w/ Kurt Bullock, Produce Dept
Also available on YouTube: youtu.be/m9YTxRsosJA
Meta just rewrote the Facebook ads playbook—and Shopify merchants need to catch up.
In this episode of The Unofficial Shopify Podcast, ad strategist Kurt Bullock (Produce Dept) breaks down how Meta’s new AI-driven ad system, Andromeda, is reshaping targeting, delivery, and scale. From embeddings and vectors to Advantage+ campaigns and creative testing, this is the update your ad strategy needs.
Whether you’re running Facebook ads yourself or just trying to understand what your agency’s talking about, this episode gives you the clarity—and tactics—to keep up.
Sponsors:
Links & Resources:
Work with Kurt:
Kurt Elster
This episode is sponsored in part by ViralSweep. When was the last time you ran a sweepstakes or giveaway? If you're worried about staying legally compliant or figuring out fair winner selection, meet Viral Sweep. The best viral marketing app for Shopify stores. With ViralSweep, you could build, run, and manage all kinds of campaigns that drive traffic and boost your brand's visibility. Want to engage your customers? Use ViralSweep's Instagram giveaway picker, or save yourself time with automated winner selection and prize distribution. ViralSweep also ensures your promotions stay legally compliant so you can run campaigns without a hitch. Plus, you'll get impactful analytics and seamless integrations to keep everything smooth. Need help? Viralsweep support team has you covered with options for managed services too. Supercharge your customer engagement today. Download ViralSweep on the Shopify App Store and start creating contests that grow your brand. New customers can get 50% off of Viral Sweep's annual plan. Just tell them in the live chat that the unofficial Shopify podcast sent you. That's viral sweep. My friends, it occurs to me it has been a while since we've touched on Facebook ads. And how can you discuss e-commerce online stores selling with Shopify without talking about Facebook ads, meta ads, Instagram ads, because that is typically one of the the number one traffic sources for paid acquisition for a lot of stores. And like any other software product, it evolves continuously. They're always adding new features, changing things. But man, is it complicated? It has been years since I have logged into Facebook Business Manager. And you know what? Ideally, I would prefer to keep it that way. So we need an expert, someone who can help educate us on, you know, what's new, what's fresh with uh with Facebook ads. And to do that, we're gonna have uh longtime recurring guest, Kurt Bullock from Produce Department, who runs uh a Facebook ad agency, meta ads agency that specializes in e-commerce at Shopify. Kurt Bullock, welcome back to the show.
Kurt Bullock
Thanks, Kurt. Thanks for having me.
Kurt Elster
I am happy to have you here. Uh and just to be clear, since we're both here and on video, uh despite having the same first name, we're different people. I did I did have a prospective client once who was like, oh, I thought that was that was you with a pseudonym. Like what? That doesn't make any sense
Kurt Bullock
Whenever I'm on email threads with you, I always put K KB or Kurt B, you know, just to keep the confusion level down.
Kurt Elster
Yeah. It's like, you know, if we were both named Brian, I don't think the same thing would happen, but oh well. All right. Totally, totally unrelated. Uh okay, so how long have you been doing Facebook and e-com?
Kurt Bullock
I was just looking back. I think it's been about 10 years that I've been doing it for e-commerce.
Kurt Elster
Any notable clients in there?
Kurt Bullock
Um, I mean, some of my favorites have been ones that I've uh partnered with you on. Like Hoonigan has been, you know, an amazing client. Yeah, Hoonigan was just a lot of fun. Um Um, a lot of the brands that, you know, I as I got into this and um started working with different clients, there's a lot of like dark horses, people that are spending a ton and doing a ton of revenue just, you know, stores that you wouldn't uh know to look at. um spending 60, 70, you know, $80,000 a day on on on meta ads, you know, and you would never know. One of them that I worked with was selling uh fruit online. It was wild. I d didn't know it was possible.
Kurt Elster
I mean those levels of ad spend would make my eyes water. But yeah, I'm not in it every day. I'm not I'm not dealing with the ad spend. Um, you know, I'm I'm in the online store. I'm like, just bring me the qualified traffic, please, and then you know, we'll try to convert them. So, okay, doing it 10 years, Facebook ads, harder, easier, the same, different. What's your take?
