Mike shares updates about two trending areas to watch this year. First is the rise of marketing mix modelling. He digs into Google's new Meridian model, including a refresher on Marketing Mix Modelling, a reminder of Meta's competing offer (Robyn), and Meridian's USPs. Then he shares a big change in the marketing activity of fan favorite Temu. We'll leave you in suspense on that one – listen to find out!
Discussed in this episode:
A recording of Mike's webinar, "What's IN and what's OUT in 2024 ecommerce?"
https://www.youtube.com/watch?v=8HF27Tjm4po
Mike shares updates about two trending areas to watch this year. First is the rise of marketing mix modelling. He digs into Google's new Meridian model, including a refresher on Marketing Mix Modelling, a reminder of Meta's competing offer (Robyn), and Meridian's USPs. Then he shares a big change in the marketing activity of fan favorite Temu. We'll leave you in suspense on that one – listen to find out!
Discussed in this episode:
A recording of Mike's webinar, "What's IN and what's OUT in 2024 ecommerce?"
https://www.youtube.com/watch?v=8HF27Tjm4po
Welcome to Growing Ecommerce. I'm your host, mike Ryan of Smarter Ecommerce, also known as Banking. Today I'm sharing updates about two trending areas I'm watching this year. One is the intersection of open source technology from ad platforms and the rise of marketing mix modeling. That's not as boring as it sounds. We'll dig into Google's new Meridian model together. And the other is a change in the marketing activity of fan favorite Temu. No wait, temu, temu. Anyway, I'll leave you in suspense about what that change is. Remember, if you enjoy this podcast, please leave a review and share with a friend, a co-worker or on social media. We really appreciate it. Okay, let's get into it.
Speaker 1:So today I want to reflect back on a couple of topics that I didn't cover on this podcast yet are not in this form. Basically, in January, very close to the start of the year, I hosted a webinar talking about what's in and what's out in 2024 at e-commerce and intended to be like a higher level kind of CMO's guide to some trends that are. Well, I think there are no trends that are really contained to a year typically, while these things, the seeds were planted long ago and we might see more leaves kind of sprouting on the plant in a given year. I find those you know this year in trends, those kind of reports, a bit artificial. But I like this what's in and what's out format. So I talked about some topics and, yeah, now we're most of the way through quarter one. The quarter's getting right now actually January somehow feels like a long time ago.
Speaker 1:So I want to follow up on a couple of those topics. The first one I want to discuss marketing mix modeling and what Google is doing there right now with their new meridian model. And the second one we just had a deep dive into Teemu, or, as I've said, as I learned since the Super Bowl, I guess it's Temu. I am always going to say Teemu or Teemu, but we learn more about Temu. And but I want to dive specifically into some changes in their advertising strategy lately. So that's something that's very new, all right, well, let's start with Google meridian. And just to recap, so this, this webinar that I hosted. It's available on YouTube. You can go watch it. If you search for, like, what's in, what's out, e commerce, tack on Mike Ryan, her smarter e commerce you'll probably find find it on YouTube and we can include the note in the scheme of the link in the show notes, but there were actually.
Speaker 1:There were sort of a confluence of trends here that I think is well captured. With meridian, I was discussing how marketing mix modeling or media mix modeling, mmm, is on a continued uptrend. You know, again, it's nothing, that's nothing new exactly. It's been trending upwards for a couple of years now. It's sort of obvious, in a way, when you see what's going on with cookie deprecation and privacy. But before we get into that, there was an it's into that a little further.
Speaker 1:There was another trend, which is that a lot of the really interesting stuff happening right now from Google's product team or solutions team is not what's happening in the app platforms themselves. They're actually publishing tons of open source yeah, little tools and pieces of software and pieces of code in their GitHub repositories. They'll publish things, for example, that tackle customer lifetime value. They have one called crystal value or they have one that's called so Tara, if I remember right. That's about profit bidding and there's really all kinds of interesting use use cases in there. Excuse me, I just looked, and that's so terrier or so, to Rhea, I don't know how to pronounce it, to be honest, but they also they have, like Google match markets. This is designed for geo testing and geo experiments. We've been using that for a while. They have lightweight MMM. That's a way to easily train Bayesian MMM and get channel attribution information.
