Transcription
It's the end of cost controls. Not really. Not quite. But it is on the right path. AI is transforming Meta ad accounts again. Meta is building a 2-gigawatt-plus data center that is so large it would cover a significant part of Manhattan. I'm going to show you exactly how Common Thread Collective, who manages paid media for some of the fastest-growing brands in the world, is setting up our ad accounts in 2025. And this is where, again, the idea that you are going to look at the last seven days of history and make a better decision than these machines is just, it's such hubris.
Think of context. Like, one of the things that people struggle with with, uh, like if you go into ChatGPT, the context is like individual data points that it can process. So ChatGPT couldn't afford to give everybody an infinite context window. It would just cost too much money. But Meta, for the sake of their ad platform, is basically, when they tell you they're funding all this money into all these giant warehouses, it's all about this. It's about the amount of contextual processing power it can have to ensure the match between ad and user is perfect, right?
What I'll tell you is that every brand we work with makes us turn this off by default, despite the fact that Meta says that there's a 27% improvement to ROAS using AI enhancements. If I have the capacity and I go to my bin of tools and there's one in there, what I know about you as an advertiser is...
Today's episode of Sharpen Your Skills is brought to you by our partners at Vermont and Mercury, two incredible contributors to our ecosystem. This wouldn't be possible without them.
Meta just announced a sweeping change to how you're going to be running ads in 2025. I'm going to break down the whole thing for you. Is this the end of media buying? I mean, let's hope so, because I, I am in the camp that myself and most of us who are doing media buying at a very manual, human level are causing more harm than good. Uh, and I think this progressive set of changes from Meta illustrates a similar set of thought, which is that there's a better way to make decisions in your ad account than to have humans trying to process small sample sets of historical data and predict behavior in the future. Uh, and that really is anchored in AI.
So today, Meta just made a big announcement, which is that they are going to be moving all ads, uh, and this isn't for every account right away. For us, we got about 50% of our ad accounts. Uh, this announcement today in, uh, early February, that they're going to be moving everything by default into what they call ASC, uh, you formerly referred to Advantage Plus Shopping Campaigns, which are now going to be Advantage Plus Sales Campaigns. So there's going to be one campaign build flow by default. Everything is going to be an Advantage Plus campaign. Um, and on top of that, that's going to be the sort of major announcement that you're going to see that a bunch of media buyers are going to overreact to. Um, they're also releasing something called Campaign Optimization Score, where they're going to be giving you a set of pre-programmed recommendations that continue to push you further out of the way and more into their AI systems.
But what I, what I don't want you to miss, that's not going to be the headline, that's a little bit nerdier, that I would suggest you go read about, is the update to Andromeda. Okay? So Andromeda is Meta's, uh, proprietary ad retrieval machine learning system built on top of Nvidia chips. So if you work with Meta in any way, you'll, you'll get a deck from them where the first slide just tells you about how much infrastructure they're building around AI and computer processing. And this is really important. We're going to cover a lot about the impact of Andromeda, the AI ad retrieval system, and how we think about how that informs our, uh, campaign development. But it's really important to understand what it enables and why it's so powerful and important and how it informs a lot of our media buying strategy in 2025.
So for this video, what I want to do is I want to start by talking a little bit about where we've been and then where we're going in terms of how we recommend now and how we're going to be building in our ad accounts, because it's pretty different. It's a pretty big shift.