Kurt Bullock
I would say it's definitely different and it's a little bit harder. Um, it used to be that um, you know, it was more technical media buying. It was all about getting in there, finding the right audience, you know, let's try this different look-alike audience. Uh let's try this different time window for this retargeting audience. Most of that has gone by the wayside now. And it's mostly about your creative. It's not so much about finding pockets. um yourself new audiences, but it's helping the algorithm to find those pockets. And so it's a it's sort of steering the algorithm differently than you used to. Before it was more direct and uh now it's more indirect, you know, through your creative.
Kurt Elster
And in the past, you know, we would get so oftentimes we could get better results by dialing in, you know, targeting, doing interspace targeting, defining the demographics based on what we knew about our customer. It seems you know from the outside looking in, it seems like today it's much better to feed the algorithm and give it as many positive buying signals as you can for it to work with, and then let it do the actual audience selection.
Kurt Bullock
Yep, yep. And that's becoming even more the way with Andromeda.
Kurt Elster
Um, what's Andromeda?
Kurt Bullock
So yeah, Andromeda. Um it's what what Meta's calling it's their latest It's their ads retrieval and delivery engine is sort of how they refer to it. And it's using AI behind the scenes in ways that it wasn't before. Um it's so it's essentially the model behind what you see, when you see it, and why you see it. Um and that's sort of all Andromeda. It's a combination of um these massive investments Facebook's pushing putting into hardware. Um, and then there's a model and systems component uh to it. But uh just this year alone, I think Facebook is slated to spend $65 billion on AI. They're opening a new data center. I can't remember where it is, but it's four million square feet, which is like 27 Costco's lined up side by side, just of racks of servers.
Kurt Elster
And this is all to run, you know, uh surely not just Facebook and Instagram, the app. This is for Llama, their big uh AI, their LLM, right?
Kurt Bullock
Right, right. So they're working with um, you know, specifically uh as it as it pertains to like just the ads, they they've put together um a partnership with NVIDIA. And so they're using this like Grace Hopper chip combo, which I think is a CPU and a GPU combo. Um, and then they've got their own purpose-built chip. And it's all uh changing the way it was really built because, you know, Facebook sees the number of creatives exploding. Uh And they're enabling that themselves with these AI enhancements. You go into ads manager now. It doesn't let you just publish an ad without asking you, you know, if you want to have AI enhance it 10, 15 different ways. Uh and if you turn those all on, then each of those is a new ad variation. And so the number of ads uh is just exploding. And then Facebook is really building everything in the direction uh for that to happen. They're encouraging that and saying that that's the way forward is. uh increasing the number of ads. And of course it's not just about number. We'll talk more about that. Quality's obviously gonna be important and you have to have to have some strategy behind that.
Kurt Elster
So it sounds like more things are happening automatically for me here. Is this part of it the thing I I hear a lot about, I see argued and discussed is um something called uh advantage plus campaigns. What the heck is that?
Kurt Bullock
Yeah, so they've it they've undergone a few changes. Uh it it's still the same acronym, but now it stands for something different with this late. So it used to be Advantage Plus Shopping Campaigns. Uh now with Andromeda, the latest model that's come out, it's Advantage Plus Sales Campaigns. Um and basically it's their uh they call it their full funnel campaign. So the idea being that before you might have sort of top, middle, and bottom of funnel audiences broken out into different campaigns. Um and uh You know, segment it that way. Now with this full funnel campaign, Meta gives you these segments. Um, so you can go in and you pre-define uh what a um engaged audience looks like. Um and so you could say, hey, that's my clavio list, that's my website visitors, that's people that engage with the Facebook page. They let you uh set up your customer list. And then when you run any campaigns now, you can do a breakdown, sort of with this full funnel approach, right? They give you visibility. So you can see, well, how many of these customers the converted are new customers versus repeat customers? uh versus people from my engaged audience. And so that's all possible in one campaign now. Uh and that's really the, you know, now when you c if you're on the latest version and they're always rolling, you know, I have three different versions of ASCs in three different ad accounts. But If you're on the latest version, uh there's not even an option to select something other than this ASC if you want to do purchases. It's just going to be the de facto campaign type.
