Growing Ecommerce – The Retail Growth Podcast

Are Your Google Ads Lying to You? - Marginal ROAS Explained

• Smarter Ecommerce • Season 4 • Episode 20

Share episode

In this episode, Mike Ryan and Chris dive into the fundamental issue plaguing advanced performance advertisers: the lack of data integrity between ad platforms and a retailer’s backend systems. Why are sophisticated retailers taking optimization decisions on ad platform data when their final bottom-line evaluation comes from a different source? Mike calls this the "optimization gulf".

The hosts argue that this disconnect is a major tactical issue that is driving platforms like Google and Meta to pursue cross-channel solutions to regain trust.

Key discussion points include:

  • The Single Source of Truth (SSOT) Problem: Why using last-click attribution and optimizing within channel silos led to platforms taking 100% credit for the same purchases.
  • The iOS 14.5 Inflection Point: How Apple's anti-tracking policies broke Meta campaigns, leading to the rise of "magic pixels" (Triple Whale, Northbeam) to fix reporting.
  • Marginal ROAS vs. Average ROAS: Why optimizing only for average ROAS is a "massive flop" and why you need to understand the concept of marginal return to avoid spending beyond your optimal point.
  • Google's Contradiction: We break down Google's claim that its bidding system (PMax) finds the cheapest conversion across all networks, and the subsequent contradiction of why the performance of those individual channels often looks bad in segmented reports.

About smec (Smarter Ecommerce):

At smec - Smarter Ecommerce, we specialize in transforming business goals into optimized ad campaigns. With over 16 years of experience in Google & Microsoft Ads, our intelligent software and expert services help retailers achieve superior results.

We're committed to giving you the tools and insights needed to stay ahead in the ever-evolving world of digital advertising.

Make sure to follow smec - Smarter Ecommerce for more performance marketing insights:

SPEAKER_00:

Welcome to another episode of Growing E commerce. I'm one of your hosts, Mike Ryan, and with me today is Chris. Hey, Chris. Happy to be here. Me too. Any small talk on your mind today, Chris?

SPEAKER_02:

Man. Yeah. We can talk about uh the tub you installed.

SPEAKER_00:

Oh yeah.

SPEAKER_02:

That was the next thing. Yeah. Let's carry on. Do you love it?

SPEAKER_00:

It's great.

SPEAKER_02:

It's great. Tub and shower, yeah. But I guess you love the tub, but I think the children love it more. Of course.

SPEAKER_00:

Of course. We're getting way off topic here. My daughter's uh phrase um is Pichypachi Pokeypaky. Can you figure out what this means? No. All right, I'll decay. Enlighten me. It's a blend of uh of English, German, and baby gibberish. Nice. Yeah. What does it tend for? She's saying like splishy splashy. Oh, okay. And then pokey paky is she's saying popo pack. And and yeah, so for non-German speakers, this is like a childish way of saying butt cheeks. So it's her way of saying splashing around naked. And for her, it's like a battle cry. Love it. Love it. Yeah.

SPEAKER_02:

So uh Did you buy the top online?

SPEAKER_00:

Just out of curiosity. No. No. Because that was through the we bought we order that through the plum. Okay. Got it. But uh I buy almost everything online. I'm I'm I'm a loyalist to e-commerce. That's good. I do buy almost everything online.

SPEAKER_02:

You're loyalists to e-commerce. Yeah. Okay. That's good. Exactly. I think there are there are there but there's there are cases where the the good old buying behavior, going into a store, you know, feeling seeing the stuff is still and for me it's it's it's there are still areas where I just don't bound.

SPEAKER_00:

Yeah, like if I have to sit on it, like a chair or couch.

SPEAKER_02:

By the way, we we got a new couch, we bought it. Online. Online. I can't imagine. Brave. Yeah, brave. But uh it was a good good choice. All right, good. May I may I drop the name where we bought it, or is it Go for it. West Wing. West Wing, yeah. Okay. Nice. Amazing, amazing online shop. Great product. Love it. Very good. So, okay. Small talk done. What do we have?

SPEAKER_00:

Yes, let's get into it. Do the letters SSOT mean anything to you? Uh no. Single source of truth. Holy Oh, sorry. Uh wow.

SPEAKER_02:

Wow. Are you trying to keep it clean? Yes, I try to keep it clean. I almost said holy shit. I was gonna say how generous of you, but uh singing, okay. This this is a this is a great topic. Yeah. Let's let's dwelve into it.