Speaker 1:So you know, there's been this for 18 months, I would say, and it's kind of increasing this tendency for Google to publish more and more tools on GitHub, and I find that really interesting for a couple of reasons. You know, it's sort of a counter narrative or counter tendency toward what's happening in the main ad platform, where we've seen promotional controls get reduced, we've seen bidding taken over by algorithm, we've seen a lot of like kind of less segmentation in the campaigns. These, these factors coalesce together to basically create a situation where it's harder to differentiate your campaigns, it's harder to stand out, it's more of a one-size-fits-all way of managing campaigns. We see that Performance Max not the first time discussed on this podcast it's the epitome of this simpler and less faceted, less buttons and dials kind of approach to campaigns. But then you've got this yeah, as I mentioned, this fascinating counter example in GitHub where there are lots of new capabilities being added, just everything from image asset management to robust A-B testing for your shopping feeds. They're putting in a lot of stuff in there and I think that a way if you want to differentiate these days, a way to do that that you can't do in the ad platform itself, is on the measurement side and optimization side. There are these very interesting pieces of code out there that you can run with. Now the thing is, this won't help your average SMB. It won't help many teams. You do need to have some kind of technical resources or a partner to help you with this stuff, and so it does kind of it's one of these de-democratizing things or something that is keeping that playing field from getting perfectly level.
Speaker 1:We can only speculate as to why Google publishes these things. I think partly they do it to keep their developers engaged and to have. Developers often want to build cool stuff and show off cool stuff, and I think it's a way that they can work on interesting projects and have these public portfolios. I think that's one reason why just on an employer brand and standpoint. But if we bring that around to the topic here of Meridian, these other things, they're not official products, they're not really production quality, they're not totally signed off by Google, like Google makes them available but they're not really marketed by Google. They don't have that full backing, and Meridian is something a little bit different, because this is something that is getting their full backing. I mentioned that they have a package called LightweightMMM and this is kind of an extension of that. That might have been. You might think of that as a prototype of Meridian.
Speaker 1:I don't know, maybe the development history is completely separate between those two, but it's actually not that novel. It's actually kind of following what we've already seen from Meta for two or three years ago I think two or three years now Meta published Robin, which is an open source MMM, and MMM's marketing makes miles there. There's things that can be different about them. They can, of course, be differentiated, but there's also a lot of things that they'll have in common and so in some ways you could view Google as two or three years late to the party here. They're definitely in a follower position behind Meta and I think one of the reasons that they published new pieces of software like this I think there is a demand in the market for them, of course, as people are looking for new measurement solutions. But also, if Meta's Robin is your source of truth for marketing mix modeling, then Google is not and it's kind of a threat toward Google Analytics. It's kind of a threat toward their position to be the arbiter of truth, to be that single source of truth. So from that standpoint it totally makes sense that they want to go out of here. They don't want some third party piece of software like recast or I don't know. There's a lot of marketing analytics tools out there that cover this kind of need, and then the big products from platforms like Meta. They don't want those people setting the tone. So I think they're going to be able to do that. So, as I mentioned, there's a lot of things that are basically just common to marketing.
Speaker 1:Mixed modeling Like this is basically going to help you understand the saturation of your channels. Basically, the optimal spend level is going to help you understand your ad stock decay. This is using a Bayesian model, like most MLM these days. But let's just unpack those things a little bit quickly. Let's just start with Bayesian why not? What does that mean exactly? It's basically a statistical approach to running lots and lots of simulations and finding more or less the most probable one, and you can insert information that you already have. You would call these things priors, like if you have previous information about incrementality testing, lift studies, if you have previous information about ROI measurement that you've done, you can integrate that in. You can also easily integrate this with future lift testing and stuff like that. So it's a way it's a bit of a more living model, you could say, and it sort of represents one of the main or larger innovations or recent changes that has helped bring marketing mixed modeling back into fashion. Besides the necessity for it, that's coming from a loss of measurement in other places.