Kurt Elster
So is this is this the default if I'm setting up ads myself, I should just go straight to this?
Kurt Bullock
You know, it is. And like I said, there's two different types now. So If you're in the latest version, you won't even have an option. It'll just put you into the ASC. Um and it, you know, it it used to prompt you, do you want to do like a manual campaign? Which you know they put sort of in brackets, you know, this might suck essentially. Uh, or do you want to use, you know, ASC, which has got all the latest optimization. So they're really pushing you towards ASC. Now they don't even ask you if you're gonna do manual So it'll put you right into ASC, but it has changed. Um, whereas the sort of the version that everybody knows as ASCs, if you've been doing this for a while, is maximum automation. You don't choose audiences. It's just a broad audience. You didn't used to be able to choose age or or any demographics or anything like that. Now they have actually enabled more age and region. But you didn't have custom audiences, so you you wouldn't select like um your last 30-day visitors or your 1% look-alike or any interest. None of that was selected. It was all just meta doing that for you.
Kurt Elster
So how does it know? Like where's it getting the how does it know what audience to target?
Kurt Bullock
Yeah, so it has to do with You know, Andromeda. And so the again, there's two iterations. All it might be helpful to go into sort of the old way that it was working versus the way that it's working now because I think it's pretty interesting and it'll give you a little bit of insight into the way that stuff works on the back end. uh which has helped my team think a little bit differently about what we're doing in the ad account. So uh is that does that sound all right for next step? Okay, all right. So all right, let's just go, I guess, one step higher. So, you know, Meta's collecting data from everybody all the time. That's, you know, anybody on an app. There's a constant stream of data, um, scroll data, view watch time data, click data, et cetera. Okay. So this is all coming in. And in the old version, meta engineers were coming up with what they call we're when you two different words that you may not have heard or that you we're going to use them differently. One is called features. And features, I think of it like features of a rock. You pick up a rock, is it smooth? Is it heavy? Is it cut, you know, what color is it? So it's those types of features. And so meta there you go. Attributes is a better way to probably explain it. But um yeah, meta's calling it features in their documentation. And it's the old features were made by data scientists. So, you know, Meta gives examples. I'll look at their document here. They say, uh, you know, ads clicked in the last 30 days, uh, brands shown in the last two weeks, uh, Facebook page likes in the last 30 days. So the point being Is that Facebook, you know, data scientists were manually coming up with these groupings, the these buckets, and they were time-based, and it was almost like um I don't know. I think of it like a sticker, you know, like you put a sticker on a pic, you know, I got a picture of me here, and I put a sticker on there that says, you know, likes mountain bikes, visited mountain bike website in the last 30 days. That's you know, we'll contrast that with the new version, which is, you know, if the old version meta was sort of pre-processing the data, putting it in these these data buckets for you. The new version, data is just opening the floodgate, meta's opening the floodgates, and they're giving all the the data to the algorithm. So um Then now the system is determining or deciding what it thinks is useful. So before, you know, we would give them, you know, uh visited this website in the last 30 days, that's what was given to the algorithm to work with. Now what's given to the algorithm to work with is all the click data, all the scroll data, all the view data. And so that's one reason why they've had to put a massive investment into these data centers is because they're processing so much more data than they were before. Um instead of being sort of like uh, you know, pre-processing food for baby, we're giving him like big boy food now, you know? So that's one one thing that has changed with the old versus the new model. The other one, well, let me say this. The new version now, I guess when they upload all of that click data is now they're including time, uh, they're including sequence and timing data. So before it was buckets, right? 30-day, 10-day, five-day buckets. Now they just include that timestamp on there. And so now Meta, just like an LLM is able to sort of predict the next most logical word or the next best word to pop up. uh now the meta algorithm is using your ads and sort of finding what's the next best um ad in this person's story to show up. You know, it's all based on sequence and and coming up with probability. So So now that the new version is, and this what's interesting is I thought this is how Meta always worked, honestly. I thought they always had all this data, but it turns out they don't. And so Um so now it's using sequence data and I like to think of the difference between an image and a video. So let's say that Meta's learning. You know, the first version they see the picture of me with the stickers on there, what I like, what I don't like, and what I visited. And we give that to you know, to the algorithm or to the salesperson and say, you know, do of this what you will, try and figure out if this person's get what they're gonna be a good candidate for and when. You Now with the new version, it's like we're giving them a video of somebody walking through, you know, the store. And so you can say whereas before, maybe you saw like, hey, this person per you know you see a picture of like uh uh they they bought a hoodie and a water bottle and a pair of shoes and that's what meta had to learn from. Now meta is able to learn from oh I saw them walk into the store, they picked up a pair of shoes, they compared it to this other brand here. They open their phone and they looked at reviews uh before, you know, grabbing a water bottle at checkout. And so they get to learn all these behaviors. And um and so the model gets fine-tuned and there's a level of nuance to the customer journey that just wasn't there before.