SPEAKER_00:

Uh absolutely. So um here's the thing. Here's something that I've I've been observing, and maybe you two. Yes. Um I mean you can you can even lead us in, but I just feel like um the amount of advertisers who are measuring and really optimizing for it based on a data source that is not the ad platform has just been growing and growing and growing over the past years. For sure. Um I don't know. What do you think about this, Chris?

SPEAKER_02:

Um yes, for sure. Uh so uh give us an example. Do you do you have an example? Uh I I've I have a couple of examples. I mean, by the way, this this is I think there's there's no right or wrong, and it's it's um it's it's an opinion here. Uh uh I'm I'm sharing with you guys. So um that's what I want to put on Front Street. But in general, what what I see is, and by the way, this is this is I'm a very biased person, and this is certainly uh recently biased. I talked to um a couple of decision makers of of rather huge online stores, almost enterprise level. And so basically it's no surprise that these companies um look at their own data in the back end. Right, of course, yeah. Which which is fine, and I think it's it's it's good to do so. Where I see just a huge challenge and something which is just not not my my my mind, I can't wrap my mind around it is you have your data set in the back end which is reflecting the true impact of your channels. Fair enough. But at the same time, these super advanced retailers take optimization decisions on the data of the given ad platform. So for instance, Google, right? PMAX. Well, if if you try to optimize your PMAX revenue based on Google Ads data, but the final evaluation of the PMAX impact on your bottom and top line is uh deferred from some other data set. Yeah. For me, and I'm I'm not an operational expert, but for me, this is a breaking point which I think is not discussed about often enough. Uh I don't even know if there are out-of-the-box solutions to tackle this. But this lack of data integrity really m moves me. I mean, I'm I'm not kidding. It it really it really bothers me because it's it's it's it's not an easy challenge to overcome. And for me, it feels like a fundamental issue. That's so there are a couple of examples I can bring up. I I don't want to drop any names here, but rather big enterprise level clients have the same issue. There is one data set in the back end, they take fun decisions on. Operation teams are looking at the ad platform uh data like Google Ads, and there is a lack of data integrity. And what the clients always always tell us is yes, they're aware of it, but this data is not matching up. Um but but there is sometimes just no solution for it. Yeah. And I think it's a fundamental issue. Yeah.

SPEAKER_00:

Yeah, I mean, I I I agree. I I view it in a pretty positive light because I think that we are these are growing pains. Um and because if I just think back to several years back, like last click attribution, going hard in good old days. Yeah, exactly. Each you know, each each each kind of channel um or platform was its own little silo, and you were optimizing exclusively within those walls. Last click going brrrr like a money printer, um, generating And you know what the cool thing was, Mike?

SPEAKER_02:

Yeah. You if you added up the revenue of all your platforms, it was 50% higher than that.

SPEAKER_00:

At least, oh man. 100%. 100%. All these platforms taking credit for the same purchases and stuff like that. Um and you know, there's just been maturity, like moving on toward multi-touch attribution. Um Marketing mixed modeling. Yeah, marketing mixed modeling becoming more and more popular last three, four years, uh having kind of a resurgence. Um and and also like to me, one of the big inflection points here was when Apple decided to just nerf uh nerf meta ads or Facebook ads. Yeah. And I'm you know, I'm referring to uh iOS, what was it, 14.5, or I can't remember anymore. Um and ATT, the anti-tracking stuff. So um this just broke everyone's meta campaigns and the way that they were measuring and optimizing. And like what happened, there was this gap and all these all these tools flooded in, like Triple Well, Rockerbox, North Beam. I'm not advocating one or the other, but they're kind of I called them magic pixels. Yeah. Um but you know, they would help fix your reporting. But then these people, they had this kind of fixed reporting, so it seemed, and then they're advertising and then they had to take actions in the platform that has no idea of what's going on. There's this disconnect. That's important. And yeah, I mean, we also a similar situation arises with Google. Maybe you have something custom in Floodlight and search S360. If you use SA360, um, and they now that's kind of harmonized a bit, but uh but yeah, like over time it's it's kind of good that we're looking at these other sources of truth. I think that's very positive in a way. Um but there there is kind of a optimization gulf between. So you'll see crazy things like um I'm going to I I need to add like a X percentage multiplier on my in-platform ROAS. Uh like I whatever my in-platform ROAS is, I need to set this much higher so that it corresponds to my approximately to my back end ROAS.

