Should your business be using AI Max?

All right, Andrew, it's great to have you on my channel in Optimizer Studio in a country that neither of us live in.

So neither of us live in, neither of us born in, neither of us is from.

Hey, I have had Andrew on before, but just before we get into it, why don't you give us a quick introduction? I think it's Andrew from Savvy Revenue.

Savvy Revenue, Andrew Lok. So I've been in Google Ads last 15 years. I've run a large agency, 200 Employees, a couple thousand clients. Now just run a boutique e-commerce Google Ads agency. So I call us a little bit more boutique, a little bit more elite, large clients. But I have the experience of smaller clients, so that's typically where my advice and my background will lean towards the bigger side. Yeah. But I have the context of the smaller.

So 15 years Google Ads, like margaritas in the rain, all of that.

Been through it all. Yeah.

Hey, I wanted to get down and have a bit of a conversation because I think we would both agree that we're really, really excited about AIO Max and really see it as a core feature of Google Ads in the future. I don't think I'm putting words in your mouth there.

No, no, no. So I'm like, I for one am not a fan of most of the things that Google come out with. Yeah. I'm vividly against PMax as a concept. I don't like most of the things that come out of Google. AI Max, I'm very excited for. Okay. I think the direction is very, very interesting. Okay.

And this is where I wanted to have a bit of a conversation because for, for me, I'm probably the opposite where I, I like AI as AI Max as a concept. I think it'll be very beneficial in the future, but I'm not sold on it yet. And I would actually put it in the same bucket as Performance Max in 2022 in that I didn't like Performance Max in 2022 as a product. Whereas now, especially on the e-commerce side, we're engaging Performance Max within 90 days of an account, most, most of the— across all of our e-commerce brands.

And I had disbanded that within 30 days of taking over an e-commerce account. Yeah.

Now, obviously for us, now both of us would agree that it comes back to the results that we're seeing inside our clients. I'm big enough to also agree that ultimately there's more than one— we say in Australia, more than one way to skin a cat. Like, you can get to you can get to different things in different ways. So I did want to talk about, so if I'll give my concerns with where I see AOMAX right now, and then I'd like to get your insights. And the reason with this is that I like to have these conversations because ultimately I go, you know, we both know what we're doing inside of Google Ads. We both are getting good results for our clients. For people listening, I'd like to give them some different points of view rather than just the marketing message of Google. And rather than just, you know, some people will go, completely hate it, other people go completely love it, whereas I'm in the middle, is in that I want to believe in AI Max, and I do see myself actually using it in the future. My issue with it now is where I say that in two things. One, I see it as 2022 PMax in that it's not the full product yet. It probably needs a couple more releases before, or releases and updates before I'm ready to use it. The other, my other big thing with Performance Max, right, sorry, with AI Max, way too many itself. Yeah, way too many maxes in, in Google at the moment. But I see, when I look at it from a current client of how we're building account structures, so I go with AI Max. And now some of this may change tomorrow, Google, Google Marketing Live, but as we sit right now, there's 3 main ways to fit, to see your ads in AI Featured Ads. It's either using broad match, either using AI Max, or either using Performance Max. We've audited some accounts that have AIO Max, and when we go through and do the review, if they're using broad max previously, I think I released some data in that it was— some of these data points may be a little bit off by 1 or 2%, but it was something like 87% of search terms that were being generated inside of— we're talking your search campaigns with AIO Max— were already being hitting with broad match. And broad match was already hitting, I think it was 92, 93% of the conversion. So we didn't really see a big uplift. So that's where I sit with AI Max at the moment. So I'd love to get your viewpoint and then we can go through and have a discussion on where we disagree and where we agree.