Kurt Elster
Ever had a customer cancel an order because they were shopping, well high, or because they broke up with their significant other. Those are real reasons customers could cancel, and they're a real headache for both you and your customers. Why should they have to email support to change, cancel, or update an order? Enter Cleverific. With Cleverific, your customers can edit their orders themselves through a self-service portal. That means fewer support tickets, faster shipping, and fewer returns. Everybody wins. Peter Manning, New York cut their order support by 99% with Clever Ific. Now it's your turn. Listeners of the unofficial Shopify Podcast get 50% off the pro plan. Just $49 a month. Go to Cleverific. com/slash unofficial and use promo code KERT50 at checkout. It's less hassle, happier customers with Cleverific. And so if I have you know if I'm a merchant and I have the Facebook pixel on my site. which it it's probably installed with the the meta sales channel. And so it's got this full data layer installed where it can send events back to to Meta and include uh you know and a unique ID for that customer. And so based on that, like even it if it wasn't because I clicked on an ad, just the fact that I visited a site that does Facebook ad advertising, they're going to be able to build this psychographic profile about my interests, aren't they?
Kurt Bullock
That's right. Yep. So as a vet as as a store owner. You you nailed it, just having the Facebook channel enabled, that's you know, that's gonna pass all the store data back to it. And it's gonna combine it with You know, so here's here's another interesting thing about the new model. Um some of it's new, some of it's not. But this was really interesting to me is the way that things are stored. uh you know in meta is um it stores everything as a vector and that was really weird for me to think about at first but let's say that you can what do we mean by vector Yeah, yeah. I mean, so all right, so I imagine it's sort of like a a galaxy, right? So you imagine a galaxy, the Andromeda galaxy, right? Okay. And um everything that is happening Is encoded by Facebook and it's along these different dimensions. So in our imaginary model, right, it's as you picture it, it's a three-dimensional model. In Meta's model, it's actually thousands of dimensions, and but we don't really need to get confused by that. I guess the important thing to think about is let's say that I upload an ad. Meta doesn't just sort of see a block of text and an image and sort of save those for retrieval to be shown to somebody later. It actually tries to understand the text using language processing. Um and it uses sort of visual processing as well, computer visual processing, and it so it'll understand the content. And then it puts that in a location in this galaxy. So let's say all, you know, going back to my mountain biking stuff. All mountain biking stuff lives over here in this corner of the, you know, of the galaxy. So my ad lives there, the content lives there, and it's given sort of a weight. I think of it almost like a little sprite, you know, like it has a color, maybe there's music playing around it. I your version of the algorithm is much more fun. Yeah, okay. Awesome. So we'll we'll go with that. But um so alright, so try not to get sidetracked with my version. What it does is it will it'll give it a location. The interesting thing about this is that then your user behavior is also given a location. So let's say that you know I'm on and I'm scrolling through reels of mountain bike uh content. It's going to encode that. And they call it the word that they, you know, this is our second vocabulary word is embedding. Embedding is basically when Meta takes this user data or ad data. and puts it in sort of a coordinate, you know, XYZ location in this space. And the reason it does that is because it's then going to measure the distance in between various objects and use that to determine relevance.
Kurt Elster
Oh wow.