SPEAKER_02:

And by the way, this this is what's what's going on something.

SPEAKER_01:

Yeah, sure, sure.

SPEAKER_02:

And we have with pretty smart people at Smart ECommerce. And and and uh one, for instance, one one huge, huge uh German client. Uh we are working with them almost 10 years, right? Uh it's a trusted and and very, very stable relationship. Um they have a pretty advanced, and we supported this client uh also setting this up, but they have a pretty advanced understanding of the true impact of each and every channel, yeah, within Google, but beyond Google. So also across ad platform for you. And this is the data basically our teams are optimizing on. So we get the daily report from the client, and what you brought up now, this is how our team so they understood what does it mean if the rowers in the back end I don't know, increase by uh 4% point. What does this mean in terms of translating this whole thing to the Google Ads ecosystem, for instance? I think it's it's it's w way better to do it like that compared to just relying on the ad platform data. That so I'm with you, this is a good thing. But it feels like, man, we're in the year 2025. Yeah, it feels so it's clunky. Yes, clunky, and and and and and so much interpretation is needed. Um where um yeah, where I'm like, okay, isn't there something more advanced out there? By the way, one one one final example I want to bring up, um, again, of course, recency biased, but I think it doesn't make it less less important. Especially advanced retailers uh are have a have a pretty established testing mindset, which I love. You know, your setup should be tested all the time. Is there a new setup which brings you more incrementality? Now, we did a pretty comprehensive uh PMEX test with a client where we we set we we basically tested a new setup. It was a cheap split test and the data within Google Ads was pretty pretty telling. The new setup won by a wide margin.

SPEAKER_01:

Yeah.

SPEAKER_02:

Now the fun part is that in back-end data and the data was not supporting the data of the Google Ads platform. Yeah, yeah. Now we are in in advanced and sophisticated discussions with the client. So what are we doing right now? One thing I can only say is the new setup optimized on Google Ads data, the old setup optimized on Google Ads Data, and within Google Ads, the new setup is clearly the better choice. Yeah. So how much coincidence is it that in the back end the data looks different?

SPEAKER_01:

Yeah.

SPEAKER_02:

Or is there a pattern? Or so I think it makes it so hard to also take optimization decisions.

SPEAKER_00:

Yeah, you know what I mean? It's that you can move the needle on on one and not see a corresponding move on top.

SPEAKER_02:

And is this a coincidence now? Is it um I I really feel like it's it's it's a big issue. Uh a big issue. It's a substantial issue. Um I think it's the right step in the right direction to not rely on on that platform data because as as honest they are, but yeah, they they claim, of course, probably more revenue uh to themselves than than it actually um should be. But this this you call it optimization golf. It there it is uh an issue.

SPEAKER_00:

Yeah. I I I mean uh and I see it I see this as threatening toward the platforms as well, because um the you know they are not trusted anymore. And um they they are not the source of truth. And all this reporting that they build, what is it all good for if no one trusts it or uses it or wants to use it? Yeah, for sure. And you know, like I to Meta's defense, they after ATT, they threw a lot of money at the problem and technology and engineering, and they restored a lot of the fidelity of their of their tracking and and being able to report back what was going on. So that some of these other tools became a little bit less necessary. But on top, I mean, in the intervening time, those tools were well adopted and they earned a lot of trust. Um and what what Meta did, I think a few months back ago at this point, and I think this could be the way of the future, they've just started integrating with these tools. If they, you know, some of these bigger tools, they they just support it. And so that they are closing that gap. And Google is not there yet. Um I think the only case, like if you have your again, SA360 probably is if you have your custom stuff going on. True that. Um But I just like that's why it's no coincidence coincidence, though, that they're both investing in MMM. Like that uh Robin. Meridian and Robin, yes. Yeah, Robin's been around for a while, Meridian is newer. But they they see that if this isn't working at the channel level, if if they're not the trusted source, um then they then they need to find the way. So they'll go cross-channel. They they need to own that conversation because you're probably not using, you probably shouldn't be using like a lot of uh multiple MMMs. Um that that's where the conversations between your CMO and your CFO, these are MMM that's uh that's you're owning really the interpretation of what's happening.