Anytime you open up a new campaign on Meta, the default is going to be ASC. If you're building for a sales campaign, your default flow, the default UI workflow, is going to be build an Advantage Plus Sales Campaign, is what it's called. Um, so there's a few different things. In, in a historical CTC, there's a video, uh, that I do in our drawing series where you'll see us reference this idea of pipes. The "pipes" media buying methodology is something we popularized around building a lot of things centered around cost controls, okay? And lots of campaigns, okay? And the idea was that every campaign contained its own unique, um, offer, angle, and audience. And any change to these three variables resulted in a new campaign. We would call this a concept. And the idea was, build as many concepts as you can. And every concept would have four to six ads in it. Uh, and we would build. If you open a CTC ad account, you're going to see a lot of campaigns all on cost controls. And each one was a pipe. And the idea was that, uh, the, the cost control is controls the flow of dollars through that pipe. And by stacking a bunch of them, you can build a lot of profitable flow, okay? So that's sort of like CTC best practice. Sometimes we're ASC, sometimes we're BAU. We kind of bounce back and forth on that. But now we are sort of following the same trend and we're going, all right, the future is Advantage Plus, uh, sales campaigns. And we're going to think about how this changes our behavior. And it changes it in a few different ways. One, we're going to let go of the idea of BAU versus ASC. Um, and we're also going to sort of really work towards consolidation at the offer level, um, and allowing for more creative to exist at the angle and audience level within the same campaign. We're going to talk about that. And it has to do with Andromeda.
So I'm going to outline what is the way I would build a campaign. And we're going to use, we've been using Born Primitive for a bunch of examples, so we're going to stick with them. And I'm going to illustrate, uh, how this would be set up. So right now, you're going to have an ASC campaign. That's going to be the optimizer, the default structure. You're not even going to have to select that. All right? Then from there, we're still going to be, uh, probably more often than not, value optimized. Okay? So this is another ongoing change that we are trying to get, um, in CTC campaigns, more and more moving from lowest cost to highest value in terms of our optimization setting. So this is moving out of lowest cost into highest value, um, as the optimization setting.
Now, why is this okay? Well, it's because for most businesses, and I say most because there are obvious exceptions. If you're a single SKU business or you run funnels to locked-in landing pages where there's only one purchase option, lowest cost can be really effective. Bid caps and cost controls can be really effective because you know the exact CPA that you're after. But in the vast majority of cases, brands are running from an ad, okay, to either a category page or a PDP with different size, color, SKUs, where the order values, so let's say this is order values, um, exist at the, and they look like this, where if this was a histogram of order types, okay? That means you're going to see a range of order values, like let's say this is $50 and this is $150. And you're going to see orders come in on a broad spectrum because the user has options, right? The user could be buying, um, the cart values are different. Exactly. So the cart values are all different, and the order values are all different. So if I bid, let's say I'm on lowest cost, okay? Um, and I set a, you know, whatever my CPA target is, $40, because my AOV might be somewhere in the $80 range, let's say that's my average, right? Right. And so I'm like, "Oh, I want to bid, I have a bid cap at $40 against an $80, that's a 2:1." But what you've done functionally is you've taken half the order values, including the, the higher range, and you've eliminated them because you can't actually bid $40 and get the $150 purchase. You'd have to be bidding closer to 75, right? And it'll always move. It'll. This is why whenever I go into a lowest cost account, we always see AOVs be lower, lower than normal. But, um, so often the modal order value is not the average order value. The median order value is not the, uh, average order value. And so if there's a bunch of disparate purchase options, more and more we're moving to value optimization with a minimum ROAS, a target ROAS. Yeah.
In the current state of Meta, though, are people having to put in their AOV still manually? How has that gone as far as? Well, you're not putting in an AOV, but you're putting in a bid if you're running bid caps or cost caps based on an expectation of AOV. And this is where I just see us be wrong all the time. And so it's one of the biggest limitations with bid caps and cost caps is that it's still a very human expectation about what you anticipate the average order value to be. And it's people are just wrong about this all the time. And it causes constant changing to your bidding. Um, and this is one of the biggest critiques, I think, that is totally fair of cost caps and bid caps, is that it's very human and it's dependence. So value optimized at target, I shouldn't say min, it's actually target ROAS is sort of a solve for this, right? Especially for the biggest brands with the biggest catalog. That's right. More SKUs and the more disparate the purchase options are, the more you want to push people up to the higher value purchases. That allows you to bid a higher CPM. You can be more aggressive and go from there. So we're still, in almost every case, and again, there are exceptions, always going to still be implementing a target for Meta. The, the part of the system that they do not understand is what your expectation of your marginal value is yet. Now, future updates are going to include profit optimization. That's going to replace revenue value. And that's a great thing. That's the, that's the future of Meta, even more, is to just keep feeding it more of your business outcome. And there's even people I know that are feeding back marginal data and doing value optimization against a marginal result, which I think is a great thing. I know that's the future on, on Meta's roadmap. But this idea, uh, once again, let's say we're doing this for a leggings campaign for Born Primitive, okay? Uh, we're going to have an ASC value optimized at low, at highest value, on target ROAS. Okay? So that's the optimization setting.