I'll skip the agreement part because that's not, it's not interesting to hear me like say the same things as you said. Yeah. But there's no doubt in AI Max has been rushed out. It was rushed out last year and it hasn't really come out. And so what we've done inside of Savi is that we have enough accounts and large enough clients. And because we're in Europe, we have usually we have the same accounts in multiple countries. So that means that we can test some things and compare the trend lines across basically the same account with running multiple different, multiple different countries. So I got tired of sitting and trying to find out where does AI Max actually work because we also saw pretty bad results, like the wrong headlines. Yeah, very unclear what AMAX was actually doing. Like, why is it showing up? Why is it being listed as an AMAX match type when we already have that as an exact match match type, etc.? The best result we ever saw was pausing Dynamic Search Ads and enabling full-on AMAX. It took over from Dynamic Search Ads within a day, which surprised me, and performance kept strong. We didn't see better results, we didn't see worse results, which I had anticipated. Our accounts regularly spend between 20% and 30% of the search spend on DSA. So when we just paused DSA, I expected it to drop, and it didn't, and I was very surprised. With that being said, a lot of the AMAX features that you expect, like AI-generated headlines and all that stuff, didn't happen a lot. The URL expansion didn't happen a lot. So it was a lot more consolidated, but overall performance didn't really drop. And my concern, where I think we both agree, I couldn't explain why. The reporting is so convoluted around AMAX today. It's like, what is an AMAX match type? What is the final URL expansion? Where do I see the reporting? And then there's some privacy things that get hidden and all all that stuff, and we can't really figure it out.

So I did see that, that post you put up the other day about the results in one of the European countries. A couple of questions I do just, just have. I actually, um, did write this on LinkedIn, but then because I was traveling, it got cut out, and I thought, oh, I'm going to talk to you in person anyway, so, so I won't post it. But I do have some, like, just structural questions. Is that one of my concerns? One of my hesitations, I would say, not concerns, hesitations with AI Max is there's no— I haven't seen any clear data around where does it perform best.

Yep.

Now the reason why I say that is because with Performance Max, when it first came out, because that was a campaign that was built around smart bidding, we could use the parameters of smart bidding where we know, look, this is an account that is ready for smart bidding. So there's a fair chance that Performance Max will work well. We would also internally build out data on— we wouldn't engage Performance Max until we saw some core metrics around click-through ratio for shopping feed, conversion metrics for shopping feed, just things that we see that we know. Because for me, for Performance Max to actually work successfully inside of an account, it needs to be 95% plus spend on the shopping for an e-commerce brand for it to be successful. So a couple things I ask with the— so you had, so you had AOMax in a separate campaign. What were the other campaigns in that account?

So one of the experiments were to have So dedicated search campaigns at DSA running separately and then creating a separate AI Max campaign. Okay.

I agree with you there because the one thing I don't like is the messaging from Google is that just to turn on AI Max within—

I'll get, I'll get to that. Okay. So that was the first experiment.

Yeah.

And the AI Max campaign did nothing. Okay. Like it's like back in the day when PMax just came out and it was this ultra priority and just turning it on took sucked all the life out of standard shopping. Uh, it's like the opposite here. Like, as long as you have DSA running, AMAX is doing nothing for you. Yeah, which was surprising to me. So there's a weighting there internally on an account that I don't think is weighted towards AMAX yet. The second experiment we did was basically the same thing— dedicated search campaigns— and then we quote unquote upgraded the DSA campaign. Yeah, by simply enabling AMAX on it, pushing the different like page content, the different rules we already had in the DSA. And AMAX as a match type got zero clicks, which makes sense because there's no keywords in the campaign. And the final URL expansion was also like lackluster. Not much happened. So there seems like there's some kind of DSA component where DSA is still weighting higher than the AMAX internally on accounts. The third experiment, What's the one you didn't like? Yeah, so that's the one where again, we continue to have the dedicated campaigns. We enable AMAX fully with AMAX, final URL expansion, and text customization. No page feed, no exclusions, no nothing. And we paused the DSA campaign. Yeah, that was the one where AMAX immediately took over.

So in that one, so you had AOMAX, just to confirm, you had AOMAX running in the 12 dedicated search campaigns. Yeah. So you had no dynamic search. So, and then you had shopping running in the account as well. So you had, you only had shopping campaigns and search campaigns with AOMAX running.

Yep.

Okay. And in those search campaigns, were their broad match previously?

No broad match, only, only phrase and exact.

Okay. So that's something that where I would go. So for me, I go, I actually think we're closer than what we probably, not that I think we're against anyway, but I would still go, we're still seeing that if that's shopping, sorry, if their search campaigns are running broad max, sorry, broad match, broad max, broad match, that there's not by turning on AI Max, that's the test that where we saw where it was engaged inside of a current search campaign that had broad match. What AI Max did is that it took credit for all of the broad match searches. So we, to pull this data out, we took the 7 weeks from when AI Max was engaged, looked at the search terms from the previous Now, yes, we are losing some data because it's in other search terms, but it did—

but it didn't go any net new.

No, it didn't go any net new.