SPEAKER_02:

I think MMM marketing mixed modeling, I think it has been around the corner for now two, three, maybe four years. I I think it it it will remain a very, very hot topic. For sure. Uh well I mean, of course, then again it's it's a it's a trust issue because at the end of the day you have to trust your marketing mix modeling. But I think at least you have a valid single point of truth you you can look at. And the MMMs are of course um trying to figure out these cross-ad platform relations, correlations. I think I think the MMMs are the the again, probably the next step uh every every online retailer should take. Because right now, the the issue of having looking at the data in the platform, let's say Google or Meta, looking at the same time your back-end data and the data is not matching up, I think this has tactical implications massive. Because if I was a CMO, I would ask my online marketing performance team lead, okay, dude, you're optimizing on Google Ads data. How how how are you making sure that this data is really supporting what what's happening in in in real life for us as a company? Um and if the answer was yeah, I'm just looking at the data and try to find out correlations, I don't know, man. It feels clunky and not state of the art. Yeah, I agree. But you can see this across all industries and across all sizes. For sure. It's it it's just a widespread issue. For sure. Um Yeah, ma'am, I think there's no out-of-the-box solution. What I think what we can agree on, uh Mike, is marketing mixed modeling is one way to get more clarity into this whole thing.

SPEAKER_00:

Yeah, but it's again, it that's at a totally different kind of altitude. That's that's way that's it's it's it's not tactical. It's it guides your your overall spend and it helps you determine things like your saturation curve and stuff like that. But it it's it's solving for different problems.

SPEAKER_02:

Multi-touch attribution is falling a bit out of fashion because of its dependence on user-level data typically, and I don't know what other alternatives are, but um the ideal scenario honestly would be because marketing links modeling, I mean, uh depending on on how granularly you set this up, but marketing links modeling is there are good models out there, and and I think they're really understanding the true impact of let's say Google PMAX versus Meta Advantage Plus. And I think you will get good scenarios presented. Like that the marketing links modeling suggests you to spend 20% more of your budget into PMAX.

SPEAKER_01:

Yeah.

SPEAKER_02:

But it's not a tactical advice, right? Because the next question would be, ah, okay, 20% more in PMAX. Yeah. But in which campaign? How? Totally.

SPEAKER_00:

Or what it's like, um actually I have open budget. My campaigns aren't spending more right now anyway.

SPEAKER_02:

Perfect. So maybe maybe we uh I think the this this last mile experts, like we are, and there are many others out there. I think uh that there has to be more synergies between the strategic marketing mix modeling and the last mile expertise. I think that's the way we have to go. However, it's an issue not talked about enough from my perspective. We can't solve it today, but we brought it up. Yes, we at least We did the right thing. We we brought problems, not solutions. It's so much easier to bring up problems. That's I mean I'm I love bringing problems. Yes, yes. But what I know is the solution of the problem is always the problem of tomorrow. So this is the issue, man.

SPEAKER_00:

I I don't know what you just said, Chris. You just you know, think about it. Think about it. Yes, I'm gonna think about that. I'm gonna think about that with a six-pack of beer, Chris. I'll tell you that. Uh we can we are we allowed to drink a beer in our Christmas episode? Um yeah, I mean, as long as we you know, company policy, I think we can't we're not supposed to drink before 4 p.m., right? So we have to take care of uh recording the episode after four.

SPEAKER_02:

Yeah, yeah.

SPEAKER_00:

And we're allowed to drink in terms of company rules. But I would not drink a beer for Christmas anyway, Chris. Spiked eggnog is the only proper. I'm only. Let's think about something special.

SPEAKER_02:

Yeah. And then we can talk about the coat. Yeah, yeah. The solution of the problem is the problem of tomorrow. Yeah, okay. Let's think about it.

SPEAKER_00:

I I think it's starting to click for me. But before we go further down that particular rabbit hole. Um, what's the next topic? Yeah. I've got something for you, Chris. Um and uh and it uh it teases a bit out of what we've just been talking about. We're talking about MMM. Something one of the things that MMM does typically offer, it will help you understand your marginal ROS per channel. Okay. Um and marginal ROS is basically the idea of like what will I get how if for the next dollar, euro, pound of ad spend, how much will I get back? And you can plot this and under you'll understand that you have a diminishing return curve, it'll flatten out. Um that's that's typically what marginal ROWS is used for. I love the concept, by the way. Aaron Powell Me too. It's very it's very important to know. Um and uh because you can always put uh actually now I'm I'm gonna steal a quote from a a guest from way back in the days, and I have to reach back in my mind to get her name. Um but she she said you can always put something in and get something back up. Yeah, but the relation. Yeah.

