SPEAKER_00 0:00

Welcome to the e commerce alley podcast. I'm really excited. This is a special episode because I actually have with us the co-founder of Breezeway, Cliff Brown. And if you're not familiar with Breezeway, you need to go check out Breezeway.co. We co-founded this tool with Cliff Brown, who's here. He came up from Atlanta and we've been spending the last two days together. Cliff, I'm so excited to be here because we're going to talk about ads. We're going to geek out a little bit. We'll follow some of the rabbit trails as we go. But man, thanks for being here. This is your first ever podcast. It is. Yep. First time I've podcasted with anybody. Oh, really? Like first overall. Yeah. Yeah. I'm honored. I'm honored. Dude, I'm so excited that you're here because we have, we've been working together for years. Cliff actually started, you I mean, you started out as a client of ours with one of your e-commerce businesses.

SPEAKER_01 0:52

Yep.

SPEAKER_00 0:52

But you've had a background in development and software for how many years?

SPEAKER_01 0:57

Oh, for my entire career. My my background is uh well, my college degrees are in electrical engineering, but that's really applied math. And that's what I've done, you know, most of my career. So probability and statistics and all that kind of stuff. I'm able to apply to e-commerce.

SPEAKER_00 1:13

Yeah, yeah. I love it. So like I think this is what's been so good about our partnership in the last. I mean, you started as a client and we've been building Brazeway for I think three years-ish right now. I think that what is what works so well with us is because you are so analytical and technical, and I'm like very feely. Uh, maybe that's a good way to put it. I'm like, hey, I I see all this data or I see all these things happening, and I feel like this is the threshold here, this is what works, this is what doesn't. And then you get to come in and be like, oh, this is what the data tells us, this is what embedding is on meta and how Andromeda works. And like you give us all this like deep dive, and I think that's why we've been so good, like as like a t a team.

SPEAKER_01 1:51

Aaron Powell Yeah, it's been a lot of fun. You know, honestly, a lot of times I kind of confirm what you intuitively knew, but other times, you know, we find, well, we can optimize things a little bit. You know, if we change things slightly, we can get a little bit better performance.

SPEAKER_00 2:02

Aaron Powell Well, or we find out that my feelings were wrong, and I'm like, ooh, I gotta change the data actually is contradicting my feelings, and I gotta change this a little bit. Well, that happens to me too. Yeah, yeah. I I think that's what's great. Like we're always for for us in like Breezeway, everything is data over emotion. Right. And I I even say that in our coaching all the time, where I'm like, hey, marketing is a mathematical equation, not emotional decision. Maybe that's one of the things that really got us a vibe and right out of the gates, because I would always say that and you are the same.

SPEAKER_01 2:31

That resonated with me, yeah. That that really did. Yeah. But you know, I mean, truthfully, you gotta trust your intuition initially, and then the math comes along later. Does can I can I confirm what I believe is going on? And and that's how you apply the math. Yeah. But you gotta know, you gotta know what to test, right? If you just randomly look for stuff, then you'll find a bunch of false positives and that'll lead you down some blind alleys. So, you know, I think this is what's going on, and then we'll we'll try and prove that mathematically.

SPEAKER_00 2:57

Yeah. Well, if anyone is not aware of what Breezeway is and and who Cliff is, well, this is Cliff. And you're gonna get to know Cliff over the next however the heck long we end up talking on this podcast because Breezeway is a it's a tool we built, honestly, out of it start it's been out of frustration, honestly. I mean, we built it out of like first it started as attribution. We're like meta's attribution is really, really bad. And so we started building Breezeway, and this is like three, I don't know, it's over it was almost three years it should go that we started. It's been a long time. It really has. So we started first it was attribution, and then we it has evolved to the point where like now we can take people's data of like we called the breeze brain in the software. It's like where it takes all of their, their, you know, their their AOV, their gross profit margins, all their metrics. It tells them their breake-even CPA, their break-even ROAS, what good targets would be. And then inside the ads manager, it even breaks down like statistical probability or you we have called a breeze score. And there's so much more. We have ad launchers, it cuts your time by like two or three X from like launching in Meta and it avoids all the dumb, annoying things that happen when you like duplicate an ad and it turns all these settings back on and AI settings. So we have been on this quest to connect. It's called the Breezeway. It connects data with confidence. And if you're unfamiliar with Breezeway, that's Breezeway in a nutshell. And so we have, I mean, how many, how many, many dollars are being managed on Breezeway right now? Do you have a ballpark? Oh gosh.

SPEAKER_01 4:23

I mean, it's in the tens of millions, certainly. I that's that's a statistic I should know off the top of my head, and I don't know.

SPEAKER_00 4:30

I guess we're growing every day. I mean, new people are growing, coming in and like so the cool thing is what what what I wanted to bring Cliff on to talk about was I want to answer this question because we've been asked this so many times. And that question is how many meta-ads should I launch? And I asked Cliff to deep dive into the numbers of this and come prepared with what he's observing from a what he's observing from like a mathematical standpoint, not just like a feely standpoint. Because I was talking with a client yesterday about this, and she's spending $2,500 a day, maybe something like that on ads. She's like, I think that I need more volume. And I was like, Well, yeah, you're launching six to seven ads per week. We need way more, we are gonna need more volume here of ads. But the question is always like, how many should I launch if I'm doing spending $50 or $100 a day or $5,000 or $2,500 to $50,000 a day? And so this is the quest that I'm on to answer with you, Cliff, today. And I'm excited to see what we come up with. So I guess the uh the question I have for you is do you observe trends in the number of ads required by spend level?

SPEAKER_01 5:42

Oh yeah, absolutely. The I mean the answer the short answer is you probably need more ads than you think. The I guess the problem that people run into, particularly if they don't have a marketing or sales background, is they hope they'll create one great ad and a and a world will beat their path to the door. But there's a reason why you don't see ads run for decades at a time, right? They they they lose their novelty and people stop watching them change channels or whatever, and they'll do the same thing in social media. So yeah, you need to continually be coming up continually coming up with novel things. The other problem is that an ad, you know, can tip typically only absorb so much spend, right? And so once uh meta has played out who it can show it to, it's gonna need something else to try and show. But it's not just the number of ads, of course, too, you know, it's also the quality of the ads. So you gotta worry about them both.

