Loading summary
Transcript61 lines
- [00:09]
A
Welcome to the Big Story, a roundtable featuring members of the Ad Exchanger editorial team. Every week we bring you an in depth discussion of key developments in digital marketing and media.
- [00:24]
B
You're a marketer, you've got a CFO breathing down your back and, and you need to pick a marketing mix, modeling or MMM system to show the power of your media. So what do you pick? When you're evaluating different solutions, you may be drawn to a free open source tech like Google's Meridian or Meta's Robin. These two big tech companies have been building MMM tools in recent years. And our senior editor James Hersher has dug in and found out everything you need to know about what Google and Meta are building and how they're positioning their MMM solutions. And we will also tap into why marketers are focusing on mmm, an old school measurement tactic that is decidedly back in style. I'm Sarah Sluice, our Editorial director. With me today is James Hersher and our Allison Schiff, our newly promoted editor in chief.
- [01:15]
A
Thanks Sarah.
- [01:16]
B
And also our longtime metabeat reporter, which is one reason why you're here. So before we dive in, let's talk about programmatic I.O. new York City. This conference is the established place to talk about disruption. We've seen it before. Ad tech upending the media ecosystem. And this year you can attend to surf the wave of all this new tech and not get sucked under the surge. As always, our listeners can use the code POD10 to shave an extra 10% off your ticket price. So lock in your plans for September 28th and 29th and we on this podcast will all see you there and say hi. So James, let's just start at the basics. Mmm. I would say there's not even agreement on what the acronym stands for. It's Marketing Mix modeling or Media Mix Modeling, depending on who you ask. So what is it and why are Google and Meta building MMM tools?
- [02:08]
C
So to try and keep it very brief, I mean. Mmm. I prefer marketing because that's just, just broader. Like media, I guess in theory is you're looking at media and marketing is sort of everything.
- [02:19]
B
Okay. That's the new term that we want to lock in on.
- [02:21]
C
Yeah, that's why I sort of go with marketing mix. But, and, and it's like it's an old fashioned way of doing measurement where you sort of, you know, it's a, you could do it with the information you have from TV campaigns and like all the kind of, you know, historical data. And what it's really known for is a looking at media and like A channel basis, like it looks at your TV and your radio and your print. Like that would be, you know, historically. And also it takes a long time so it doesn't really fit in. It definitely like fell out of vogue like, you know, when sort of multi touch attribution, all the online stuff came into play and has come really back around in the past five or so years in a big way probably. Probably because just the data is going away. So to do sophisticated measurement, you're sort of falling back to this. You no longer can do real like user level tracking and all that and
- [03:23]
B
just to kind of maybe interrupt on the timing piece. So historically, MMM was done yearly. So it was like, okay, we got the reports back, like now we can plan TV for, for next year. And I feel like even when I talk to marketers that are like on the innovation end of the spectrum, it's
- [03:38]
C
like quarterly, maybe, maybe a little bit faster than that. That definitely is like there is this category of startups that are doing it. You hear the term lightweight MMM a lot. Like, you know, like if, oh, like North Beam or Mutant X. Like there's a lot of companies that have sort of light lightweight. MMM is meant to imply like you just kind of implement this. Like it's more software ish, it fits in better. And I, you know, I feel like they're trying to shave it down to like weeks as opposed to like, oh, you ran this campaign and you find out, you know, next year how things went. Like you, you know, a little, it isn't like real time, but maybe it informs. Like if you run a two month campaign, it can inform your like next campaign. And they're trying to get it down to like informing what you have in flight.
- [04:25]
A
I mean they have to do that.
- [04:26]
B
This is actually.
- [04:27]
C
Yeah, right. People, people definitely sell that as it's something you can.
- [04:31]
B
And maybe this is a good segue to kind of compare because I think historically Google with search, meta with social, I mean you're looking, you're optimizing your campaign daily, not yearly. And you can kind of see these conversions flow through quite quickly. You've got seven day conversion windows, 30 day conversion windows, you're kind of tracking that way, way faster. So what's with these companies going into the, the dustin off the mmm?