SPEAKER_02:

What before we move on, Mike, uh the the concept of marginal ROAS, um is it widely adopted within within the e-commerce world? And are question A, question B, and are online retailers really looking at that metric, try to understand, okay, I can get more revenue now, but I I buy this revenue at at a margin burning rate.

SPEAKER_00:

Um what what what's your take on this? I I think yes to an extent because um and and you know, here now we're referring again to platforms as a source of truth and stuff like that. If you look inside of Google, Google Ads, for example, you'll see like a keyword planner and um some other some other tools that do show this curve. Um and it It's Google's estimation of that. And I don't know exactly how they arrive at it. I don't think that they're like slapping a Bayesian algorithm on and putting priors in and stuff like that. It's probably closer to some sort of heuristic or it might be an average effect or something.

SPEAKER_02:

They got better at this, by the way. Yeah.

SPEAKER_00:

Michael, because why am why I'm asking this? But otherwise, sorry, otherwise you need MMM to really solve for this. And then now then if you're asking how adopted that is, then we're talking about the adoption of MMM. Aaron Powell Okay. Fair enough.

SPEAKER_02:

In general, the concept of marginal ROAS is something because why I'm bringing this up, it's because what what Google has been very, very good at, and we talked about this many times, P-Mex is is performing. So uh Nomenestoma, performance max. It's bringing your performance, it basically hits the RAWAS, the target rowers you put on campaign level. Mostly if there's enough numbers velocity, it's slightly beyond the ROAS target, so it's performing. However, I my guess would be that a lot of clients are happy by just hitting the ROAS goal. Right? A lot of online retailers are just, yeah, okay, my ROAS goal is eight. PMAX campaign is running at 8.3. Everything is A OK. Yeah. Because the margin on ROAS concept is not that visible. And that's I don't know if it's that important for the online retailer. As long as the average ROAS is hit. I think it's a massive flaw, by the way. Yeah.

SPEAKER_00:

But would you cosign that? Yeah, for sure. I mean, um, I I think in in practice, another way like people might look at their at their impression share and see if they can increase it if they're happy with that, if they want to increase it. But that there you'll see these effects exactly kick in because um you know you don't want to have a hundred percent impression share because you're far beyond like um at that point you're you're getting impressions that are no longer perfect for you. Yeah, exactly. And um, but this is finding the ideal impression share is actually kind of uh diminishing. I know what I mean. It's all related. I know, I know, okay. Yeah, and but that's right, people are typically looking at their average rose. Uh they you know, they want to see um they have a target in place and it's important that the campaign like daily fluctuations, of course, given. But and Google's promise is that they'll fulfill it uh uh on a on a kind of a monthly level. Um but you brought up marginal marginal ROS.

SPEAKER_02:

Yeah. Uh what's going on with Google?

SPEAKER_00:

Well, I'm bringing I'm bringing that up because Google has been bringing up marginal ROAS lately. They've taken an interest in it. Um their interpretation of it or what they want to achieve with it is somehow a bit different because they're not really talking about that curve. They're not they don't seem particularly interested in that. Their point is that um they're they're trying to build a story here about why cross-network advertising makes sense. So these are campaign types like performance max, demand gen. Um and actually almost every campaign at this point is cross-network or has the option to be cross-network in to some degree or another. And they have some, you know, pretty convincing, at least at face value, some pretty convincing argumentation here. They're talking about when we were complaining about channel silos earlier, they're saying that if you're looking at each ad network within Google as its own, like you know, you've got a search campaign for your search, you have shopping for shopping, display for display, and so on. Um, that within these campaigns, there could be valuable conversions lingering just beyond the borders of that little silo, and you're missing them.

SPEAKER_02:

Um meaning that my next conversion within my search campaign is let's say on CPA level, is more expensive than this incremental conversion with demand chain. Exactly. There could be a cheaper conversion somewhere else.

SPEAKER_00:

Within Google, of course. Yeah, exactly. And that you might hit um that you might hit a point where due to the constraints like your efficiency constraints, whether your Rose target or something, where you can't buy more conversions. And or if you have a CPA target for whatever the case might be.