So when you— before, when you talked about like, okay, so you have like a decision tree on like these factors need to be, be in a certain— these conditions need to be set in order to run PMax. The same thing. I'm like, I feel like we're there with AMax. Yeah, because we can, we can learn a lot from dynamic search ads. We can learn a lot from broad match. So, uh, in my agency we run very little broad match. Okay. And the reason why we've never engaged really in it, we've tested a bunch of times, but we believe in like the holy grail of Google Ads is the search term needs to match the ad, which needs to match the landing page. Broad match fundamentally breaks that. So if you have a keyword like mountain bikes and you have another keyword that's called, uh, like What is like racing, racing bikes. If you pause the racing bikes, mountain bikes can take over that search term via broad match. But then you show a mountain bike ad to a racing bike search term, performance drops. CTR is lower, conversion rate is lower. So it's just, it's a poor experience. Broad match as a whole is a poor experience for the user and as an advertiser. And that's why I think that, and this is going to be my big claim, I think AMAX is going to kill broad match. Because I think that the, the broad match functionality as a match type is embedded in AI Max. That's why if you turn it on in an account, it will say you shouldn't run broad match. And but the ability to write a better headline that matches the search query and sending to a different part of the website where it's a better match, that's the strong part of AI Max. The very uncomfortable part about all this is trying to figure out what the heck is going on.

Yeah. So we, we're— I'm very pro-broad match, but what I would say is the way that we view it internally is that the most important guardrail is actually your conversion data. Yeah. So we would then heavily rely on being able to set the right, uh, so in e-commerce setting, that's the right target ROAS goals. We definitely do have some guidelines we set around negative keywords, but we're we're loosening them more and more. What we generally find—

just go no negative keywords. Yeah, just come to the dark side, Aaron Young. Come to the dark side. No negative keywords ever.

What we generally find is the better conversion data, so in terms of actual volume of conversions that you have inside of an account, the less guardrails you need. So for me, we're really framing it that your conversion data is becoming I, I have to be careful when I say this because people will run off with it, but I go, I think the— and we've seen this inside of data— that the conversion data and the bidding strategy that you're using inside of campaigns is just as important as the keyword targeting that you're using.

Pretty much like it's, it's today, it's everything.

Yeah. I'd like to finish on this. Have you got any internal frameworks? Because we haven't got frameworks yet from Google about, um with AI Max, what should be the minimum? So let me phrase it like this.

Yeah, I have, I have thought. Yeah.

Is there a minimum number of spend or clicks or impressions or even conversions that you want to see in a month, let's say for a 30-day period before you even think about engaging AI Max?

So I would take one step back. I would say people have gotten very lazy running Google Ads today because some people were brought up with standard shopping doing 70, 80, 80, 95% of the conversions in an account. And then the work that's associated to running search ads is 5 to 10 times greater than just running a standard shopping campaign. PMax came out and that just flipped everything on its head. Very few people are engaged in, in building out real search campaign structures today. So I'll say before you do AMax, make sure you just build out the fundamental campaigns that you're supposed to have. You don't have to be as detailed as before where we used to have, okay, so you have your generic keywords, your category keywords, your, the brands you sell on the site, the combination of brand and category. You have the products and then you have all the custom things for like the model number, series, series and stuff like that. But just build out the basic campaign structure for search first and foremost, because AI Max is going to expand what you already have. And you want to have it do that from a base. Yeah. So I would say start there, start with building it out, and it doesn't take long. Just do it once. Claude can help you. It's, it's not that complicated, but just do it once. And then once you've done that, it's— you need— I go very high in minimum conversions. We also work with very large accounts. Yeah. But I say just be really careful. Anything less than 100 conversions a month It's just— then we're talking 3 conversions per day across multiple campaigns. It's very little data. So when, when I say that, people just— they, they look flabbergasted and like, what are you talking about? It has to be at least 100, and, and I like 300 to 500+ per month for some of these things really to work. But then the really big thing to look at is your footprint on your search terms, because AI Max is supposed to expand it. And this is where I think a lot of people are talking about AMAX wrong. And it's the same conversation that went wrong with broad match. Broad match sometimes works exceptionally well if you have something to expand into.

Yeah.