SPEAKER_00 6:34

Now you something that we've talked about in the past, and you you did this, you did this, you showed me a graph once. And I I don't know if I've told you, I've taken this and I include this in our onboarding for all new clients every week. Uh-huh. The shelf life of an ad. Okay. I when you talked about it, it made so much sense to me. And we've incorporated it into our onboarding. It's a I took the screenshot you sent me of the chart, and you had analyzed something like 69,000 ads in Breezeway. So it was 69,000 and some change, almost 70,000 ads that you analyzed. And you shared the shelf life of like how many got spend, how many ads only got one day of spend, two days of spend, three days, four days, all the way up to like 366 plus days or whatever. Or a lot of ads get zero. Zero spend. Yeah. Oh yeah. There that's zero, and then one day and two. And then it was basically how long does an ad get spent before Metadrist stops spending any dollars or even pennies on it, just zeroes out. And it the it was, I think it was eight days.

SPEAKER_01 7:32

Yeah, it's not very long for a typical ad, right? Now, what you're hoping for is to find ads that will run for weeks or months even. But most of your ads are going to die pretty quickly, and that's probably an expectation that most people don't have when they first start advertising. So what we've found, and we've kind of confirmed this with other people that have published similar statistics, is you know, only roughly 5% of your ads are going to get a significant amount of your spend. So 5% of your ads might get 80% of your spend or 85%. It varies a little bit from company to company. But the point is that the people that are successful at this have to create a lot of ads. They have to create a lot of good ads, too. It's just not volume. And that's how you win at this game. You know, it's uh there's other things you can do to tune, you know, in terms of how you configure your campaigns and your ad sets. And of course, everything else has to be set up right. But at the end of the day, you got to be really good at creating this stream of ads to test.

SPEAKER_00 8:32

Yeah. Yeah. I mean, that's that that that was always the reason we include it, by the way, is that because it was eight, it's what we have is eight days of average shelf life spent, and we use it to just reiterate to our new clients why we teach a weekly cadence of launching. Like you need to be injecting new ads. And what I see sometimes is people th look at ads as like this optional thing, and they're like, well, I'll just launch ads when I find time, and nobody ever finds time to make ads because a lot of people don't want to make ads. And so they launch when they feel like it. And it's like every two or three weeks or once a month. And I'm like, you have to treat this as a non-negotiable, the same way you wouldn't just be like, ah, you know what, we'll ship out orders like whenever we get to it. It's like, you ship out orders because it's the lifeblood of your business. And we try to tell people like the lifeblood of your acquisition and growth is meta-ads, and you have to do it every week. And so your grow your your data allows me to back that up and tell them like, and it's not just me saying this, the team at Breezeway analyzed 70,000 ads, and here's what they see eight days.

SPEAKER_01 9:34

Well, you come from a sales and marketing background, so it seems obvious to you, but you know, somebody like me that comes from a technical background or somebody that's starting a business usually doesn't appreciate the fact that they're gonna have to spend a significant amount of their resources on sales and marketing to be successful. And usually it's gonna be, you know, half of whatever it is, half your time, half your spend, you know, half your money or more to make that successful. And people, you know, just starting out don't want to believe that. They want to think, you know, well, I've got such a great product that, you know, people will discover it and it'll go viral. But that almost never happens. You've got to have that sales and marketing component. And if you're not willing to commit to that, then you're probably not gonna be successful in in meta advertising.

SPEAKER_00 10:13

Yeah, and we definitely see that in like all the the businesses and clients who we work with that are the most successful, they're just they're they're so meticulous and specific about the fact that they cannot drop the ball. And I have clients who like, they're like, I'm going on vacation. Do I launch ads this week on vacation? I like I'm not gonna be able to film anything on the week. I'm like, hey, just get ahead, or it's okay if you skip one week or two weeks a year. Like because they're just so, but they've grown as a result of that. Now, I think this is also important. The reason I want to have this conversation about how many do I need to add is because I also believe that some people they get comfortable with their cadence of how many ads that they launch. So some people, step one, be consistent in launching every week. We tell all of our, we kind of break our, we've talked about this since you've been up here, we break the ad spend level of our clients into three tiers, zero to three hundred dollars a day, three hundred to two thousand a day, and then two thousand plus. And I think a lot of people, they get in the cadence of weekly, but they get really comfortable with the number of ads. And so this is where one of the clients we're just talking about, and we had a chart we were looking at, was launching on average 6.2 ads per week, yet spending s over $2,000 per day. Right. And so the the volume the the spend volume doesn't it maybe it's inverse. The the the launch volume doesn't support the spend or the spend doesn't support the volume. It needs more volume in that. And so I think most people get comfortable in the the cadence and the volume they're doing, and then they hit a ceiling and they wonder why. So are you saying that volume that volume has to rise as spend volume rises? Yeah, it yeah, it does. I've got a chart.

SPEAKER_01 11:48

Do you want to this is a good time to show it? Let's pull let's pull it up. Yeah. Okay. Let's do it. So just to explain this chart, this is not average, these are actually companies that are successful. And a lot of these are NPM clients. They're not all NPM clients. But along the X axis here is an average weekly ad spend. And along the uh Y axis here is the number of ads that they typically launch. And you can see, particularly at the low end, it's pretty noisy, but you'd sort of expect that because these are people that are just starting out. They haven't really quite figured out, you know, how to do ads yet. Frankly, it's really frustrating when you're just starting out because you don't know how to create good ads, right? And you need you know, five, hopefully, good ads a week. And and so that that that's a challenge.

SPEAKER_00 12:32

But if you look at this screen, real fast, for anyone who's listening on audio, uh, anyone who's listening on audio, we are we're we're on on YouTube right now, we're cutting to a screen that Cliff is actually screen sharing. So if you're on audio right now and you want to give some visuals, we're looking at a scatter plot that that Cliff is showing right here, and it's dotting out the the spend the number of ads by spend volume weekly. Yeah, it's a weekly spend volume? It's weekly spend volume, right? And the number of ads they launch per week, and we're looking at a little more than a hundred companies. And what is the requirement of the companies within this scatter plot?

SPEAKER_01 13:03

Well, it's adjustable, but for this one, they have to be spending at least an average of $100 a day.

SPEAKER_00 13:08

Okay.

SPEAKER_01 13:09

And they've got it, we've got to have data going back, I think, 20 weeks. So again, this is not the average. These are these are comp this is aspirational. This is what you should be shooting for. Because these companies are successful enough that they're surviving on.