Now I want to talk about, um, the, the different options as it relates to what your window of optimization is that you're going to give to Meta. And my current school of thought on this, um, historically we have been by default a 7-day click, almost universally trying to press towards more and more 7-day click as the default optimization setting that we're using. And it's because we have found the, the most problems caused in an ad account by an overemphasis on view attribution to existing customers in a way that conflates the, the, uh, the attribution. And often you'll see Meta overreporting its impact when you're running 7-day click, one-day view versus when you're running a tighter expectation rate on click. And we often see 7-day click is slightly underreporting or is closest to the actual result. Now, when would you run an even tighter window than 7-day click? Um, almost never. I, so this, this is, this is what I would say is that I am moving more and more my school of thought closer to 7-day click, one-day view. Yon Levy, if you're watching, you're going to, you're going to like this. The key is though, that I, in order to do that, I have to have an incrementality study, right? That gives me the actual causal result of the 7-day click, one-day view. And again, this is all about Meta signal. To Meta, signal to Meta, giving them the as much information as possible about who is the kind of person that actually buys from me. Now, what I find is that, let's say in Meta, I have a view of all three of these. I might get like one-day click is reporting a 1.1, 7-day click reports a 1.5, and 7-day click, one-day view reports a 2.1, right? This is often what you see is that the 7-day click, one-day view overreports the impact. But if I have an incrementality, uh, ROAS adjustment that allows me to see that the actual result is like a 1.7, as an example, well, then I can create a factor that allows me to set my targets. So if essentially if my goal is to get to a two, then I might need to be at a two and a half on 7-day, one-day view. But I'm still giving Meta the most purchases possible. So this is a change in CTC world. Um, there's still going to be cases where, uh, we might be on 7-day click, especially if the brand tends to have like a ton of organic traffic or other things that really conflate this. It does create some complexity because the percentage click and view could change over time. So it, it is, it does present challenges and it's not perfect. But I'm just, I'm more committed to that, whatever setting you're using, uh, on the optimization, run an incrementality study on that setting so that you can get the results and act accordingly. And the more signal you have, especially for like, we work with some brands that have like $2,500, $5,000 AOVs, like you just can't get enough signal on a 7-day click basis, um, to really get to optimization. Now, the good news is AI is reducing the need for the amount of signal that we have. So, but regardless, I want to give Meta all the information I can. That's going to be a common theme that you hear as we talk through this.
Does it make the learning slower then? Well, ideally faster. Faster, because you're giving more purchases back. Okay? Right. But one of the things with Andromeda and other things is that this idea of the 50 purchases conversion is coming down. It requires you learn faster. The amount of signal Meta processing is, is, uh, increased. Okay?
So we've got our Advantage Plus, Advantage Plus Sales campaign. We're going to run value optimized. We're going to have a target ROAS. And we're probably going to be on 7-day, one-day. Now, at the audience level, we are still going to be very much, uh, broad audience. Okay? Is still absolutely going to be the setting. And this is even more related to again, Andromeda, which is allowing for more contextual signal to better perfectly match each ad to each individual user.
Rising CAC, tariffs, inflation, being profitable in e-commerce is hard. That's why managing your finances shouldn't be. Our partner Mercury helps you combine banking, credit, and working capital into one seamless, easy-to-manage financial system. You get banking with no hidden fees, so important. You get 1.5% cash back on all credit purchases. And you get predictable payback on all of your working capital loans. In addition, you're going to get a beautiful, easy-to-manage cash in, cash out dashboard to track all of your financial health in one location. And if you sign up now, you're going to be eligible for two different rewards: $250 when you deposit $5,000 into your bank in the first 90 days, and another $250 when you spend $5,000 on your credit card in the first 90 days. That's $500 in rewards within 90 days of joining Mercury. Join over 200,000 business owners who trust Mercury to manage their finances. And of course, tell them Taylor sent you.