If you're already covered all your core keywords— let's say I'm selling running shoes and I'm only selling one very particular type of running shoe, then there aren't a lot of keywords I can have. So if we're, if we're focusing on this type of running shoe, let's just say a marathon running shoe, um, and that's it. So I don't have trail running shoes, I don't have hiking, I don't have walking, I don't have all that stuff. But I have a search campaign set up that's focused around the marathon running shoes that I, that I'm selling on the site. If you engage in AIMAX, it's going to try all the other search queries. That's a bad idea because it's most likely not going to convert. Just like I have a jewelry client, we sell jewelry for men. Being just having the keyword bracelets is a horrible idea. 98% of the people in the demographic that searches for bracelets are women. So we can't bid for that no matter what we do. So I think that's the most important thing, like, is your site supposed to be expanded? Now we have another client that sells, like, they have 500,000 items or SKUs on the website. It is built for AMAX. Yeah, because it can— I can't build out enough keywords that, that preserve the search term to add to landing page relationship. I can't build it out. I can't keep it, I can't keep it updated. But that's what Dynamic Search Ads did. So amazingly well, and that's what I believe AI Max will do, just even better.

But the clarity there is you'd be saying that you're, you're looking at accounts— so you— not to put a number on it, but are you putting a number of you'd be wanting at least 100 conversions a month before you start going into AI Max?

Yeah, I'll be very concerned about anything less. And the reason why is because if you don't have conversions, then smart bidding doesn't know what it's supposed to be bidding for. Yeah, it's like having shopping ads on manual bidding today, and I hope people don't do it anymore, but I've done a couple of times just to try to see what happens. It's like, it's that like smart bidding bids higher and lower at the search term level. That's what it does inside of AI Max campaigns as well. So if you don't have any conversions on the various search terms that AI Max is expanding you to, then it doesn't know if it's good or bad. And, and people will ask me a lot like, what, what do I do if I have 5 or 10 or 20 conversions per month? And I go I feel bad for you. Like, it's the worst place to be. Like, it's, it's so hard because you usually can't get really— like, the people who are really good at what we do, we take the clients that have a lot of, a lot of conversions. Like, a 5 conversion, you can't— it's such a hard— it's such a hard place to be.

I am a firm believer, and I say this statement all the time, I think Google Ads is harder to start. This is talking if you've got a normal budget, not even a small budget, a normal budget. Google Ads is harder to start than what it was 5 years ago, but so much more easier to scale because once you've got a data, once you've got that right data coming through, you can leverage all of these tools.

It used to be hard, like at scale Google Ads used to be hard because you did manual bidding. Yeah. So you had to be really good at designing that manual bidding system. You don't have to do that today. And so in that sense, it's, it's amazing. I do much like I do much better now than I have ever done, and I was really Good at manual bidding.

Okay, so I look at— by the end of it, I look— I really appreciate your insights. I would say we probably agree more, um, from what I've taken from that is that you would definitely— you would still start search and shopping, only engaging AI Max once you've got that high, high, high level of, of quality of, of conversion data.

I would say engage in AI Max earlier if you have a large SKU count on your site. Like, if you're only selling one type of shampoo, then like enabling AI Max to expand into more ways of saying curly hair shampoo is a bad idea because there are no more ways of saying curly hair shampoo. Like, it's just, it doesn't exist. Now, maybe in the future when AI Mode runs at a certain volume, then maybe the ability to custom or personalize the ad through AI Max in, in AI mode, that might be a thing where it becomes an important thing. It's not quite there yet, but I would say definitely like high SKU count, I would engage earlier because I think it just works. DSA worked.

And I think, just thinking with your experiments, you would also say, not to put words in your mouth, that run a test first. So whether if you're running across multiple countries, try it in a country, or if you're running, say for example, all in the US, maybe just segmented out to a separate state, um, depending on data though.

That's the hard— yeah, so I would say I wouldn't be— I, I'm a huge fan of testing, experimenting, because I believe that that's the only— that's the best way for us to understand and avoid this, this black hole of everything Google has is a black box, because you just go, I don't know what was happening. And that's— it's so far from the truth. There's many things you can do, there's many things you can learn. I think that AirMax generally has a bad rep right now because of nothing really happens when you turn it on. You can't really find out what's going on. The experience that we've done lately, and again, this is from the last month, shows that you can actually just do it. It doesn't look like it's a bad thing. You of course need data to go that route, but I think more people should just try to switch, switch over to AirMax from DSA and not be as worried about it.

Andrew, pleasure to have you on again, and great to be in person this time. And thank you so much for sharing your insights on what you're seeing in real-life accounts with AO Max.

Always a pleasure, Darren.