SPEAKER_00 13:22

Yeah. And I'm just looking down here, looking down here in the corner. So the very top, like it you see the line going up. So the average moves, right? Right. What is the are you able to hover over? Is this just a screenshot or can you hover over this? Yeah, no, I can hover over. What's the very top right? So the very top, top right, how much are they spending per week? Oh man, they're spending 36,000 per week, and they're doing 50 new ads a week a week. Yeah. Yeah. So like you could definitely now there are some in here that are launching a really low like a low volume. What's that very bottom one but has a decent amount of spend? Yeah, what's it?

SPEAKER_01 13:52

Right. So this guy has got 18,000 in spend and he's only doing two ads a week. Now that's possible, right? Every business is different. So in that case, and I I don't I can't tell from this who he is. I I could tell, but we're obviously not sure. We're hiding. Yeah, we're hiding the company's names. But there are businesses where you can get away with that. You know, for example, if your audience is changing all the time, then you don't need fresh ads. So if you're something like a wedding photographer or a personal injury attorney or or something where the need is kind of short term and your audience is changing, then yeah, you don't need as much of a flow of ads as, for instance, a you know, a DTC cosmetics type advertiser. Because there the ads are continually being shown to the same audience. And it's got to be new and fresh, right, for them to look at it and read your message and hopefully click. So it does vary by by business, but most of the people that we're working with, and I think most of the people that NPM's working with are in that DTC space. And so the the that line, that dotted line there is probably a pretty good representation of of what they should be shooting for.

SPEAKER_00 14:55

Is there so is there like a trend in like let's say I'm spending, are we able to see like what are the the people that are spending less than uh $1,000 a day, what how many ads on average they do? Is there any way that we could see something like that? Yeah, absolutely.

SPEAKER_01 15:08

It's it's here in the graph, right? So if you look at this yellow line or orange line, I'm not sure exactly how it's showing up, you can see it intersects right around five ads per week. So that's probably the minimum anybody should be doing.

SPEAKER_00 15:21

Five ads a week. Okay. And we recommend for zero to three hundred. Let me can I'm curious to measure this, because this is the uh this is feeling Josh. Right. Observation Josh against Data Cliff. Okay. All right. So this is actually really great. Now I went and I just filmed over 40 videos for our clients, so I hope I don't have to go redo all of it. But here's what here's what was my feelings. When you're less than $300 a day in spend, uh we're recommending three to five ads, and I say minimum because you can you can do more, sure. But three to five ads per week is what we've been recommending. And obviously diverse, not just like only change a headline. It's like, no, no, this is completely different ads every time. Three to five ads. And then once you start spending between 300 and 2,000, uh, we just kind of double the recommendation to like at least five to ten ads per week for up to 2K.

SPEAKER_01 16:05

Yeah, let's see where it kind of hits that 10 ads per week. So that was 14. That's 12, that's seven. So it's kind of in here. So what would that be if we find the closest dot? So uh, you know, according to the data here, and again, uh this is you know, data data is typically noisy, so you know, take these as guidelines. There's nothing chiseled in stone about this. But you know, it looks like about 7,500 or so you want to be at that 10 ads per week.

SPEAKER_00 16:35

Okay. So really once you start to touch over that $1,000 a day, essentially.

SPEAKER_01 16:39

Yeah. Well, I mean, the the more good ads the better, right? Yeah. So you know we've got some other charts that we can show, but it only a few of your ads are going to carry the majority of your spend. Would it make sense to pull one of those up?

SPEAKER_00 16:54

So yes, and this would be this is really another interesting concept. While you pull that up, I'll keep talking here. Sure. Because one of the questions we always get is like, well, or an obs observation and then a question. Observation is meta puts all of my spend on this one ad, and I put all my heart and soul into all 10 ads that I launched. And it's heavy weighting, and no matter how many times I inject new ads, it always keeps going back to the OG ad that has been winning, and it goes back to that and back to that. And it almost like we're like emotionally burdened by it. And my stance has always been, eh, you just gotta you have to, it just means you you didn't create something better than what Meta already had, and you just gotta get better at the thing. Right. And so that's been my feeling observations, what I've seen. So, like, how do you what's the typical is there like an 80-20 on like 20% of the ads carry 80%? Like, what's the what's the observations you have?

SPEAKER_01 17:44

Well, it's even sharper than that. You know, it it's like the bottom between five and ten percent of your ads are probably gonna carry 80 to 90 percent of your spend. And that's just because of the way Meta works, which we can talk a little bit about. But here's one, this is a this is a pretty heavy hitter. They're spending $5,000 a day. But you can see the vast majority of their ads are getting less than $200 in spend. So even lifetime? Yeah, and then when you're lifetime. So even though they're pumping out almost 50 ads a week, the vast majority of them aren't really doing anything. Now, the problem, in some sense, is worse for them, right? Because they're they're trying to achieve this diversity, which we should probably talk about. But the more ads that you create, that hard that's the harder it is to find things that are diverse, right? Yeah. Now, for a smaller advertiser, you're trying to find that one winner, right? And if it's five percent of your ads are gonna be winners, and that means we're probably gonna have to try something like 20, 20 ads before we find one. Probably more than that, because if you're just beginning, you're not really good at ads yet. Yeah. So the mentality to come in is that I've got to get really good at creating this flow of ads in order to find these winners. And the ads need to be diverse, which is so which is very hard to define.

SPEAKER_00 18:59

Yeah. And and I'd be curious, like, how do you define ad diversity? So there's like something when Andromeda came out in it was the tail end of March 25, I believe. I spent two full days. I dropped everything, canceled all my meetings. I'm like, I spent two full eight-hour, 10-hour days analyzing everything I could, thinking about it, thinking through strategy. Connor's in town, and and we're like talking through all of this stuff. And I'm like, I believe Meta wants two things, volume and variety, volume and variety. And then everyone's talking about ad diversity. And so I view variety as diversity. But like, what do you view as diversity? Like, how would you clut c uh categorize it, clarify? Like, what would you look at it as? Well, it's interesting.