One of the things that I want us to understand about Meta that is different than almost every other media buying mechanism, if you're buying television or if you're buying display or you're buying placements in a newsletter, is you are buying bulk placement, um, across a broad set of users. In Meta, this is all user-based advertising. Okay? Each individual user is inventory that has a set of behavior and a set of response history to different ad formats and creative types. And so what, uh, Andromeda really does, and I'm going to stop here to explain a little bit about this, is that it expands the context window, okay, for analysis of user behavior. Okay? I want you to imagine that when a human processes, journey is much the amount of data that Meta can process about Mickey and the context of your historical actions is massively increased. That's what these giant data centers are all about. It's about the amount of context and processing power I can have about your behavior. So when humans do data analysis, they tend to time bucket things. So they'll tend to say, "Give me data for the last seven days and let me analyze that and make a behavioral decision." But the problem is, Mickey, I don't know, when did you sign up for Facebook? Signed up for Facebook 2010, right? So we're talking about 12 years of history on the platform, things you've clicked on, things you've purchased. Almost 12 years. Yeah, like, sorry, 15 years. I'm bad at math. This is why humans shouldn't do this. 15 years of behavioral history that builds a profile and a pattern of how you act. Mhm. The kinds of ads you respond to. Does Mickey click on videos or images more? Does he like, you know, this kind of copy, that copy? Is he on IG Stories or IG Feed? Does he like, and, and then the, the myriad of data that is on top of that, the friends you like, the things you share, the depending on who it comes from, whether you're more likely to interact with it, like who, who influences you. Window then grow. If I saw something six months ago, a year behind. That's right. So think of context. Like one of the things that people struggle with with, uh, like if you go into ChatGPT, the context is related to the amount of tokens, or think of it as like individual data points that it can process. And if you try to go upload, like if you do this right now, if you export, let's say 30 days of individual ad history in Meta and you try and upload it into GPT, it's like, it's not going to work. The first two and then hallucinate. That's right. So like, it, because the context window and the processing power, because it literally takes compute to process that much information. And so ChatGPT couldn't afford to give everybody an infinite context window. It would just cost too much money. But Meta, for the sake of their ad platform, is basically, when they tell you they're funding all this money into all these giant warehouses, this is a, this is a data center out in Utah or wherever, right? Like billions of dollars in these, you know, these giant data warehouses. It's all about this. It's about the amount of contextual processing power it can have to ensure the match between ad and user is perfect, right? And this is where again, the idea that you are going to look at the last seven days of history and make a better decision than these machines is just, it's such hubris. I can't even begin to to understand like how, uh, how how big the gap is between our human processing power. And that's the same and true on defining an audience. If you think that you are going to define by selecting a 1% lookalike or an interest group a better audience for Meta than saying, no, no, no, we're going to allow you to understand the individual user, the information about the ad that I have, including what, uh, what creative formats exist, the copy, the URL, who else has liked it, clicked it, etc., and allow you to design the perfect match between user and ad. That's not even close. Then you're crazy. You're literally crazy. And so, and that was true back when it was Meta. So if you think about what Meta engineers did before AI, is it was still very human in the sense that you had, uh, smart data science processing individual time buckets of data and designing algorithms based on that behavior. But now what happens is the AI is contextualizing an, like a much larger set of behavior, um, to be able to create this match, right?