Kurt Bullock
Yeah, so like if your scrolling behavior lives in the same neighborhood as uh my ad and the attributes that Meta understands, you know, my ad to have. uh it'll measure the distance between those and if they're close together, it sees that as a positive signal and that, you know, you're we might be a good match. And and that's one That was that was sort of an unlock for me, just thinking about how this is all stored. Um, because now all of a sudden you can think about different neighborhoods of creative and different neighborhoods of audiences. and trying to make your way into those different neighborhoods uh you know with our main lever which is ad creative.
Kurt Elster
Okay. So what we're really this truly is uh A leap forward in machine learning is what they've done here. Where because this is how the LLMs think in in neural networks. Like they have built a a deep learning network around Buying behavior and shopping behavior. That's quite clever. Um, hmm. So we're able to get, you know, ultimately the goal is I want to show more relevant ads. to the right people. You know, if I'm a customer, if I'm just on Instagram scrolling, being interrupted by ads that are irrelevant to me is more annoying than being interrupted by an ad that is relevant to me. So it's like net Net good for everyone involved is what we're going for here. Um, at, you know, some huge expense to build the data center, you know, and process that data. But that's what we're going for, right?
Kurt Bullock
Yep. Yep.
Kurt Elster
Okay. And so all these set like that, there's no settings for me. You know, all that stuff that the those are just algorithm updates behind the scenes. And then I'm not choosing There are fewer settings and and toggles and buttons to press in you know in my campaign with this new system. And then as far as like the data layer integrating it with my store. That's plug and play. That's uh the the meta sales channel and Shopify. So we've ruled everything else out is being done for us. It sounds like creative's the only thing that left. What where do I is it the person, what do I get to do?
Kurt Bullock
Yeah, I mean, so this is something that is unfolding a little bit. So as I mentioned, there's two different versions of ASCs right now. The old version, it was just broad targeting. You didn't have any targeting. The new version that just rolled out into accounts is like a combination of the old and the new. And where is it actually going to land? I don't know. We'll see. They maybe they're trying to figure that out as well. But so now they've given us access to um ad sets again. So old ASC, there was no ad set. It was all just one ad set and one campaign. You would throw everything in the same bucket, let Meta sort it out. With this latest version, they've given us a little bit of that control back so we can create different ad sets that have targeting interest and look-alike targeting. Um and we can have exclusions. Um for those uh you know of of you listening that use the old ASC, there was a customer um Let me see. This you could determine with a slider essentially how much of your spend was going to go towards existing customers. So it was the kiss existing customer spend cap. That's I would usually set that to five or ten percent um because it was all in the same campaign. With this new version They took away the existing customer spend cap, um, but they broke it out into separate ad sets for you. So you can exclude customers again completely. Like you used to be able to. So where's it gonna land? I'm not sure, but I think the the message is exactly what you've described, and that is Meta doesn't really want us fiddling with all the controls in the same way. It's obvious, you know, as sort of we explained with all of the this sort of um the modeling and the data that they're collecting on everyone, they've got infinitely more data points than I have access to. And so really the way that I think about interceding is if I need to teach, you know, meta something, because if we rewind just a little bit back to the when we were talking about embedding and vectors. You know, when you put a new ad in, there's a lot of talk on Twitter about like, you know, you put a new ad in, it gets no spend. How can Meta know if this was a good ad or not based on, you know, 50 cents or a dollar worth of spend? And I think the answer is that it can't know entirely. But what it does know is this. It will embed your ads, so it'll convert it into that vector, place it in a location, and then it will look at other ads nearby. And see how those performed. And it will sort of infer the performance of your ad based on what other stuff has done, since you know they're all in that close sort of physical proximity to each other. It'll make assumptions. It's not always right though. So I've thrown things into ASCs where uh ASC you know didn't spend on on this ad. It just spent on its favorite two or three. And um I was trying to get spend of this new ad that we had, you know, just produced. Um, and so what what I do now is I'll break them out. into a separate campaign by itself, a concept by itself, with, you know, let's say three variations. So if I've got one one video and I'll open with three different hooks, let's say. So it's the same concept. And I'll force spend through it. And what that does for Meta is that it adds color. So it's got that location, uh, but that's not fixed. It actually changes the more Data that meta gets. So if you force some spin through it, it'll get a more accurate picture of where that should land. And sometimes it can unlock performance. So for instance, one of my best performing ads in in this account that I, you know, that I'm thinking of right now, uh, cleaning supplies. We could not get it to spend. We broke it out into its own separate campaign, forced spend through it, and then we reintroduced it back into the ASC campaign. And uh it has totally changed the ad account. It brought us from not profitable to profitable. Why wouldn't Meta spend on that ad before? I don't know. You know, but it helps sometimes to force some spend through it, give Meta a little bit more complete picture, uh, some test data to work with and instead of just, you know, going based off of uh you know, all the information that it's sort of um I guess all of its inferences about how it s suspects the ad will perform.