SPEAKER_01 19:42

I mean, there are some things that we can point out that seem to be diverse, right? So, like, you know, the obvious things are different formats. You could have video, you could have static, and then within video, you could have, I think you guys call it a Yapper ad, you could have Yeah, styles. Right, UGC, you could have a founders ad, unboxing, right, right, all those things unboxing. But the irony is, is that you know, Meta is all about diversity. We need creative diversity. And they can't tell us what creative diversity is. They they don't, and the reason they don't, it's not because they're trying to hide it from, is because they don't actually know. And this gets into a little bit of wonkiness that you and I talked about, something called embeddings, right? And so the way Meta worked prior to 2025 was there were literally categories of different interests interests. Interests. Interesting interests, right? And you you could find ways of figuring out what all these interests were. So, you know, Josh, you were interested, you're interested in music, you're interested in running, you're interested in coffee, and there were interests for that. That somebody at Meta, some engineer or or business manager or whatever, said, you know, I want to create an interest for this. And then Josh would get grouped with that category, and then ads would get grouped with that category. And that's how Meta would know to match up your ad. Andromeda changed all that. There's no the these categories don't really exist anymore. And what it does is this thing called an embedding, which is it's kind of a bizarre mathematical concept, but we'll I'll give an analogy that'll hopefully make this clear. But basically, it computes this huge string of numbers that represent Josh. And there's a huge string of numbers that represent a coffee ad and represent a, you know, an ad for musical instruments or or whatever. And basically, it takes these two numbers and based on those, it decides this is a good ad for Josh or this is not a good ad for Josh. The thing that's frustrating about this is that there's no way to look at that group of numbers and be able to pick out this is the reason why this is a good ad for Josh. It just can't be done. Because it's just it's just too complex a problem. But it works. And so you're kind of thinking, well, how could that possibly be? Well, the analogy that I like is if you've got you know a young toddler and they're you they're learning a language, you're teaching A new word, right? You have an 18-month-old son, right? He's learning a lot of new words now. So he probably at this point recognizes a dog. Yep. Maxwell. Maxwell. Okay. Maxwell's the dog, right? Your son's William. Anyway, so I mean, that I imagine this happened over time, right? The first dog he probably recognized was Maxwell, but then he began to recognize other dogs. Probably you'd point and say dog, and he would repeat back dog. And after this happened dozens of times or maybe hundreds of times, it got to the point where he could point to a dog that he's never seen before and still recognize it was a dog. Right? I mean, he knows he knows what a dog is. Well, if you were to ask him or ask anybody, his pediatrician or neurologist or anyone, what's the process that he's doing to recognize that dog? Nobody could tell you. Because it's buried deep in a human neural network, if you will, for that process. And you can't bury that out. I mean, he's not saying, well, it's got a snout that's this long and I can see his canine. I mean, it's none of that, right? There's something else going on that we can't describe. And in effect, Meta is doing a similar sort of thing, right? It's looking at all these inputs, features, and it knows that if I have these features, it seems to work really well with people like Josh that have these set of features, and that's how it's serving up your ads. So the frustration for an advertiser is we really can't back out what that is. You know, what makes it creatively diverse or so that it'll do Josh and it will do Cliff and it will do, you know, whoever else. The only way to figure it out is to test, yeah. Is to try different stuff. And that's why advertisers need to be continually trying different stuff. Because there really isn't a way we can back out why it is that Meta thinks this is creatively diverse. There's no way Meta can do it. It's just it's buried too deep inside.

SPEAKER_00 23:50

And you have now you've explained this to me and you said like it's beyond 3D. You're like, this is embedding is multi-dimensional. Right. And million and and meta runs these billions of things, what is billions of just times a day, it's doing some kind of analysis.

SPEAKER_01 24:06

Billions of times a day, it's doing these calculations. Yeah.

SPEAKER_00 24:08

It's doing these calculations. And and so what how I kind of look at this is is like, so I'm going to use my son William, who's 18 months old. He d he does understand what a dog is now. And he at first he just like he would call Maxwell meh, and it was just meh, meh, meh. And then it became dog, dog, dog, dog. Then it became Maxwell, Maxwell. So like he almost like his it sounds like am I wrong in thinking that his experience has shaped what he believes it is, or it shapes his understanding, but everyone's experience is different in how they how they view and perceive and what they like and what their interests are. And so are you saying that Meta is able to learn base it's able to pull in all of the data that it has on somebody that it has this multidimensional aspect to it, to where it can say, based on all of these things that are so the thousands of dimensions of this user, we can present this ad based on all this data, and not even we can tell you why it ends up presenting the ad. Yeah.

SPEAKER_01 25:11

So to go back to the example of William and Maxwell, right? I assume William at this point recognizes Maxwell, also recognized Maxwell as a dog. Yes. He probably recognizes other dogs, but he also knows they're not Maxwell, right? Yes. So so he's he's developed a certain level of sophistication. So the way this has become possible in Meta is two things. First of all, you need massive amounts of computing power, which has obviously happened over the last you know 10 years or so. You know, everybody's heard of NVD and those kinds of companies providing the hardware that's really good at doing these kinds of calculations really fast. But the other thing is you need is massive amounts of data to train these neural networks. So, you know, in LLM, they basically sucked every piece of data that there was on the internet. Meta had its own data, right? It it knew what people responded to which kind of ads, and it fed that into this model to figure out, okay, well, if if we can identify these things about this person and we can identify these things about this ad, it looks like there's a pretty good chance that this ad will resonate. And that's basically how they matched them up. But the only way they were able to pull it off was because they just had huge amounts of data. And they didn't suddenly start doing this in 2005. I think they they started working on this like five or six years before, you know, and they're kind of warning advertisers about this. You know, stop using categories to target. You know, start targeting using your uh your copy or your creative or whatever. And that was in anticipation that they were going to do this. And they claim it works a little bit better than the category system, but yeah, the truth of the matter is it it makes their operations much more efficient. They don't need to have people invit constantly trying to figure out what categories to invent or changing their code to accommodate more categories. It's all done automatically now through their learning algorithm. And you know, that allowed them to lay off several thousand workers, I guess, uh a month or two ago.

SPEAKER_00 27:03

So I guess one thing that I I think about when I think of this. So like maybe I'm listening, I'm like, I don't I still don't understand what embedding is. Well, it sounds like meta doesn't even fully understand how everything works and they just call it embedding.