So we are still going to be value optimized. It's still going to be broad. But I want to talk about something here. I want to talk about, uh, exclusions. Okay? One of the things that, um, is still a thing that, uh, brands care a lot about is the distinction between new and returning customers. And Meta allows you to define at the camp, at the at the account level now, right? At the account level, um, existing customers and engaged customers. Okay? And I want to tell you what we're doing with this because I think this is unique relative to the way that most people are handling it. We are calling existing customers active customers. And this is where we are making, I'm going to make a big push to force our industry into letting go of the idea that someone who bought from you five years ago is an active customer that's going to buy from you and needs to be excluded from your ads. You have to stop this. This is suffocating. You go watch my "Suffocation Starving Giants" video. So an active customer, again, just to define this, we look at the sequence of purchase history. So if this is day zero and this is day 500, in a C, the aggregate average history of customer behavior, the time between purchases, between purchase one and the next purchase, where's that 80% mark? Okay? Where's that 80th percentile where people are not likely to purchase? Let's say it's 130 days. Then my active customer set is everybody who's bought within the last 130 days. Does that make sense? Okay? Now, the engaged audience, we still want to understand the distinction between a net new customer in entirety and customers that we would call lapsed. So think of for us, an engaged definition in Meta's term is lapsed. So we are going to take this as non-active or lapsed. Okay? These are people outside of this window. Okay? And we want to win these people back. Now, what that's going to mean is by doing this, you're going to be able to see at the account level how much delivery you're getting to each customer type: new, existing, and engaged, or for us, lapsed. I don't actually care about remarketing in the sense that the definition is somebody who went to your website 30 days ago. That's not important to me. I don't actually care about that.
What is the current way that brands are segmenting this? Most of the time, engaged is website visitors. That's it. That is what I see. Pixel shoes up. But I, I don't care about that distinction. I care about this, the definition in customer state, uh, and then having optimization according to these settings. Again, as long as I have an incrementality study on this, and the best incrementality platforms will break it out for you of the incremental impact on new versus returning customers. This is going to be fine. So at the audience level, now, this is what we're doing. So this is the campaign setting. This is at the ad setter setting level. And then the key is I'm building each of these campaigns around an offer distinction or a product distinction. So when we say offer, what we mean is what you're selling. Mhm. Okay? And now this is where in the past, we would create a different campaign for every angle and audience. So let's say I was selling leggings to, uh, moms versus to students. Yeah. I would create two separate campaigns for those to try and aim the pixel and get creative, different creative delivery, and also for creative reporting. And because the BAU best practice wasn't to load a hundred ads into the campaign, because again, we didn't have Andromeda, we didn't have this large-scale ad retrieval system and with sequential learning that allowed us to process way, way more information. So the other big change for us is we're no longer creating new campaigns based on angle and audience. This is just, we're going to move my graph here. This is now just going to be tons of creative inside of this campaign.
What does that look like practically? Is that the headline? Is that the? So we, It's a great question. So we would like to create as many diverse ads, okay? So diverse ads meaning recognized by Meta as distinct. So things that often don't get recognized as distinct or unique ads is if you just change the color of the background. If you were to, uh, maybe just change the copy on the same image in various ways. These are ads that are distinct, um, ideally distinct in their ability to access placement. So as an example, as a creative practice, right now we make a rule because Meta, um, prioritizes it and makes you eligible for rewards if, uh, you have a vertical. Every campaign must have vertical video with sound on. So every campaign that we have has to have vertical video with sound on. It makes you eligible for rewards, which is not a signal to you that this is of value. So even if it's taking a static image, creating an animation of the logo and putting some music in the background, we do not launch a campaign unless it has vertical with video with sound on. Is a, is a rule we're implementing into our system. We know creative volume is everything. But let me show you a tool that we're using at Common Thread Collective to help our clients produce exponentially more ad creative. It's called Vermont. Vermont allows you to take the hundreds of ads that you've just developed using CTC's exciting new system and take what is the reality that right now, every one of these ads is probably driving to a single landing page. And so what that means is that, let's say you have a hundred ads times one landing page is going to give you 100 funnels. But let's imagine we could do this in a better way. What if each individual ad was able to drive to not just its own individual loner, but what if you could create four variations? And maybe that this one needs its own bundle, or maybe this one needs a price test between two different ads. What if this one needs to have one featuring older people and one featuring younger people? What if every ad you created was able to have its own perfectly designed message and funnel so that you could exponentially increase the amount of funnels? So instead of just one landing page, we went to hundreds. Fast, easy, no code, no developer needed. That's what Vermont enables. Their tool allows you to quickly, lightning fast, build as many variations of your landing page, and not just landing pages, but full cart and checkout experiences instantaneously. They give you AI recommendations for ideation to go from just having a hundred funnels to thousands of funnels instantaneously within minutes. You can be testing tons of different variations of their pages to produce the best results and exponentially increase your creative volume. That's the kind of math that Pierre de Vermont himself could get behind.