- [04:58]
C
I think partly it's like they're just responding to what advertisers want to a degree. Like, like I said too, like they're just, they're falling back on this because MTA is just totally gone. Like, I think that's Just not a valid like it's not feasible anymore. Nobody really wants to do last. There probably is a lot of like definitely default to last click out there but the platforms don't like last click anymore either. Like they, they want to do a better job showing of like how YouTube shows up and they're like streaming whatever and all this other fun stuff that isn't so last clickable. So you know they're, they're very self incentivized and I think also the, the sort of case I'm making in the book is that this is a way of in their minds like delivering transparency, like delivering on the advertiser like the ceaseless clamor for transparency because it's like oh here's an open source model. Like there's no black box, there's no jury rigging this. Like it's your, you know, do what you want with it. So I think that's a very compelling aspect of it too of like the sort of sales pitch from Google's perspective, from Meta's perspective,
- [06:12]
A
why they leave it open source or why they.
- [06:14]
C
And why. Yeah, I think it is open, right? Well that's why it's open source. They, they can like lean on that transparency and say you know, you can trust Meridian. Like Meridian, like you can just look in like what is Meridian doing? You see the code. Like it's not like a sort of the Google machines and it is a way, you know, I think like they're getting, it's not like they're changing performance max. Like it's not like their ad platforms actually have to get more transparent. They could just sort of bump the, the scope of measurement up so much higher that like you're just looking at Google or like you know, YouTube or sort of these big channels and say like oh we can be transparent at this, this height and you can almost like forget how untransparent your, you know, the ad platform is.
- [07:01]
A
Handy.
- [07:05]
B
I feel like one. There's always been a little bit of tension I think between search and social and TV in terms of are they stealing the attributions or are they stealing the credit? Because everyone knows you run a TV ad, all of a sudden your social ads start performing better, your search ads pop. And if you're occluding marketer, you know, the real reason for that is is the TV for example, not the lower funnel searches that are happening after that. So does this kind of maybe amplify those concerns that marketers might have about Google and Meta taking more credit than they deserve or does it kind of head them off and contextualize their role better?
- [07:53]
C
I think they are trying to sort of like divert or address those concerns behind by sort of pushing for adoption. And also this is like a moment in time when sort of measurement, how, how measurement and attribution works is kind of resetting. And so you know, Google and Meta and these companies, they want to, they want to set that the groundwork. Like they want to understand the rules and be the one to write the rules. Like that sort of is probably a pretty comfortable place for Google to be.
- [08:22]
A
I know there's a lot more to say about Meta and Google and their open source products, but can we divert for a second and talk about Amazon's MMM product? Because it isn't open source and I think it's really interesting you point this out in your story. Amazon is totally upfront about the fact that the whole point is having it make Amazon look as good as possible, both as a media seller and a retailer. There's just like no pretense there. So does that kind of transparency, we were just talking about transparency, does it weirdly make Amazon's approach like more honest than Google and Meta?
- [08:57]
B
I do think so.
- [08:58]
C
Like Amazon, right? Amazon is selling something much more straightforward. It's like their, their MMM product. Like you said, it's not open source. It's just like the MMM data feed within Amazon. And yeah, the point is to make Amazon look good. It's a way to hook up your like Amazon data with your retail store and like your other, you know, it's going to boost Amazon's performance. And the whole point is that it like, you know, it's an Amazon like ROAS improver and generator, all that. And there is a bit of like a right, you know, it's kind of like facade of the open source, like okay, here's an open source like transparent product that Google probably a lot of marketers don't even know that they're using Meridian under the hood of like whatever their agency or vendor is using because you know, it's almost like Chromium, like it is like a default adopted white labeled kind of product. And you know, a lot of people I talk to are say like, of course. Like unless you're like a hardcore marketing mix modeling startup that's like all about selling your model and like owning that. Like even some of the most sophisticated vendors that you would sort of think of in that category would say like, oh yeah, if I were building hours today, I'd do it on Meridian. Like that Just sort of makes sense. And you know, everyone's spending so much on Google and it has, you know, all these like great little Google things. It's mapped the Google verse so beautifully. So it just like makes sense. That's how that, you know, that's how Google works. Like they just make this like it's almost you know, a spiteful decision not to use it in a way. And you know, they're giving it away for free. But I think there is this danger too in that, you know, people I talk to say like, okay, like oh, it's transparent, like there's no black box. But at the same time the default like off the shelf settings are very much like absorb Google data, like favor Google. And it is made to be customized but probably a lot of people do not customize it or are not like sophisticated, not data scientists or something and they're just kind of using the product off the shelf. So yeah, it is kind of like a buyer beware. This isn't just something you can click and use. You really have to put a lot of time and effort into it and if you don't, it's just, it is going to end up favoring Google or Robin and Metaspace.