Kurt Elster
So an ad it has not run any spend two. It can review it and then analyze it and then decide it's not gonna bother.
Kurt Bullock
Right. Basically, you know, so like if you've ever uploaded an image to Chat GPT and ad, you know, in my case I do it all the time with ads and say like, what's this ad about? Um, who is this targeting? Um, what's the offer? They're doing the same thing um with the ads. And so they're storing all this data about, you know, who is this about? What's it, you know, who is it targeting? What's the product in here? What's the offer? And so it's not actually storing, let's say, the add image in that, you know, embedding space. It's just storing what it needs to know about the image and and the meaning of the image. And so that was one of the big, you know. takeaways I guess as I learned more about this is that changing one word uh in your headline is not going to cause um the ad to be embedded in a different location. It's going to show up, you know, in the same spot. It's not meaningfully different. The only way that words uh you know in your headline are going to change where it's embedded is if it has a different meaning, if it actually changes the meaning of that phrase or sentence. Um so that you know Meta has a different understanding because it's not cat it's not storing every little word, it's storing the meaning of the words. So so what you need is for the meaning of the words to change, and that's how you get to a different location.
Kurt Elster
And you attempt to predict this by just having a different AI review at first. Yeah, you could you Llama, you know, Meta's, you could get use that on uh for free on Hugging Chat. Which kind of a fun tool uh as an alternative chat GPT. You could try having it review the ads and see what it comes back with.
Kurt Bullock
Yeah, so uh that's a I actually haven't tried. Did you say hugging chat?
Kurt Elster
Hugging chat is yeah, it's like a uh platform where you could pick different LLMs and they have Llama, which is Meta's. It's kind of fun.
Kurt Bullock
That's that's a good idea. The way that I've done it is I've I put together a custom GPT, um, you know, in chat GPT, and and basically I included documents that explain all of this embedding stuff uh and links to Meta's engineering documents. Um, and so I've put that in there and the the the GPT is there just to give me different ideas about my ad. So I'll I'll upload an ad image or I'll upload copy and then it'll spit out. You know, have you thought about trying this, that, and the other so that you can make this meaningfully uh different? And You know, I've actually got a list of some of the questions that I can share with everybody. We can go through some of those too if that's useful.
Kurt Elster
But wait, what are the questions?
Kurt Bullock
So yeah, I've I've got a big list and and so I you know it's probably not gonna be good content to go through all of the questions, but I'll be a couple of your favorites. Yeah, yeah. I'll I'll I'll give you a few. So What I'll start out with, and a lot of this just really goes back to fundamentals of marketing. Okay. Um, so one of the ways that we've unlocked audiences is just Um, you know, we're thinking about what's a different, a completely different reason that somebody might buy this. So I've been looking through, you know, I'm always in the ad library, and there's an electrolyte brand that I follow that I think does a really good job with this. And I noticed in the, you know, in the ads library, they've got a ton for, you know, sort of the benefit of helps you recover, which you would expect, right? Electrolytes. Okay, so we're targeting athletes, et cetera. But as you scroll through their ad library, you also see they have um, you know, improves focus, which is like maybe for keyboard warriors, right? Like guys like me that are, you know, working in the office and typing. Um so You know, helps with brain fog, helps with focus. So that's another um different reason maybe that somebody might buy the same product. Um and then they also, interestingly, they had one for like kicking the sugar habit. Uh, so is as an alternative to like, you know, Gatorade and sugary drinks. So that maybe targets a different segment. And it has a, you know, you're you're gonna land in a different neighborhood because you're talking to a different person.
