are unofficially building. Like I think the coolest things happening in Google ads right now. Are in GitHub and not in Google ads. And I think it’s because they need to keep their engineers happy because they’ll go crazy otherwise. And I, and I, and it’s a way that you can, that you can differentiate yourself that we haven’t talked about is by checking out the, these, these packages.

FREDERICK VALLAEYS: Yeah, no, I love that. And like, check out scripts that people are building, check out different tools, check out GPTs, check out GitHub. Right. And ultimately the hope is that some of this automation that Google is building means we don’t have to spend quite as much time like hunting down that perfect negative keyword, like you can automate that and now you can deploy.

Like you said, better geo testing. And spend some time on that on that. And then hopefully that’s going to be a bigger impact than like that one negative keyword that took 5 a month and clicks away. Right? Like, no, now we’re talking about a big geo inside, like, Oh my God, this market actually adds you know, this much more volume, right?

So, and that could be millions of dollars as opposed to 5.

We could keep talking for hours. So hopefully we’ll have you back on another show, but let’s do a few rapid fire questions, Mike, if that’s okay with you. So the first rapid fire I have is what’s something you wished you had known before you got into PPC.

MIKE RYAN: Hmm. Let me think about that. I, I, I’m not sure. I, you know, I think it’s a classic question, but I, I honestly, I really don’t know, I mean, I don’t know. I will say that I I held Google on a very high pedestal back then. And I still, I, I like all the Googlers that I work with and I, and I have a lot of respect for this company and I love using their products, but I understand Google in a totally different way than I did several years back.

So that’s just something that has changed. It’s totally changed. I think everyone, as you get older, you start to understand the way the world works. And it’s not always like happy things. So but being an ex Googler I’ve worked with amazing people. Everybody was out there trying to change the world for the better, but there’s also a reality, I think of what the world is and we have to operate within that.

FREDERICK VALLAEYS: We may have covered this before as well, but what’s your favorite AI tool and how do you use it?

MIKE RYAN: It is chat gpt. I do I do have You know a premium subscription or we switched over to the company subscription and I have my frustrations with it I think there’s you might have heard these reports about gpt being acting kind of lazy and I I have experienced that too It’s amazing.

I mean I Just on a personal level, like I, I, I bought a smartphone entirely using GPT. And it’s concerning maybe for some people because I didn’t do any searches on any. Search engines. I wasn’t exposed to all those blog post roundups full of affiliate links. I didn’t see a single shopping ad or search ad.

And it was powered by search, you know, it was powered by Bing search, but I didn’t touch a search engine. And that’s another I’m using Google. I’m using Bing. I’m using it in a search like way. Or in instances where I would have used Google search in the past. I’m using it that way way more often.

But again, I’m not, I’m not really using Bing either. I’m just, it’s powered by Bing. Yeah, but then on like a, on a professional productivity level. I love it for helping me write more complex formulas than I could do on my own, or that would have taken me an hour of research to figure out stapling together, checking out different stack exchange and stuff like that.

And it’s just happens. So I love that. And yeah, it’s a really cool tool.

FREDERICK VALLAEYS: And I like what you said about buying a phone completely on the Chat GPT or the Microsoft experience. Like one thing that frustrates me is the amount of times that I have to pull out my phone during the day to do mundane things like, hey, we’re out of toilet paper.

Let me put it in my shopping cart from Target and let me get that redelivered. I’m like, and I love the Alexa voice assistant. It’s so convenient. But it’s limited because everything you get is from amazon and I don’t want everything from amazon. I want some like I want unbiased reviews from across the internet and that’s sort of the promise that gpt holds because it has ingested the whole internet and its results in the case of microsoft are still driven by the top 10 organic listings that might be digested into a summary of like here’s the phone you need to buy and then You know It works with the ads to send you to the vendor that can sell you that phone and take the action to buy the phone.

And of course on Google side, it’s generative AI does the same thing. It runs on top of the search engine, but that’s sort of the promise for me is that that day when I can just have like a little button where I push the button and then I, I speak to it and it accesses the whole internet and I don’t have to look at screens anymore.

So that whole voice search thing that was so. overhyped. I think that might be coming back in conjunction with the smart assistance and a GPT capability.

MIKE RYAN: Yeah, I mean, I’m so tired of typing on my smartphone that I can’t describe it. I don’t know why, but I just I’m done with it somehow. And but I’m not.

I’m not convinced about these pins either. I still think until there’s a I think maybe glasses could be the way I’m not,

FREDERICK VALLAEYS: I’d love to be the humane pin. That’s starting shipments in about a month. And then I got my vision pro that’s arriving any day now. Again, I don’t see myself running around with these and then they’re going to get smaller.

Right. But. But like, I just want to live in the world and be able to ultimately have my thoughts. I don’t even really want to speak it. Like I’m just thinking, Oh my God, I need to do this thing. Like, why can’t that just be done? Like if I went to chat GPT, it would do a good job helping me, but I have to type it.

I have to formulate, I have to read it. I have to then go to a website, like, and it’s all the pieces seem to be there to take us very soon to that next level of. Usability that we’ve all seen in movies like Minority Report and, and Elon Musk is working on Neuralink. So he’s done the first embedding of a chip in somebody’s brain.

It works right. And his big vision is like when AI is out there to get us like human plus machine better than the machine alone. That’s ultimately what Neuralink is going to be able to do is put a supercomputer generative AI inside all of our brains and no more screens, no more typing, just, you think it, you got it.

MIKE RYAN: Yeah, it’s, it’s a wild idea. And also, I don’t know, it’s, it scares me a little to imagine what that would be like, because I think it’s a, it’s a different way of existing. Yeah.

FREDERICK VALLAEYS: I mean, maybe he’s going to escape to Mars and then leave all of us here to figure out how that neural link and GPT and all of that changes the world.

And if it looks okay, they’ll come back.

MIKE RYAN: Yeah, exactly.

FREDERICK VALLAEYS: But also Mike, this has been a really fun conversation. Everyone, I hope you’ve enjoyed it. Taking away a few tips from Mike Ryan, please follow Mike on LinkedIn and check out his company, smarter e commerce. And then if you’ve enjoyed this, please subscribe, hit the button, like it, and we’ll be back with another episode of PPC town hall very soon.

Thanks for watching.