Speaker 1:And again, the goal of this just I mentioned channel saturation, I mentioned ad stock decay. So just to get into those, like you can understand that sort of dose response curve between the budget in your that you're putting into a channel and the revenue that's going to come out of that. You can understand the shape of that curve and how these different channels compared to each other. So that can help you decide, like what is the right spend level for a given channel? Am I overspending on meta? Am I under spending in Google ads? Should I throw some more money at YouTube? And of course Google has an interest in answering these questions because this is getting to the heart of what CMOs do. This is getting toward what CFOs do in terms of budget allocation and deciding which channels are going to get spent on. Now you could easily at that point get conspiratorial and say, well, does Google kind of self favor themselves a bit? And I don't think so, because it's open source. You can inspect every line of code. I don't know, maybe they could hide something away in there, but I think it's risky business to do that when the code is available for inspection.
Speaker 1:But we will talk about a couple of USP's unique selling propositions that they try to have here. I just want to mention that other thing first ad stock decay. So this is another kind of key thing that you can learn from marketing. Mix. Modeling would be ad stock decay just refers to the lag defect of advertising or another way to look at sort of the half-life of advertising, like after there's been an impression, how long does that count for, how long does that have some kind of lift? And this can then help you understand what is right frequency, so that you're not just kind of guessing. So you know there's questions here that people of course want to know the answer to like are these channels incremental? Where can I spend my next best dollar or euro or pound? How often should I be showing my ads, which? How are these channels really comparing to each other in an unbiased way? So these are the kind of questions that you try to answer there.
Speaker 1:And in regards to what makes the meridian model special, yeah, I think there is a level of robustness and investment that Google can bring to a project like that that maybe a third party piece of software can't. That's probably debatable hot take, I don't know. But two big things that they focus on. They want this to be kind of definitive in regard to paid search measurement, and I think it's obvious why they want that. But they talk about search volume as being this kind of confounding variable within marketing mix modeling, that there are these fluctuations in search volume that other marketing mix models are not aware of and Google has access to that data and can incorporate it better than anyone else. I'm a touch skeptical about this, like, their kind of argument is that search volume is a very close proxy for demand or consumer demand, and that's not necessarily the case.
Speaker 1:I'll share a short anecdote with you from a while back. I can't say who I do this analysis for, but it was one of the one of those large shoe brands, shoe and streetwear brands out there, so you can fill in the blanks mentally and I'll add it us, nike, puma I'll let you make your own assumption there. But it was really interesting because I looked at the relationship between search volume and other metrics, kpis like conversion rate, for example. So that brand, because I had access to their to not just to the search volume data but to their performance data, so this was possible. And that brand they saw three major spikes during the year One in March, one in August and then again in November. So it's pretty obvious what was going on with these. March, this was in the US. March was March madness, of course, when the then be excuse me NC double a is going strong college basketball in the US. And then August is back to school, shopping season One, of course, sportswear and streetwear are going to. There's going to be an increase in demand there. And then November, that's obviously the holiday season. Now, as I said, there were big spikes in branded search volume at each of those points, but only two of them actually correlated with increased conversion rate and with increased conversion volume, and those were August back to school and November for holiday shopping. The March madness period saw increases, but there was at least no obvious intent there that we couldn't see that being expressed in the campaigns. That's just a big example I think that there are.
Speaker 1:For any given category or brands, there's going to be fluctuation in search volume and sometimes it's going to be meaningful and sometimes it's not. It really depends on the quality of their implementation. But they've been surfacing more about search volume in their campaign insights for a while now and I've always been a bit skeptical of that because there's really not always a one-to-one line between those things. Now what's to me slightly more interesting is YouTube. They of course, have access to that reach and frequency data. The model works with or without reach and frequency data, but it's more powerful when it's in there and it might be possible to furnish other models with this data, but that should work more out of the box in the case of Meridian. So these are kind of their USPs is that they can better model paid search based on basically factoring in demand using search volume as a proxy, and that they've got a very good grasp on YouTube reach and frequency.