Kurt Elster
Right. Very different buying reasons. You're right.
Kurt Bullock
Yeah, right. And then uh interestingly, I also saw they're targeting like nursing mothers who want to stay hydrated, uh people. uh nursing hangovers, uh people that have been out drinking and partying, you know. Um and so I think they've taken that idea and Uh that's a they've executed really well and I'd love to see how the ad account looks, you know, the results. But from the outside, it looks like they're doing a great job at finding these different buying reasons. So that's one of the first things I'll start with. is just thinking about um you know different reasons that somebody might buy the product the the uh product. Um there's visual style and I think that this is you know not going to be new information to anybody so this is stuff like us versus them and UGC and founder ads. Um right. So making sure that when I look in an ads library, I do sort of what I call the squint test. It's like if I look at your ads library and everything looks generally the same, it's all using, you know, brand colors and it's all like super, let's say, product focused or not product focused. Whatever it is, if it all looks basically the same, that's not a diverse ad account. That's not what Meta is asking for. And you know, you're going to be reaching largely the same audience. Which if it's working, that's great. And that's part of the strategy. Uh, you know, is when you find something that works, you want to you know they call it explore and expand. So first you're exploring, finding these audiences. Once you find them though, you want to expand and build out. Once you find messaging that works, great. Try that messaging as an us versus them and try it with you know, three different creators and you know, sort of expand on that. But um going through you know some of the other questions though is just an interesting one was environment and location. So am I showing the product in the same setting every time? So Um for instance that cleaning product, you we were always using it in the bathroom and the kitchen. That's sort of what our ad library looked like.
Kurt Elster
Then that would be like the obvious choice, right? Cleaning a window, cleaning a counter, cleaning a bathroom.
Kurt Bullock
So we started trying things like um we took it out to the park and started cleaning like signs, you know, traffic signs and like playground equipment and things like that.
Kurt Elster
Something unexpected but familiar.
Kurt Bullock
Exactly. Yep. So another it's just getting us out of that same space that we were in. You know, and it's so easy when you're coming up with your ads to just sort of have your, you know, we're a cleaning product. We're gonna be this is gonna be filmed in the kitchen. Um and so these other these other videos that we've created have uh done really well organically. Um so they've sort of taken on a life of their own and uh worked really well in the ad account as well. And I think that That's sort of a um an expand. We're not necessarily going to be reaching there's people exist along the spectrum in lots of places in these neighborhoods, right? But I think we're reaching a different corner of the neighborhood by doing that.
Kurt Elster
Are your upsells stuck in the Stone Age? Your store runs on data, so why aren't your upsells? Outdated upsell apps make you guess, test, and tweak manually. That's costing you money. Zippify one-click upsell changes everything. Its AI engine automates upsells, analyzing every cart in real time to offer the highest converting product. No guesswork, no setup, just results. Over 10,000 Shopify stores trust OCU, and it's already generated nearly a billion dollars in extra revenue. So flip the switch and let AI handle your upsells. Try Zippify one click upsell free for thirty days at zipify. com slash curt. That's zipify dot com slash K U R T. Don't leave money on the table. Start AI powered upselling today. Yeah, this is really this kind of changes the you're right. This very much changes the approach. It is important to understand the the methodology here to you know how these these deep learning networks, these neural networks think. Have you, you mentioned custom GPTs. Obviously the algorithm itself is very AI-based. Have you used any uh AI-generated ads?
Kurt Bullock
I have and it's a it's a little bit of a a struggle with um brand guidelines uh in terms of oh so I guess what I'm trying to say is this. I've I've used it and I think it comes up with great images, but the labels are always messed up.
Kurt Elster
I know.
Kurt Bullock
It changes the logo. Yeah. It changes the colors on the package. Um, and you zoom in and sometimes some of the words are just gibberish. And you know, when you if you just see it as an ad, you may not zoom in and see that the words are gibberish, right? Consumer probably will never know that. But when I send it to my clients, they're like, what is this?