- [11:24]
A
Yeah, I mean Henry Innes, who you talked to, the CEO of Mutinex, he had a very like funny quote that I really liked in your piece that using Meridian and getting it for free is like getting a free puppy and not a free beer. Like a free beer. You just down it and enjoy the buzz and you're good to go. Like you have to feed a puppy. Like you gotta take that puppy for a walk. That's an investment of time. And if you don't have the data scientists to make it work or you're not sophisticated or whatever, I don't know, it would be like giving me access to a professional kitchen. I don't know how to cook and I'm just using the microwave or something. Like it does work. Like my food gets hot. But yeah, like you said it would be a very Google friendly version.
- [12:05]
C
And it is, there's like real data science. Like it's, you know, that's definitely beyond me doing all a whole mess of interviews and briefings for this. Like, you know, people are always talking about like multicollinearity and you know a lot about Bayesian causal inference. You're going to hear that a lot from people who can't explain it to you. So it's, you know, it's like, yeah, it is like not for the faint of heart. So it is misleading almost that it can be used off the shelf.
- [12:34]
A
Right. That it's just like here's a free product and who doesn't love free? But there are other considerations for sure. I thought it was also interesting this other point that you made. It was kind of just as a side point. But it's very salient that Google apparently prefers when brands don't dig in too deeply and they don't realize that their MMM is running on Google code. And that does make sense because then, well, if you don't know, then you don't ask uncomfortable questions about whose interests the model is sort of optimized for. Right. So there's an apparent third party neutrality vibe. But if you don't know then you, you don't know and you just, you accept the results. And they're probably close to reality. Right. I mean Google is Google. Like we're not talking about something Podunk.
- [13:21]
C
Yeah. I mean and they, they've invested a lot in the product itself. Like what I've been told is that Google has put a lot of sort of engineering resources. Yeah. And sales behind Meridian, which is strange. Like for an open source product. Multiple people told me that there's like real KPI, like their salespeople are commissioned to and like you said, like they're, you know, a win for them is an advertiser, an account that's using it and they don't even have to know it's Google Meridian. They want this kind of framework adopted and there's nothing like that. I've actually heard the opposite from, from in terms of like meta. Yeah, they're pulling back, disinvesting, which I think is probably something to do with like there is no Google Analytics part of Meta. Like Meta truly is just like a media company and Google like Meridian is now plugged into Google Analytics. It's very much like a part of that.
- [14:19]
B
Do you need to like store the data on Google Cloud, James? Is this, is this part of it as well?
- [14:24]
C
I don't. Because it's open source, like in theory you can just take it and use it anywhere you do. But all the like fun Google exclusive stuff I imagine has to be, you know, all these clouds. Amazon does the same thing with its MMM data feed by the way. They funnel you into their like cloud based ad tech like suite. So. So yeah, that's. That isn't. What's that. That's pretty standard.
- [14:50]
A
Is there more to say though about why Meta is pulling back? Like I think that's really interesting because we were just Talking a few minutes ago about how marketers demand transparency. It's like all anybody talks about. And it's, it would make sense for a platform to invest in this kind of product. Maybe not if you're Amazon, I guess, but it's interesting to me that they would invest in Robin and then abandon it or seem to abandon it.