And you're saying that Meta is going to be the one to fill that in? If it's a one-by-one, they're going to make it a nine-by-six? I'm saying that you have a choice. Your choices, uh, you could choose. And this is, this is really where I think if you really want to solve this, and I'm going to tell you something about Andromeda right now. What you do is you turn on AI enhancements. So AI enhancements is going to be our recommendation for every brand that you turn this on because we believe so much in creative volume as a variable to unlock scale. And this is the easiest way to do that. Now, what I'll tell you, what I'll tell you is that every brand by default wants this off. Yeah. What if it looks bad? What if I don't like the way it looks? What if I don't like the music that's on in the background? What if the image? Yeah, there's a thousand reasons why people right now, almost every brand we work with makes us turn this off by default, despite the fact that Meta says that there's a 27% improvement to ROAS using AI enhancements. Okay? And the reason is, I'm going back to my friend Andromeda. And I'm going to keep beating the, the horse. We're going to come back. Seems high. That's really high. 27%. That's really high. That's the difference between profitability and loss on the bottom line. That's double your turn. So Andromeda's new ad retrieval system prioritizes, okay, prioritizes advertisers with more options. This is so important. If I have the capacity to very quickly process what every user wants and know exactly what I need to deliver to them from placement to copy to style to format to text to whatever, and I go back and I go to my bin of tools and there's one in there, what I know about you as an advertiser is you're unserious. And to get it to slot into different, do we have the Lego? Right? You're going to force me to take a square peg and jam it into a round hole a bunch of times. And so Meta is going to deprioritize you. When you turn off AI enhancements and you don't feed a bunch of creative volume, you are being deprioritized as a delivery partner for Meta. This is like, I, I can't even tell you how important this is. And why creative volume is the future of this advertising. And the easiest way, because I know like if we go back to our creative operating system, many brands don't have the capacity to produce as many ads as they would like, is to use AI enhancements. But if you're unwilling to do that, then your solution is you have to create a ton of ads, okay, into the campaign and to continually feed new ads into ASC. Now, what I'll say is the other problem in the world that I live in as a service provider is that clients hate, hate, hate, hate making ads that don't get spent. If I could offer you anything in the world, it would be to release that. It would be to release that. But if you demand it, then we can create ad sets within the ASC. You can still do that and obligate certain spend onto ads. But it is not the right decision to make. It is to allow Meta and its ad retrieval system that pairs better to user historical context, which now allows them to also include the historical sequencing of Mickey's entire life, every action he's ever taken, not in a small window, but across everything he's ever done to better pair creative to user, right? Which minimizes the amount of time and energy and testing budget required to get to answers and to prioritize advertisers with more options. So we are going to focus our energy on narrowing the amount of campaigns that we exist. So reducing campaigns by audience and angle into just offer. So the only way we're going to build a new campaign is under two scenarios: one, the offer changes. So we go from selling leggings to selling shoes, okay? That's a new campaign. Now, it may actually be the case if a brand says, we don't actually have limitations on inventory, or if you were dropshipping, that I would just put all of these into the same campaign. I wouldn't actually even care at the offer level. But that's not usually the reality for e-commerce. E-commerce has inventory obligations where you need to create volume at every core business objective or skew. And so those, they'll still need to exist in separate ASC campaigns. Okay.