SPEAKER_01 27:16

Aaron Powell Well, they understand what embedding is, but but you can't look at the embedding and pull the information back out. And it's not just meta, that's the way these neural networks work in general. So, for example, you know, one of the big breakthroughs that people saw occur when this machine learning stuff or AI uh occurred was they talked about radiologists, that the pattern recognition that AI could do was as good or better than somebody that had a medical degree in radiology and or medical specialization in radiology. And so, you know, you present these images to this AI and it says, ah, that's cancer. And it's pretty good at predicting that that's cancer, but you can't ask it or can't back out how it knew it was cancer. Because it's just it's it's in this neural network that's got thousands, tens of thousands, hundreds of thousands of parameters, and there's just no way to have it tell you how I knew or how it knew. I mean, I'm anthropomorph anthropomorphizing it, but how how it could tell that that was cancer. The same way that William can't tell you how he recognizes a dog. It just does.

SPEAKER_00 28:20

So when I think of all of this, the thing that I think about is like I think this is just telling us it's it's less about how what we believe meta wants. It's like it's like it we shouldn't, we should stop trying to interpret all of this stuff and just know that Meta, we need to let Meta interpret the stuff that we give it to in pretty much give up control and let it have the control of what to do. And so essentially it's telling us we do need to test a lot. We need a lot of volume, but we also need a ton of variety to fuel meta the ability to go and embed and go find all of these different audiences. And that's why we don't just want to change a simple headline into creative, because that may not tell a whole lot of actually we it may not tell a whole lot of what to run with. There's not enough contrast between that. So it sounds what I'm gathering is like even if embedding is still it is still kind of fuzzy to me, uh, I might have to spend more time on this. It's telling me that I just need I do need more volume and I do need more variety. And it's just I need to trust Meta to take the thing and go find the right people.

SPEAKER_01 29:28

Yeah, you don't want to fight the algorithm, but I guess the point was is that you you know, you know, what what what you'd like is Meta says, well, tell me what I need to be creatively diverse. And it can't tell you because it doesn't know. So you've got to figure that out on yourself. You've got to try enough different things to try and you know decide, well, what's create what's diverse? And we you know you talk about that all the time, different angles and different formats and and different surfaces.

SPEAKER_00 29:51

This is even, and I'll even share this. So like I had a conversation two days ago, we're on a coaching call, and I had two questions come up. One question was, Lee, shout out to Lee if she's listening to this, she's amazing. She's a jewelry business, and we're talking about calling out ICPs. And she's like, well, when you're you used to talk about like, hey, write ads for this ICP, and my perspective has changed because we used to like use an ICP as a lens of create building a creative. And that can be useful. Like I could say, hey, I want to talk to professional cleaners versus consumer people that just need cleaning products. If I if I create my ads through the lens of professional cleaners, it will help with that. But and so in some scenarios, looking through the lens of ICP makes a lot of sense. But in other scenarios, let me hop to supplements. I could talk about a use case or a pain point, and I believe that this is for this ICP. I could say, oh, I'm gonna talk about joint pain pain. I'm gonna say this is I believe that this is through for the ICP of people that are over the age of 50, I don't know, joint pain, right? But that joint meta knows that joint pain is actually a can uh a challenge or a pain point of way more than just that ICP. Right. And so I am more of a fan of saying ICP is a good lens to consider when you're coming up with unique angles and creative, but I'm not trying to like be like, oh, these ads are for this ICP. It's like, no, no, no. Meta will find who the heck the the ads are for. And the same goes with top of funnel, middle of funnel, and bottom of funnel. The conversation shifted then to someone else who's saying, well, how do you view it? I usually build ads with top of funnel for these people, middle of funnel, bottom of funnel for these people. I'm like, how do you know that the message that you think is middle of funnel is actually top or bottom of funnel for somebody else? Because like I don't know where people are at in their learning journey. And I'm making an assumption that I actually know that this messaging is 100% for this person in this stage, right? But it's like that is very like linear of my brain and how I would my experience would view middle, top, and bottom of funnel. Right. But that's not the experience of someone else or where they're at and how meta would interpret it. What are your thoughts on that?

SPEAKER_01 32:07

Well, I I I think all those are good things to try, right? You gotta, I mean, you can think of different ICPs, you can think of different funnels, but don't assume that that's how Meta's thinking about it, right? That that that's how those are good ideas to try, okay? And and then and then act on the feedback. One of the concerns I would have about top-of-funnel ads is, and this is a little controversial because not everybody believes this, but I I think the data supports it, is that when Meta enters, shows your impression to someone, it's trying to close the deal. So if you're running four sales, it's doing that based on the probability that it thinks it's a good thing. Right. Expected expect action rate.

SPEAKER_00 32:53

Yeah.

SPEAKER_01 32:53

There's not the meta doesn't know about I'm gonna prime this guy today and then tomorrow I'm gonna show him another ad to close the deal. There isn't any mechanism inside Meta that does that, at least that I'm aware of. You know, I've spent a lot of time reading their documents. If there was, you'd think they'd tell us about it. But there's a lot of influencers that seem to think that that's the case. And so if you're doing top of funnel, which is fine, don't depend on Meta to create that funnel. What you're hopefully gonna do is make them problem aware, and then by getting them to sign up for your lead magnet, you can you can get them into your funnel with email or whatever. Because what Meta's gonna do, it's gonna say, well, this guy's interested in this particular problem. And yeah, they might show another ad that you developed to this guy the next day, but they're gonna show ads from all your competitors too. You know, they're they're not building a funnel for you, right? They're just trying to maximize their profits in terms of you know, showing the the impressions that they think are most likely to convert. So Yeah. So I'd be I'd be careful thinking about top of funnel. I think it's fine to sh to have top of funnel ads or to think in terms of that.

SPEAKER_00 34:00

Yeah.

SPEAKER_01 34:01

But don't depend on meta to create that funnel. That's something you're gonna have to do.

SPEAKER_00 34:04

Aaron Powell And I think that's something that I have learned that I've kind of truthfully gotten wrong over time. Because I remember when we first started talking about Andromeda well over a year ago. Right. You and I are meeting, we're looking at this stuff, and I'm we're talking about okay, there's Andromeda, and then there is just two years ago. Yeah. Was it what? Was it two years ago? It was in San Francisco, right? Oh my gosh, we've been talking about this forever. I'm sorry, I just forget. It's just only got popular in the last year, it feels like. So, like, so we're talking Andromeda, we're talking oh she what's the M or what's the what's this second? There's sequence learning. What's the other one? Oh blink. Lattice. Lattice, lattice.

SPEAKER_02 34:38

Yeah.