Speaker 1:So I would expect this model to be flattering towards YouTube. It'll likely look more flattering than just the numbers in and of themselves, which is not necessarily a bad thing. Those are the kind of campaigns where the contribution can be tougher to measure. It can be more inherently not invisible but intransparent, and it's great if they can shed some light on that. I think it'll also be like we've looked at Pmax with Meta's Robin that I mentioned. We've looked at Performance Max using that before and it was flattering toward Performance Max as well. I think that these campaigns will look generally good, but of course, the other side is a double-edged sword for Google, like people can also determine that they're overspending on these things as well. But generally, I think Meridian is very cool. It's very customizable. I think it's another thing that, similarly, I mentioned these GitHub tools from Google that are more valuable toward larger teams with technical resources or with technical consultant partners, and it'll be the same for marketing mix modeling. You need a data scientist or an analyst, you need someone experienced in the mix. It doesn't just work out of the box. So it does actually continue this tendency of the haves and have nots.
Speaker 1:I want to just mention, by the way, one thing from the press release that was pretty eyebrow-raising as a last note before I move on and credit for this goes to Harpal Singh. On Twitter, he noticed that they claim that 60% of advertisers are using MMM in the US, and this is absolutely nonsense. That is definitely not true. That is far overstated. I don't know the percentage, but it's for sure not. I mean, that's almost 2 thirds of advertisers, more than half. That's definitely not the case. Google's research partner there was Kentar, and I don't know but motivations there might be for that. Maybe I'm wrong here. I don't think so, though, but maybe they just want to use as big a number as possible to kind of motivate others, give them some FOMO, that they should be using MMM too. If they genuinely believe that, then I worry for them, because then I think they're really over-sizing the market opportunity from their standpoint here, and they're probably going to be disappointed when 60% of their advertisers are definitely not using this in 12 months from now. But I don't want to be so hard on Google all the time here. I think this is a cool tool they built. I think they're building a lot of cool tools in GitHub, so do check those out, and then we'll move on to the next topic, which is a change in the marketing activity of Temu.
Speaker 1:I still can't. It doesn't feel right saying Temu, oh man. So I actually considered that last podcast episode like two episodes back about a month ago with Stefan Vensel have a listen, he's so smart. But I consider that kind of the capstone of my work on Temu for now, because I've been following Temu since last June, shortly after they launched here in Europe. They've been in the US a bit longer, publishing data and just thinking about them a lot and I actually started to get kind of I was starting to get a little over the topic in a way. But interestingly, interest to hold on that phrase doesn't work. But interestingly, it seems like everyone else was just getting started.
Speaker 1:I ended up I'm really flattered by this. I mean I ended up speaking with Modern Retail and the Financial Times and Forbes and the Wall Street Journal and it's insane, all these people about Tim was advertising. But I just felt that the story is okay, they're spending a lot. I think their business model is interesting to watch. I think it is a suspenseful story if this can be maintained. But I said, okay, I'm not looking at this anymore until something in the picture changes. And then something changed. So here we are again.
Speaker 1:Yeah, I don't, I can't directly tell you what Temu is spending on Google ads or Meta or something like that. I can only look at proxies that show how aggressively they are marketing or how much are marketing, and the main way that I'm looking at that is custom metrics. That I call competitive prevalence and, by the way, temu was another one of the themes in the webinar, so check that out on YouTube. But competitive prevalence basically works like this. It's nothing fancy. We have access to over 800 accounts in one of our Google ads MCC's. There's more in the other one, but I just I just look at one at a time and we can. What we can then do is find out how many of these accounts face competition from Temu at that account level at a very high level. It can also drill down deeper, but for how many of them is Temu a notable competitor account wide.