SPEAKER_02:

So Mike, just to make it clear, because I think it's a fascinating example, and Google is very good at at bringing complex topics in a in a not simple, stupid, but simple substantial way to the audiences. But just to understand it, let's for the sake of simplicity, let's talk about CPA. I have a campaign running with a CPA target of let's say six. And Google and it's a search campaign. And Google is buying uh conversions within my CPA target of six. Yeah. And then there might be a conversion, let's say the last conversion with my incremental dollar ad spent where the CPA would be at seven. Exactly. Meaning that conversion is not bought.

SPEAKER_00:

Yes.

SPEAKER_02:

But there would be a conversion at the CPA of four within shopping.

SPEAKER_00:

Exactly. Yeah, and they have a table uh they compare it to fishing in a bunch of little ponds instead of fishing in the ocean or something like that. And they and they have a table similar to what you've said where it's like, okay, I'm buying um a conversion on search at a buck, at two bucks, three bucks, four bucks. I'm always buying the next cheapest one, uh, but they're it's getting more expensive. Um whereas if I would be using cross network, I'm finding one at a buck on search, a buck on shopping, a buck on display, et cetera. And so and they show that in principle you can end up getting more um from this blend.

SPEAKER_02:

There there must be an issue, man. No, come on. Otherwise, I wouldn't be talking about it. Let's talk about it.

SPEAKER_00:

I told you I bring problems, Chris. I'm a problem bringer. Yes, let's go with it. Today's the the the episodes of problems, not solutions. Yeah, that should be my LinkedIn tag line. I I bring problems. I strive in problems. I like to make everything everything muddy and complicated and then and then walk away. What is the problem, man? So I mean, so first off, I've yeah, then so they're saying that PMAX, that the bidding logic works like this, that it's going to find you the next cheapest conversion, wherever that might be. And um that's that's for sure the way that I've understood Meta's advertising to work for years. They've always said that it basically works like that. That is not quite the way I've understood Google to work. Google, my understanding, has been more like there is they're trying to reach your efficiency on average. Exactly. And so every single conversion, they're they're making a prediction and you know, it might be a little above or a little below, but ultimately this thing's gonna average out. Um and not that they are trying to always find the next cheapest conversion. And um, you know, also like when PMAX was launched, I I think we maybe talked about it on here before. I I've I've mentioned this a few times in uh various places. Like certainly my understanding was that PMAX was full funnel or full funnel-ish. Um and that's why it had display in there and all these other things. Yeah. But actually, as we found out, it's very bottom funnel. And that's not the sense it it's actually more remarketing, like you mentioned, just going on in these things. Trevor Burrus, Jr.: So we were talking about shopping share of a PMAX campaign, even with assets, is 90 percent. Well, that's the feed-the feed-based share is definitely 90 percent. But well, but what we see now with the channel performance report is that it that's that's not all there's a lot of YouTube feed-based ads and a lot of feed-based di- um, you know, and a feed-a feed-based Google display ad is basically dynamic remarkable.

SPEAKER_02:

But it's it's rather bottom funnel. Yeah. That's what we know. Yeah, exactly. But uh back to your point. So it's full funnel is yeah.

SPEAKER_00:

So that that my understanding was that it was full funnel, and they used to talk a lot about this very that these channels were playing together like teammates and kind of this multi-touch attribution thing. They talked back then a lot about micro moments and these other nauseating buzzwords, but basically, right ad, right moment for the right user. And we'll find that user on any of these channels, and they're all supporting each other. And it and if and but let's okay, so the bot the full funnel stuff is gone. But also this this this sort of teamwork or multi-touch attribution to me is a bit gone because they're actually sort of describing that these channels in a way are in competition with each other a bit. And maybe it's an oversimplification, but it's the way that they explain it. Because actually PMAX is going to bid on wherever the next cheapest conversion can be bought. And so does the market chain. Yeah, exactly. And and okay, like fair fair enough, I guess, but they also they also tell us to go easy on them when we look at the PMAX channel reporting that they've made available. Because they're saying like you should understand when you're looking at the the ROWS, for example, of one of these these other channels like Display or YouTube or whatever, um that's not telling the full picture.

SPEAKER_02:

Yeah, not explaining the full picture.