- [15:16]
C
Yeah, I think the, the one person described it as more of like a science project within Meta, which is how it started in Google. Like it was just someone's project inside Google who did it open source. And like there's lots, there's hundreds of types of things. So that's why it's very rare for Meridian to have been like plucked out and given all this like distinguished backing and all that. So it like, that isn't so crazy that like Meta sort of just treats this as this little like science. Oh, like we made this, it's available. I don't think there is such like an incentive for them to really push it the way that Google has. And there's like, it is. Yeah, it's not, they're not going to be the, the one here. Like people aren't going to use a numerous, like people use an incrementality vendor and another attribution vendor and like this kind of measurement vendor and an MMM vendor, but they're not going to use like numerous MMM models. Like people are going to fall back on Meridian and I think Meta is focusing more now on like shaping how it shows up in Google Analytics in these sort of other. Mmm. Sort of. Yeah. Like if you go through Google and we want to still improve how we
- [16:31]
A
show up in MMM reports, like the Engage through attribution. Is that what it's called when.
- [16:37]
B
Yeah.
- [16:37]
C
Click through attribution. I think that. And they also have an Engage through category. But yeah, earlier this year they did sort of make a subtle but important change to how they, how they pass data to Google Analytics. Namely like it was, you know, it wasn't just that, but that's, you know, we're talking about Google Analytics, there's Adobe analytics, then there's kind of nobody else. And yeah, that is when it, when this wasn't like an interview but they did, they allowed press to, you know, Q and A, some people, some product marketers about it and they, you know, their marketing team talk brought about, brought up like this is about showing up in MMM feeds and, and they've made other things too where they've like changed how, how like their ad clicks work, how things that are in the background. You might not even notice. Unless you're like a user who feels like huh. Like I feels like I kind of accidentally click or like on a lot more Instagram or Facebook feed ads. And it's because they increase what's called the hot zone like area where they consider like a click and they're. Yeah. So they're feeding that into these systems to show up better in these reports that once upon a time just were like search last click, search wins and they want to diminish search. They're kind of using co opting the same search language.
- [17:55]
A
Oh and conversion windows. They're definitely expanding conversion windows too. Right. I mean that's another tweak that you can make.
- [18:02]
C
Yeah. And especially as that is one of those things too because I think as people adopt. Mmm it does let you take a longer look. It just naturally like stretches. You know. I think a lot of people probably get like very heads down and they're like, oh, our campaign performance today or whatever. And it's, you know, they don't, they don't look at like a more natural timeline. So that probably is like a healthy just result of the MMM trend too is people you know, more open to a longer window and like okay, how is this actually working over months?
- [18:35]
A
Yeah, it's a better reflection of reality how people move through a funnel. They don't move through a funnel in like five seconds.
- [18:40]
B
When you. A while ago, James, you're talking about kind of like the. How Google might not even want marketers to know that like they're using the Meridian code. I know sometimes it's been white labeled. So like what, what would that actually look like? Like how would a marketer be using Meridian without maybe knowing?
- [18:58]
C
They are a lot of these vendors. There are a lot of. First of all, there are a whole like category of Google sort of services vendors that help brands like with just this kind of thing, the ad swerves and jellyfish and stuff like that. And you know they're like this is what they're experts in. I think too like it's, you know, if you're a measurement vendor or especially if you're like an agency and it's not like really exactly what you do. Like why, why not? Like why build our own model for this when Google has something like pretty good that we can just sort of say we've got. Customize. Yeah. And truly like you can customize. Most of these agencies have like data scientists who should be, you know, who can do mmm reporting and all that. Yeah. Modeling all this stuff. So. So it does it makes total sense to me why you would fall back on Meridian as opposed to. Yeah, like, there are, you know, there are some vendors that are really expensive. Some people spend like tens, hundreds of thousands of dollars for a modeling service. So, yeah, like, there's a very. There's a very like sort of clear rationale.
- [20:14]
B
I think, I think some of it's just Excel spreadsheets, like, you know, very more advanced ones. But then also you mentioned like the data science aspect of that too, which is maybe the way it's going versus the weighted Excel spreadsheet.
- [20:28]
C
People mix them too. I have heard it like that. That was also one of these sort of trends. Like a year ago, did a story with adsorv that was working with. Yeah. An airline, Alaska Airlines.
- [20:39]
B
Yeah, I remember that.