SPEAKER_00 34:39

Lattice, and then there is uh gem, which is the generative ads model, which is the brain. So when we're talking about all these things, uh sorry, we're geeking out. We're nerding out here for uh for any of you that are like, what are you guys talking about? So when we talk about sequence learning, when I presented to you, I said, I believe this means that Meta can predict the sequence of ads in which somebody needs to see in order for somebody to then take an action. Where I inject a ton of ads and then maybe meta views this one is top of fun, this is like someone who's unaware pro you know, unaware uh that they have a problem or unaware of a solution, whatever it is. And it will identify this is an unaware for this type of person, fuel it to them, fuel them middle funnel, then fuel them the bottom of funnel. And I remember I presented that to you, and you're like, eh, Meta's not that good. Meta can't predict what is top of funnel, then fuel them this and know that they have to see this exact thing in this modality or on this not placement surface, right? On this surface of what is what they call it. I presented that to you and I was like, this is what I'm saying. And you're like, Meta's not that good at predicting. And I'm like, well, let's see how this plays out. And then over time, we we did learn Meta's not that good at predicting. Right. But the only thing I would say is there are ads that I would consider, from my interpretation, as more top of funnel that fuel much lower cost per click, that does get more volume of people into the ecosystem when meta like finds the ad. And usually Meta will latch onto an ad that gets a lot of engagement first. That's just observation there. I don't know if that's truth or not. But if Meta gets a lot of clicks, it usually a lot of times prioritizes for a certain period of time, and then it weans off of it once it realizes the purchases aren't coming through. But it's fueled a lot of people into the ecosystem that buy from other ads. So we have this so there I do see that sometimes happen. But at the end of the day, you just made a good point I never thought about. And that is that if you're optimizing for lowest cost, highest value purchase conversions on a day, Meta's just looking to spend your budget fully every day and just get the lowest cost conversions and then most people to buy as possible. So it is kind of I never thought of it like that. That what you're trying to do top of funnel when you're when Meta is built to optimize for estimated action rate. So why would you try to intentionally feed it to something that will have a low estimated action rate? Meta's not trying to do that. It's trying to get you the end goal of a purchase. That's an interesting thing.

SPEAKER_01 37:00

I mean it is a problem, but particularly if you're you're selling something that's higher AOV and it's not an impulse decision. You know, they may have to see an ad a few times, they may have to think about it before they actually make a per a decision. So I'm not saying that that can't happen, but uh it it's not something that Meta optimizes for. It's very difficult to prove a negative, right? I can't prove something doesn't happen. But the and you're certainly not the first person to interpret their sequence learning paper that way. A lot of people did and continue to. But if you read the paper, first of all, it's a marketing paper, it's not a technical paper, right? So you got to kind of read between the lines. And the biggest problem was they use sequencing in a different context in earlier stuff that they published, right? Where sequencing was a sequence of ads that that people saw. But if you read that paper carefully, they're using sequencing to avoid saying tracking, right? Because what they're what that paper is really about is keeping track of what you're doing on Meta. The sequence is the sequence behind you. Well, it's a sequence of things you're doing. Which ads are you clicking on? How recently did you click on this ad? That's what it's keep that's the sequence learning that it's doing, and that's what it's feeding into the embedding, right? This huge string of numbers. So but that's the way, yeah, that we interpret sequence learning. I've got, you know, guys on the Breezeway staff, but we've got at least one guy that's actually got a degree in artificial intelligence, and he, you know, he reads it the same way. That it's not, it's not the sequence of ads that are being shown, it's the sequence of actions that an individual is taking. That's what sequence learning is. But if they called it tracking, then what then you know Zuckerberg would have been asked for it, asked about it the next time you appear before Congress. So yeah. So they went back and said, They went back. Sequencing is a better word. They banned that tracking word, right.

SPEAKER_00 38:47

And they and they and they use sequencing. But people don't like being tracked, but they don't care if they're sequenced. Yeah. Yeah.

SPEAKER_01 38:55

Exactly.

SPEAKER_00 38:56

So this is kind of interesting. When I think of embedding and I think of like volume and variety, because I do want to come back, by the way, because the question we're trying to answer here is like, how many meta-ads should I launch? And I do want to come back to that. And I want to give some like some directional understanding for anybody listening to to close the loop on that. And then I have a few more questions after that. Sure. So with that said, how many ads, if I'm spending $100 a day, how many ads should I be launching? I would say you should be doing five a week.

SPEAKER_01 39:28

Five a week. And yeah. And your expectation should be that it may be five weeks before you find one that, you know, really dominates and captures the majority of your spend. And that frustrates frustrates people when they see that, but that's actually a good thing because it means that Meta has decided that this ad is materially better than everything else that you have, and and it's finding an audience for that. Hopefully it's finding a profitable audience, right? If it's consuming 80 of your sent 80% of your spend and you're not profitable, then that that's not a good thing. And it will, in fact, do that. It'll it'll decide that, you know, I'm gonna spend the money here because that's that's the best option available to it.

SPEAKER_00 40:05

Yeah. We we call this, so I have this as like a quadrant in our our playbooks for our clients. And I look at it as meta has a vote and the customer has a vote. Meta's vote is essentially how much meta once spends on it. So if meta puts a lot of spend on it and the c customer vote is do customers buy it? So if something gets a lot of spend and no purchaser, so a lot of meta vote, no customer vote, then that's in the bottom right quadrant. And that is like something that you don't chase, but you learn from. Because if meta likes it, there's something about it. We just need to raise the intent of the thing to try to get to the upper right quadrant, which is Meta loves it, spends a lot of percentage, a lot of dollars on it, and your customers love it, has good performance. That's a unicorn, and we want as many of those as possible so we could really scale. So I I I that makes a lot of sense to me. So let me tell, let me ask this question then. Let's in this bucket of I'm spending $100 or less than $300 a day. Sure. What do you say to someone who is stuck at a certain spend threshold, maybe it's even beyond beyond the 300, someone who's stuck at a spend threshold? What do you tell them?