Quick question then, what does this mean for product dev? Is it easier to launch new products? More difficult? Well, I, I think that what it offers you is every new offer is a new campaign opportunity. You're just feeding it into the catalog. That's right. And, and your job is to continue to feed an endless supply into these campaigns of ads so that you can keep. The other thing we're going to do here, the, the last piece of this. So if we've got, this is the campaign setting, this is the audience setting, this is my creative setting. Is ideally I'm going to have AI enhancements. If not, I'm going to for sure at least have this, and I'm going to feed as many ads as I possibly can. That's going to be the ad level. The other thing is the budget setting. So the way that we're going to run this is still inflated budgets, meaning I'm going to set the budget way above what I think it will spend in any given day with the target ROAS. So if I want to spend $1,000 a day, it's going to be set at $5,000 a day. And I'm going to keep feeding this machine till it can find $5,000 a day at my target result. And that is, that is going to be the work I do on media. We're going to reduce the amount of media buying behavior. I want less human actions, less individual, uh, decision-making, and more work feeding and finding more creative fuel. Um, so inflated budgets, these settings, this audience, this creative structure, and understand that if you want to win the Meta game, you're going to prior, they are going to prioritize advertisers with more options. The media buyer is moving toward the creative strategist producer. It always has, because I, I genuinely believe. Oh, and I moved this over, hopefully I didn't get out of frame here. Um, this is, this may be a controversial statement. I, I just want to be clear that I'm including myself in this. My experience is that almost, let's call it 99% of media buyer actions are net negative on the business. That the decision-making is so chasing data, it's all, it's all reactive to historical response. My bid did this, my AOV did this, my spend was bad. I do this and I react to a very small sample set of data. That it is really almost in every case, uh, not a creative to the business, um, at all. So this is a pretty fundamental change. We, if I have watched anything about AI, so if I, if I think about some of the other ways we're using AI at CTC, is every morning we do something called One Map Notes, where our media buyers go analyze an ad account and they process that and provide an update to every client under the format of what, so what, now what? Right? Uh, the amount of time it takes them to do that is about 20 to 25 minutes per client across three to five clients every day of every week across 30 people. It's, it's hundreds of hours of work, um, that the AI now can instantly process and provide better information, more thoughtful notes, better data analysis, every day and instantaneously. Um, and the end result of this all is the more information. If I could, if I can get Meta, my inventory position, and I can get it the marginal value of every sale, and I can get it at the LTV of every customer, then eventually, if we think about this growing context window in these giant data centers that exist, what you're understanding is that our processing power is too limited for this activity. Mhm. We're not built for this kind of decision-making. And this is where the hedge fundification, I call it, of media buying. You're going to see that in my other video where I talk about IMR, incremental marginal return, is the future of where this is all headed. Is that we can't process fast enough. We don't, we sleep, we go on vacation, we're out on weekends, we're not available to make decisions all the time. And so the on the present, um, able to process an ever-growing context window of information to better align creative type to individual user desire based on the historical behavior pattern of that individual user is where this advertising is headed. And freaking, it's exciting. It should, it should reduce for all of us. We have to figure out margin expansion as a business. And the labor profile is a place that we have to do this. And the reality is that media buying, as a person that goes in and is uploading things into the ad account, clicking buttons, which is fundamentally data entry to build a campaign, and then monitoring and making manual adjustments, is like, it's poor human labor. It just really is. And more and more, there's going to be oversight and thoughtfulness that oversees the systems to ensure that it's delivering against the business outcome that you want. But the day-to-day media buying, to your initial question when we started, probably is dead. And that's great news for all of us because it allows us to evolve the ways that we impact business. But this is what you're going to see inside of a CTC ad account. The pipes, um, delivery is going to probably seed itself to, we're going to sell the best SKUs and the products that have to go live, um, and we're going to build these large ASC with tons of different creative at the value optimized, more often. There's going to be cases where there's still going to be bid cap and cost caps, um, against broad targeting with this structure of exclusions. We're going to be, for especially the more mature businesses, we're going to care a lot about lapsed customers. We're going to segment out how we measure RME by actually flowing the budget into different categories. We're going to start modeling RME, um, and thinking about the price of lapsed customers. We're still going to be running incrementality studies to set the T-R. And then when the brands are brave enough and courageous enough, we're going to have AI enhancements on. And if not, we're going to make sure that we get the right formats to get the most delivery we can and to feed as much creative as possible. Change is good. Welcome to 2025.
[Music]