SPEAKER_00:

You'll see the average ROWS of that channel. But that could give you the wrong conclusion because it's not talking about the average ROWS, it's talking about the marginal. So those campaigns were stepping in when no other conversion could have been bought cheaper or stuff like this. But actually, the ROAS on these is i is often quite bad. And I I'm just getting confused here, Chris, because um if it's ri if it's really this sort of competition of like the where's the cheapest conversion out there in the Google ecosystem, then the segmented view on the ROAS per channel should should be holy grail, right? Yeah, they exactly. They they should the the performance should basically hold up there. For sure.

SPEAKER_02:

But Google is saying uh marginal ROAS, yes, but the segmented view on it's it's it's not that important. It's not telling the full picture.

SPEAKER_00:

Like if if if this multi-touch attribution thing would be in place, I could get behind it because all right, the the performance of this channel doesn't look good, but that's because it's assisting something else. Yeah, exactly. But they're saying no, it's not really full funnel, it's not really multi-touch, it it's a bit different than that. And at least as I interpret it, and maybe Google can tell me that I have it all wrong. They will, for sure.

SPEAKER_02:

But the question is uh who we should trust more. I mean, I I for sure uh I mean what what I can tell you, Mark, is I I think it's it's very fascinating. And again, we we give Google sometimes shit where it's due, but uh, I think Google has done a lot of right things in the last couple of months. We've talked about this many, many times. Uh so shout out to Google. I think they're they're on the right track again uh from the perspective of an online marketing expert. You can do a lot of things again and and and you control, more insights and so forth. I want to put that on Front Street, Front Street. And this marginal ROAS idea is 100% the right one. We have been preaching about this for the last 10 years, man. Can you remember when uh the average ROAS is actually not the most advanced way to optimize your campaigns, is about understanding the marginal ROAS. So I think I love it. But again, your point, the problem you brought up is is Google contradicting itself right there.

SPEAKER_00:

Yeah. Yeah. Yeah, it's a total like in their in their in you know, in their example, there's they're saying like they look at this single channel where you're getting escalating costs, and even then at a point your your volume gets limited. And they propose this alternative view where basically it's the best of both worlds. You're you're getting more conversion value or more revenue, more conversions, and the efficiency is is better too. But if that would be the case, then it shouldn't be a problem to segment it. Because it actually look it actually looks like um these channels are are stepping in at a point where you're pretty saturated. They're stepping at a pretty saturated point on the curve. And actually the question arises whether it should have bid for that conversion at all. Or if it should have, you know, if if the high like what are the likelihood? I'm sure that there are some of some conversions on these channels that perform quite well. They're probably existing customers, they're probably dynamic remarketing, stuff like this. Um that's why they perform so well. Uh but I if that's not the case, I mean, I I I just to me, the whole thing doesn't really hold water. I'm not sure what the children's like.

SPEAKER_02:

No, I know what I mean. Uh uh from my per perspective, again, I'm I'm not I'm I'm not that deeply into this whole thing, but my commercial understanding would be if this marginal ROAS thing is is a real thing, by the way, I I would love it because I I think that's the right way to look at every dollar spent you you're willing to spend. Fair fair enough. But then the segmented view of each and every channel you're active in, and the numbers on in in that segmented view uh should should should be taken up with utmost seriousness. And Google is claiming don't look at that too much in detail because it's not telling you. So there is the contradiction. Yeah. Um however, let's let's see how this plays out. My question, Mark, would be how this marginal uh ROS thing, how how to apply it? Well, is it like an optimization tactic or what or is it just a reporting functionality?

SPEAKER_00:

No, I mean they don't there's there's not much you can do about it. They're letting they're letting you know that that is their bidding logic. Okay, got it. Exactly. And um and actually I would find I would be much more convinced and kind of more satisfied to hear that it's this multi-touch thing that's going on. It would make a lot more sense to me based on what I see. But uh no, there's nothing particularly actionable that you can get out of that. Um it's supposed to be their assurance that the system is functioning, that there's a benefit in cross-network. There could be a benefit in cross-network. Um, but I think that it it just looks to me, when I look at the numbers, it looks like these other channels appear to be far overshooting any like they've gone beyond on that curve where they should have gone. And so it's funny for me to for Google to raise the topic of this curve because I I don't think that it's being applied correctly. Um so I don't know what's going on if it's just, you know, they've because you have to imagine how complicated this auction environment is. And this idea of like the next cheapest conversion, it it sort of implies that like you're buying them all through the day, but actually there's all kinds of inter intraday seasonality. And it it must be orders of magnitude more complex. And so maybe it's a failure to communicate. It's a hell of a claim.