- [20:41]
C
So they were going. They used Robin and then they'd used a hybrid like Robin, Meridian, some other stuff. And then they were just like, eff it, we're going to Meridian.
- [20:53]
A
They called it a Frankenstein's monster. I remember.
- [20:55]
C
Yeah, that was their terms. I felt bad using that in the story. I don't even think I did. And then it wasn't in the story.
- [21:00]
A
And then you. But it was in the interview.
- [21:03]
C
Yeah. So I was like, all right, well, that was, that was frank. Yeah. So there. And there are. I'd heard of other cases of that too. A lot of people were on Robin early on because they were the only one. They were the only open source model on the market for they were first, like 20 years.
- [21:20]
A
Yeah.
- [21:20]
C
Yeah. Uber released one. Like, lots of companies sort of like took a swing at it that sort of didn't make any sense. And I think they just sort of wanted these into these models. Like, hey, like, consider our point of view. Like, Uber's gonna show up better if you just like paste this into your model.
- [21:36]
A
Yeah.
- [21:36]
C
And so, yeah, it makes more sense to just now I think a lot of people will sort of fall back to Meridian, which has had a lot of investment in new products. And like at Google Marketing Live, it got like some sweet new integrations and updates and all that kind of thing. So, yeah, like, and they're really pushing it.
- [21:57]
A
And the advantage too. I mean, we didn't really talk about why it's good. We were just immediately talking about the grading, your own homework aspect, black box or whatever. But it has very deep integrations with Google's own channel. So it's cross channel, but gets all of this search data and YouTube. It's like data that only Google can provide. And that is a big advantage.
- [22:20]
B
And I guess maybe let's bring it all home. James. We'll have a succinct podcast. So I'm a marketer and I tell you, hey, I'm going to go with Google Meridian or I'm going to go with Robin or I'm going to go with Amazon. Like what's kind of your kind of coffee shop. Quick read reaction in terms of cool. But also watch out for this or like that'll be helpful for this. Like what's, what's your, what's your read?
- [22:50]
C
Yeah, I don't, I mean I think, I think it is a good idea to, if you're like a big brand, to, to invest in media mix modeling. Like maybe you did it once upon a time and don't anymore. I said media mix marketing mix modeling.
- [23:06]
B
Okay, so, so one, you should be doing it.
- [23:09]
C
Yeah, I think you should be doing it. If you aren't doing it, start your channel spend. Like if you're a search and social kind of brand or something, you're a TV and retail and you're not anywhere us. Like it's, it's really for companies that have lots of channels that they're marketing in influencers, podcasts, radio, tv, blah blah, blah. And they also are like it's part of it is about being able to account for random variables. So like if you're a business where like you know, you would, you're. You sold a ton of coats because it rained or had a big winter or something, or the opposite celebrity said something about you. Yeah, like there's going to be, you know, the Knicks win the championships and like booze and like some other, you know, there's going to be a huge spike that you have to be careful. Like, oh, is that your, was that your TV campaign or did the Knicks win the championships that week and something crazy happened.
- [24:01]
B
Right.
- [24:01]
C
So it's, it is about like marketing mix modeling is about building in. That's where like the Bayesian inference stuff gets in. But like these random variables that can affect things being able to account for all of these things at once. So if you're a brand for that makes sense, you should invest in it. I think it is just like a better healthier system even than like the deterministic kind of daily performance in the trenches. Like yeah, I don't know, laser vision. And just be, yeah, be mindful if you use Meridian, like no, no shame in that. It's a perfectly good open source model. But understand the work that you are putting on yourself and the like burden to personalize this product, like, make this a really strong reflection of, like, what actually makes sense in your business or otherwise. It's just. You're just gonna get, like, the Google off the shelf, which is gonna, you know, not be. Be a distant truth. Yeah.
- [25:04]
A
You're adopting a puppy.
- [25:06]
C
Yeah. Right.
- [25:08]
B
And I like that point, too, James, about the mmm decision being about what channels your marketing is running in and the value varying depending on that piece. So thanks for filling us in on your story. Let's read the whole thing. Let's all read the story, if you haven't already. And we'll see you next week.
- [25:36]
C
It.