SPEAKER_01 41:11

If they were stuck at a spend threshold, then probably, you know, their I would guess that their ads are limited in the amount of, you know, good viewers that Meta's finding for them. So they really got to try and figure out, you know, is there another angle? Is there another audience? Is there, you know, different platform or something that I haven't tried that I can try? And you know, you've got lists of those things and you go find list of things on the internet to try different formats, to try different formats is actually pretty easy, trying different angles, right? So you know, what is the reason that someone's gonna, you know, part with whatever, whatever the cost of your thing is, and can you come up with different ones? The real sad case, and you know, we've both seen this, is somebody finds an ad that just blows it out of the park for for months, even, and then just dies, right? And the the emotional incentive is to blame Facebook, right? And it's like, well, no, it's not Facebook. It's like, you know, you showed that ad to everybody or almost everybody that would be inspired by it, but you need to find something else. And and so that's why you got to be continually trying new things, putting them out there, and hopefully Meta can find, you know, other pockets of people with these embeddings that will find those ads compelling.

SPEAKER_00 42:30

So you believe that if you hit a threshold in a ceiling, it might be in let's say you're trying loads of variety and diverse styles of creative to a particular hook and angle. Are you saying that you just believe that pocket of the audience for that particular angle might be tapped and you need to explore different angles for those audiences, or maybe consider building ads for different types of audiences?

SPEAKER_01 42:53

Or both? I think both. I mean, if you're consistent, if you if if you're at a consistent place, right, you're you're you're operating profitably at a $200 spend or something, and you can't get above that, that's not necessarily a bad thing, right? You got some breathing room to go experience some stuff, but you know, take some some wild leaps to try and get away from you know whatever your core ad is. You know, look look at what other people are doing, look at what your competitor's doing. Is there something you can do to make it different than what your current ads are and and test that? But yeah, constantly be looking for the new thing or the new new way to present this thing. Because ideally, right, Meta will find a couple of different ads and will be going after different people, and that's gonna, you know, result in Meta having a much easier time finding people that want your ads, and you're gonna be able to increase your spend and scale at that point.

SPEAKER_00 43:46

Yeah. I would say that like so I did an episode, a podcast episode. Anyone who's listening, don't quote me on this. It was audio only, and it was maybe like w one or two podcasts ago. And I did it on how we identify the ad testing ladder. It's Like I first look at what are angles that resonate, which are like this overarching umbrella. Sure. And then I go into hooks for the angle. And then I go into once I find a hook, we take that and run it wide with a ton of different styles. And then we'll shift format. So we usually start with image, climb the ladder. Then once we have proven hooks and all that, then we move over to videos that say the same thing in a hook, say that hook in a different way. And so that's how we've found we could continue to get variety. And it makes a lot of sense. I'll even share something that's worked really well for us. And I and I just included this in the newsletter that we sent out today, is that calling out ICPs has been working really, really well. And it's something new that we try that, like we've talked about, you know, in e-commerce alley, most many people are burned by agencies. And so we've used that language forever. And then when we decided to simply say, let's call out specific industries like that have been burned by agencies, even that layer was different enough that Meta loved it. It cut our CPA almost in half for the last four weeks. It's been crazy where it's like, hey, scale with meta ads minus expensive agency fees, scale your apparel brand, scale your tallow brand, scale your coffee brand. Like we had like all of these different, we started exploring wide, and Meta just loved it. Yeah, which is crazy.

SPEAKER_01 45:14

That's brilliant. That's a really good idea. Because right, you're not really offering anything different to the apparel brand than you are to the coffee brand at the end of the day. But Meta is using that to try and reach out and find, you know, different people to show it to. And also it's targeting apparel and it's targeting coffee. And some of those might be cheap, right? That yeah, or cheap's the wrong word, they might be efficient or or or profitable for you.

SPEAKER_00 45:37

So yeah, no, I think that's a that was a fantastic idea. So like I just the volume and variety, I just think that like if you I would say if you're feeling stuck, just look at all your stuff. I'm willing to bet a lot of it, a lot of it has to look pretty similar. Because met, I mean, if it was that diverse and good, meta would be allowing you to have high performance to sustain at a higher level. If you feel stuck, my hunch is you might just be deep in an angle, a hook, a style, or a format that is like maybe not format, but like a particular thing that is just too close to each other that meta's not seeing the distance between the ads too much. Right. And it's just not doing it.

SPEAKER_01 46:12

Yeah. So if you're not getting a lot of spend or maybe even zero spend on an ad, that's likely because Meta says, well, that's just like this other ad. So it's computing this embedding, and these two embeddings in, you know, we can kind of interpret in 3D space, right? They could be far apart or they could be really close together. If they're really close together, Meta says, well, these are basically the same thing. So I'm not going to dilute the signals that I get on either one. I'm going to pick one and go with it and not put any money on this one because from as far as Meta's concerned, it's the same. So yeah, if if you're seeing that, you're continually launching ads and the the same one is always getting all the spend, or the similar one is always getting the spend, then yeah, you haven't really found that diversity lever as far as meta is concerned. And I it's definitely frustrating because Meta can't tell you what it thinks is diverse. So it's up to you to figure it out.

SPEAKER_00 47:00

Well, I'll I want to share this from findings, and then I want to hop to the next two tiers of like, hey, how many if I'm spending a little more, how many ads should I potentially be launching-ish? Uh because it's obviously, like you said, directional guidelines, like these aren't like written in stone, it'll it'll vary. But we did this test with a client of ours. We took, we wanted to talk to different ICPs because we found this working for us. We're like, we want to show you this is working for us. Let's try this. So we went ahead and we used the same background image. We did a static run, same background image, and we just changed the text on the front of it to call out different ones. Now, by the way, ours has a different image for each the the text it structure is the same minus the industry in each ad. And then but we did change the image of the products on it. And so we tested the same background image and we just called out different ICPs for her. Meta put out of the 10 ads, I or 10 ads, I think, nine of them got not even one penny of spend. Really? Okay. Not even one penny, which was like extreme. And it just lashed onto one, ran with the one. Well, if it didn't put one penny in it, then yeah, and the algorithm said these are all the same ads. So we were so what we're trying to say is does Meta look at headlines as diversity? And that showed it actually doesn't use headlines for diversity. So we're like, okay, which was my assumption, right? It's like iteration. So, okay, so now we're gonna keep the same headlines. All we're gonna do is we're gonna tailor the background to the ICP and what's front. Suddenly we launched those, we just launched those the next batch of ads, and then everything else started getting a little bit of spend, and then of course latched onto the one that saw the most potential. And so it was just it was showing me that it's like, yeah, meta like it uses the text maybe to help target, right? But it doesn't actually view the diet, the creative is the diversity. That was just a really isolated, we're like, let's just do this same exact image, different text, meta didn't care. But we did different image, same exact text, and it was able to go find those types of people and if you and then it started dil delivering some spend until it found the one. Right. And I thought that was so interesting.