SPEAKER_02:

It's a hell of a claim. I mean, if like I said, if if they can execute on it, then again the question is how can you control it as a as an online retailer? But it's a hell of a claim. And again, if if that claim turns out to be true, I think the segmented view on your channel performance is is is the one thing which which should be highly, highly relevant. Yeah. Right? Because that that's where you probably understand that the best and most sophisticated way, what this incremental conversion actually costs you, or what the incremental performance of each and every channel is. Uh yeah, that's certainly contradicting. Um let's see. I I think the the the message. I mean, we read the same article. Uh great job here because it's it's simple, it's powerful.

SPEAKER_00:

Yeah. I I think if you spend a couple of minutes reading through it, it'll it's very convincing. But it's very convincing. And maybe it's less paragraph. It it just feels a little bit weird. But I think we'll learn more about that soon, and hopefully we can cover this on a on an upcoming episode because um they will be right now. You you have to go into single accounts to see this channel performance. And um, you will now be able to see this at the shell level, the MCC level. So just to uh to explain that, we have a lot of accounts. We have data for a lot of accounts, and this will allow us to see that from one place, and then we'll be able to look at much more, you know, we can find trends, we can find the fun stuff when you aggregate data. Um and we'll see, you know, what is the range of outcomes, what is typical. Because right now it's all spot checking. Of course, you can't get a picture. Yeah. Of course. Yeah.

SPEAKER_02:

Mike, um, because man, time time flies again, not just not just with regards to this episode, but with regards to the whole year. We have now two episodes left.

SPEAKER_00:

Exactly. We're next, next will be we're gonna do a Black Friday recap. This will be a big one. Yeah. Love it.

SPEAKER_02:

So you will bring bring some numbers. I will bring the data. And then we have a Christmas episode.

SPEAKER_00:

Yeah, which we've been teasing for a while now because we're very festive.

SPEAKER_02:

We have festive, we love Christmas, and um we'll show up in some festive outfits.

SPEAKER_00:

Yeah, let's leave it at that. I know. Chris is playing, he's playing shy here. He knows exactly, he has a plan. Uh we will uh open the marketing budget for for uh for our festive outfits. Exactly. And um and let me just say that episode will be forecast to air on December 23rd, just in time. Perfect. Yeah, I'm looking forward to the reason I'm so excited, Chris. I have a I have a I love Christmas podcast. I have a I have a playlist just of Christmas episodes of my favorite podcast. I love it.

SPEAKER_02:

I love everything around Christmas, man. Now I'm I'm you know I'm I'm extremely into the esports, and there the US is doing a lot of things the right way, some things not the right way. It's always a question of perspective. Where they are clearly the best that is uh the the decoration of the studios, man. They they sell it's it it just brings the positive vibes. I love it. Yeah. Watching NFL on Sunday, now with it's it's it's a pleasure.

SPEAKER_00:

Uh the Austrians get it too, man. There's nothing that beats an Austrian Christmas market. And you're right. Yeah, I was why I had to go to the US Embassy a while back, and so I think it was like November 1st or something. I walked past the Ritz Carlton on my way in Vienna. In Vienna. And uh It's all it. Yeah, it was it was November 1st, and they had it looked the Christmas decorations. I was like, guys, it's November 1st. Yeah. David East. You know, they went full Mariah. Man, they went full Mariah on that thing.

SPEAKER_01:

Yeah.

SPEAKER_00:

All right, man.

SPEAKER_02:

So we we will try to, you know, I don't know, bring the same energy.

SPEAKER_00:

Oh, it's called We too, dear listeners, shall go full Mariah. And if that's not your thing, too bad. We can't, yeah, too bad. It's not our fault. Time for the intro. So thank you very much. This has been another episode of Growing E-Commerce, brought to you as always by Smarter E-Commerce, also known as Smack. You can learn more at smarter-ecommerce.com. And again, if you enjoy this cut podcast, please do share. Give us um, you know, a follow, uh a like, uh, a review. We really appreciate it. We appreciate it. Every single one of them.

SPEAKER_02:

Mike was a pleasure.

SPEAKER_00:

Likewise.

SPEAKER_02:

Thanks a lot, sir.