Speaker 1:So I saw this number peaking in December at around 80%, which is very comparable to where Amazon is. Just for reference, I saw a similar number published from my friends at TNUITY in the US Love that team, andy Taylor, mark Ballard from the research team and they had a number at, I think, 90% in the US, which is, yeah, stunning, and I think there could be the discrepancy between the 80% and the 90%. It could be a market factor, or it could be about some difference in the filtering or methodology we use behind the scenes. I purposely made mine a touch more conservative, so that could be why. But in case your highs are glazing over, put dead simple eight out of 10 advertisers were facing notable competitive pressure from Temu in their Google Ads auctions, and that started that was at zero here in Europe back in April, so this was within six months. They got to this level. That's nearly unprecedented.
Speaker 1:We saw a similar movement with Wish back in the days, and go back and listen to that podcast episode with Stefan Wenzel where he explains that team Temu and Wish are actually not that comparable. They have different business models. I learned something out of that myself. And then I'll just tell you one more thing. I was speaking with a friend. I phoned a friend who I'm not going to talk about this data in quite as much detail. It was not mine and I don't want to share more than what I'm supposed to but this person is collecting the number of active Temu campaigns from the Meta Ads library and historicizing that on a daily basis.
Speaker 1:And here we get to the interesting part we can get out of this context and get to the interesting part. There was a notable drop on both of these kind of proxies of marketing activity. There was a notable drop in February with Temu. So for Meta that meant that their number of daily active campaigns, which had peaked in late December or early January, rapidly collapsed, and it rebounded a bit since then and then with Google. That means that eight in 10 advertisers was seeing competition. Then it rapidly collapsed to six in 10 advertisers Again just right before the Super Bowl, and it has also rebounded a bit to about seven in 10. But they lost about a quarter's worth of progress in a very short time. These numbers were comparable to where they were at in September of last year and I don't know if this is related to the Super Bowl.
Speaker 1:If they wanted to kind of purposely wind down their advertising activity a little bit so that there'd be less noise for the Super Bowl and they could kind of gauge the effect of the Super Bowl more, this sounds to me like way too advanced for what Temu was to be. Parents, just spending, spending, spending. They blasted through the late Christmas or late holiday period with ad spend, even though it was totally unrealistic that any of their packages would arrive in time. They've done nothing but go up and to the right in terms of these marketing activity proxies and, from what we can tell their actual spend, they've done nothing but increase, increase, increase over time. So this to me, marks a really interesting possible inflection point. I don't know, it could be. There are different reasons for this. They could be just wanting to spend less. They might have reached us a saturation point. They're deciding that they're going to spend less now. They might be expecting more efficiency from their campaigns and that might have kind of the automatic effect of giving out less budget. They might be trying to narrow which products they're focusing on.
Speaker 1:I mean, for Google, I split this out per retail category and the trend was pretty similar between all categories. The ones I looked, I think Health and Beauty took a bit more of a hit. Oh, and hold on, I'll show. I'll just tell you one last thing. This isn't just these two proxy metrics. There's one other view of this data, which is that in Google Merchant Center you can view the businesses with the highest visibility per country and category, and they won't tell you yeah, really the numbers. But it's a ranked list and you can filter for either ads or organic visibility or both of those combined.
Speaker 1:But I looked at Teamoo, filtered for their paid visibility and they lost. They lost rank as well. So this was, you know, they lost rank in February compared to in previous months. So, let's say, like in Germany fashion, for example, they were the number one in terms of paid visibility, which is a good, a very close proxy for Spenda, I would say. They were the number one ahead of even a local specialist marketplace like Tlando, ahead of Sheehan, ahead of Amazon, and now they dropped down to a number three position. Now that could be that the other advertisers spent more, started spending more, and Teamoo spent the same.
Speaker 1:But when you combine it with this other view that these other marketing proxies are slipping, I feel confident saying that Teamoo is slowing things down a bit, taking stock. I don't know, and we'll see how long that trend continues, but I've looked at this through the first week of March and they're just at a different level right now than they were in December or January. To just talk quickly about the broader implications of that, you know this could maybe be bad for the stock prices of these platforms like Meta and Alphabet, google's parent. I don't know that, but we do know that they're the number one spender on Meta and they're the top five for Google. So when you're number one or your top five customers start spending less, definitely something to keep an eye on, definitely could impact your revenues. On the other hand, I do think that the amount that they're spending now, even if they've slowed down a bit, it's still much higher than it was a year ago. So in year over year terms, in that kind of comparison, it looks fine. It could be in like a quarter over quarter comparison, it might look less favorable and even broader than that, let's assume. I think it's a fair assumption.