SPEAKER_01 49:04

That is, I mean, it's really counterintuitive, right? You would think that it would have a much easier time interpreting the text and and figuring out how to diversify across that. But yeah, I mean, the the these AI models can, you know, suck in a creative, suck in a a video or a a static image and calculate its embedding and discover that, well, these people like red or these people like, you know, bucolic outdoor scenes or whatever. Yeah. And build that in, right? To to to figuring out how to how to match it to. Unfortunately, like I said, it it it'll never tell you that it did that. So we're kind of flying blind.

SPEAKER_00 49:41

Hmm. Sorry, my phone's starting to go off. I normally have my ringer off here. But let me just say this. So, like, so if you are like under several hundred dollars a day, you're saying five-ish is a good rule of thumb, what you're at least saying. Sure. What examples do you have for maybe some of those thousand, two thousand dollar a day spenders? What's the average you're seeing that the successful ones are launching per week?

SPEAKER_01 50:05

Okay. Well, we looked at that before, right? We looked at 10 ads. It looked to be, and again, this is just looking at the data that we've got. So this is these are companies that are successful, and this is what they're doing. But you can see you're showing that that's that it's pretty noisy. But there's definitely a trend here. So it looks like you know, around $7,500 a week, it's about 10 ads, right? So let's figure out what was the number you wanted.

SPEAKER_00 50:31

So a thousand, maybe maybe two grand a day. Okay. What do we have around that range? So 14,000 a week. Let's see. Four fourteen thousand a week. So it looks like that's gonna be about fifteen ads, something like that. Fifteen ads. So almost it's like you go from zero to a couple hundred dollars a day, really, five is good as you start to find the winners. Right. And then once you find winners, you know, you now are gonna have to let those ride as you continue increasing volume to find more to be able to support the level of spend, right? Is what I'm gathering here. Yep. Because there is a level of volume required to support a certain number of spend. And then give me something like five thousand a day. Five thousand a day? Yeah, what do we have for like five K a day?

SPEAKER_01 51:14

Uh it's not much above the five. It's probably seven, something like that.

SPEAKER_00 51:18

So yeah, 50% more than that initial five a day. Well, I'm sorry, five thousand in spend a day. So it's like thirty five thousand, thirty thousand a month a week. I was giving you, yeah, the the weekly spend. Yeah.

SPEAKER_01 51:28

30,000, yeah, it starts to get up there. It looks like we're probably 25 a day, something like that. 25 ads a day.

SPEAKER_00 51:34

You're talking about it. 25 ads a week. 25 ads a week. So I'm just recapping. I'm just walking the line because I know that there's somebody listening in each of these categories, and I want to just say several hundred, you need to aim for like five ads a week, and that's a good rule of thumb. You start to push a thousand, ten ads a week is a good rule of thumb, or at least what we're seeing, two thousand. So when you double, it's not like you have to that's 15 ads a week. And then when you start pushing five thousand a day, you're talking twenty-five ads a week. Aaron Powell, there's actually a formula here.

SPEAKER_01 52:01

I don't know if anybody can actually read it, but it's it's basically three quarters of an ad per thousand dollars per week.

SPEAKER_00 52:07

Aaron Powell Three quarters of an ad per thousand dollars a week.

SPEAKER_01 52:11

That's an interesting role of probably should uh yeah, we probably should normalize on the ads, right? Ads per $1,000 per week.

SPEAKER_00 52:17

So But there is a minimum floor though, because if you're spending $100 a day, you don't want to spent make .7 to five ads a week. Yeah, there is a minimum threshold. I'd start at five.

SPEAKER_01 52:27

Because you're g because at that level, you gotta get good at ads, right? You've got to do your reps, I think is a way that I hear you put it.

SPEAKER_00 52:32

Yeah.

SPEAKER_01 52:33

Because you're not gonna be that great at ads when you start out. It's a skill that you gotta learn, like everything else. And in just cranking them out, you'll get better at it, you'll get faster at it. Hopefully you'll you know discover some tools like Breezeway that'll make it easier to launch ads. And that's a skill you're gonna have to develop because again, you know, kind of how we started out this conversation is the people or most of the people that are successful on meta advertising have figured out how to continually create a stream of new ads to test.

SPEAKER_00 53:03

Yeah, yeah. Well, Cliff, I appreciate you being on here, man. Well, I appreciate you having me. Yes. And I just want to say anybody who's listening, um, if you don't have Breezeway, go get a trial. Breezeway is incredible. We use it, we survey our we work with a little over 200 e-commerce clients right now, and probably 50%, uh, maybe a little less than 50%, 40%, 50% have Breezeway. We survey every quarter, we ask the number one tools that they use. And Breezeway is the number two used tool to Shopify. It is the number two most logged in and used tool of any client we have that is using Breezeway. And so if you're not using Breezeway, it's going to make your ad management a breeze. It's going to connect your confidence to data so that that way you can make really good decisions. On top of that, we have the ads launcher, which will literally cut your ad launching time in half, at least. And it is way less frustrating than the ads manager. It's easier, it's intuitive. We even have pr some pr some things pre-selected on the AI enhancements that we recommend to make it a lot faster. It gives you a CEO dashboard that shows you data so that you can look at your business from a high level and it pulls in everything from like your Amazon into there, it pulls Google Ad spend, it gives you blended ad spend percentages. It allows you to make good decisions inside your business, even beyond only meta ads. And so if you have a benchmark tool, we have a bad day detector built into it. The benchmarks will show you against others for like AOV and conversion rate and ROA ads and CPA. It gives you all these metrics again uh against the benchmarks to show what metrics you need to be working on in your business. And so Breezeway is obviously built meta first as a priority. However, there's a lot of stuff that you will get out of it beyond just meta insights. And so thank you for listening to this. If you found this valuable, go ahead and drop a uh drop a comment. If you're listening on YouTube, what was a takeaway that you had? We would love to hear it. And also, if you are on iTunes or your or Apple Music, I guess it's called now, or Spotify, please go ahead and leave us a review. A little bit of rating, it means the world to us is we're on a quest to become the number one e-commerce podcast on the planet. Cliff, thanks for joining me. And everyone, we'll see you in the academy. Go create some ads, everybody. Let's make some ads. All right.