Speaker 1:We hear like the CEO of Etsy was very outspoken about this. We hear that their activity is so strong and so aggressive that it's forcing the budgets and the bids, the unit costs of all advertisers. It's just making these channels more expensive. So they're kind of raising the tide, so to say. So it could have like a bigger secondary effect. When they start spending less. It might lower that tide a little bit, which is great for advertisers, not quite as good for Google.
Speaker 1:On meta, but I'm generally like I mean, there's nothing new about what team was doing in this regard. They have some business model innovation, but in terms of spamming these channels with budget, it's not the first time that we've seen this. I think Amazon pioneered it to this kind of extent and scope here in Europe. That was back around 2018. I don't know if it was exactly the same in North America, but that's when they really turned on the gas in like Q4 of 2018. And the reason why they're just playing a different game than every other advertiser. Their goal isn't to just like convert web traffic or make a transaction. Their goal is to poach a fish out of Google's pond and put it into Amazon's pond. Their goal is prime subscriptions. Their goal is we know our website has great AOV effects. We know we have good customer lifetime effects, so we can spend. Like the math looks totally different from everyone else. Basically, they were Trojan horsing Google ads. It's like they're a big competitor here.
Speaker 1:For product search or for people searching for products online is Google. They want those people to be searching for products on Amazon. So you just spam the options with your ads and you start pulling people out of those Google searches and onto your other ecosystem. So Tmoo is more of the same in that regard and I get a little concerned for the future of these channels, thinking that others are doing the same. Like we've recently seen that TikTok shop is starting to have ads on Google and it's an absolutely terrible customer experience. It doesn't make a lot of sense from the user journey because it's not like you just click the ad and you're on the landing page right ago. Like you need to make that purchase in TikTok shop in an app and shop the campaigns. Give me the app from Shopify. They're also going to start advertising on Google ads and meta. These are these potentially deep pocketed. We'll see what kind of an investment they make. I'm not saying they'll make a Tmoo sized investment, but they're potentially deep pocketed. Marketplaces and platforms and apps All these Amazon and Timu and shop and TikTok shop they're not like just an e-commerce site or something like that.
Speaker 1:They're just looking to grab people into their ecosystems and kind of divert traffic and they're willing to spend nearly anything to do that. And that makes it very hard when you're just an old advertiser. So, excuse me. So I would be glad if Timu backs off a little bit. Yes, I don't, I'm not a fan of what's been going on there, but I think that's really what I wanted to share with you today. I wanted to let you know that Google has the Meridian model available now for MMM. I want to let you know that there's been a change in Timu's spend levels in February and, so far, in early March as well, and I don't think it's seasonality and I think it's they're reducing their commitment a bit.
Speaker 1:And I also want to remind you to check out my webinar what's In and what's Out Trends for e-commerce in 2024 on YouTube, because the cool thing about that is I talk about a lot more trends very fast. It's not as wordy as his podcast is. I run through them as quick as I possibly can, so there's a lot of information all in one place. And, yeah, I want to thank you very much for listening. Let's scrap that, because I'll actually thank them for listening in the outro. So I want to encourage you to head over to YouTube and check that out. And if you're using Metas, robin or Google's Meridian and you have experiences to share, reach out to me. Or if you're seeing a difference in what's going on with Timu, reach out to me. I'd love to hear from you. Thanks for listening to Growing Ecommerce and if you enjoyed this podcast, please consider sharing it with coworkers, friends or within your professional network. We really appreciate it. This podcast is produced by Smarter Ecommerce, so known as Mech. To learn more, visit Smarter-Ecommercecom.