Kurt Elster
And then when you have to like, you know, at the point Well, my we've done this, you know, to get positioning um content. And then my designer will be like, well, you know, we could Photoshop the labels on. Like at that point, we may as well just take the photo.
Kurt Bullock
Yep. Uh but you know, I have used AI-generated ads and sometimes it does a great job. I think it has to do with sometimes the complexity of the label, or if it's, you know, like this cleaning product that I keep referring to, it has a hard time reproducing the label. But things like backpacks and stuff, it's done better with, where it's not trying to reproduce the label. It's just it knows the features, you know, of the zippers and pockets and the color and the fabric and it can get that right. So Um, so yeah, that's something I'm definitely playing with. And a lot of times I'll do it in multiple phases. I'll I'll maybe have the environment generated. So this is one thing we did. We'll we'll say, hey. picture somebody, you know, on the rug, you know, cleaning with a rag, and then we'll describe the scenario. You know, there's sun coming in from the window. There's like you know, a glass of water on the wooden table and it's a rug on a tile floor. Just sort of paint the whole picture. Uh and it'll give me that environment. And then we can go in with our designers afterwards and place the text overlays and sometimes just lay the product on there or adjust the label.
Kurt Elster
I think that's probably the smart way to do it. Plus it's and it's so good for ideation. Like you really, it wouldn't be realistic to be like, hey, we're gonna take this photo ten different ways in 10 different locations with 10 different people. You know, it's you could do it, but it'd be so time consuming and not practical. Versus if, you know, I could just come up with like, hey, here's 10 ideas, and now, you know, we're gonna punch one up. And when it does well, okay, now let's do that photo shoot for real. I think that's probably the approach.
Kurt Bullock
Yep, it works really well like for creative briefs. So well you We'll do just that. We'll have AI come up with a few different images, send that to one of our creators just to kind of give them the idea of what we're thinking about, and then they create it, uh, you know, using their own resources. So it can really be helpful in that respect.
Kurt Elster
Meta's doing all this automation, but it really is putting the focus back on the ad creative, which I can I like that. Makes things you know it's things are still very technical, but Maybe less so as far as like, well, I you know, if if I just get the right magic combination of settings, my campaigns will do better. Right. Um, but then if everything's in this algorithmic black box, have I lost My ability to test in tune, you know, to learn from findings, or is it is it easier? What's your take?
Kurt Bullock
I think that um For better or worse, we've eliminated some of the variables that we have control over. Right. Um, and so if I look at it for, you know, for for better, I think Sometimes we would get in trouble trying to to create and tweak our way into performance, uh, you know, with inside the meta-ads manager. And sometimes we were doing more harm than good. It's hard to know. Um and so, you know, if we're looking at this as a positive, I would say, well, getting rid of that variable is great. And we can focus really on what Meta's asking us to do, which is come up with good creative. and good messaging, think through your buyer personas and why they want your product. What desires did they have that you can meet, right? And how you can express that as an ad. Um, so you know, we still have a testing methodology. It's not all just throwing stuff against the wall, you know, just to see what lands. Um as I mentioned a little bit before, we'll launch campaigns into what they call ABOs, which is basically just an ad set where you know you have a set budget for it. And that's where I'll usually launch new ads. Um, so we're testing that way. And if something works, we do what they call scale in place, right? So you can just Ramp up the budget. If something works, great. Don't don't change it. You can increase the budget right there. Um it doesn't even need to be called a testing campaign, right? It's just whenever you launch something new, it's always a test. And so if it works and you have it broken out into ABOs, you have more control and you can push more spend through it. But then if something does work, I always move it into my main ASC, my scaling campaign. And so my my ASC is full of just winners, right? It's stuff that's worked in those ABOs. I can feel confident putting it in there with my best ad and It won't necessarily get spent, but here's the thing about ASCs is that they, since they're full funnel campaigns, you know, what is a funnel? It's a sequence, right? And so that's what this is all about. Andromeda is all about the introduction of sequence into the algorithm. And so That's powerful. Yeah, so I think that we can we can feel okay putting more creative into that ASC. And if it doesn't all get the spend that we wanted it to. That's okay too, because Meta now has more nuance into the customer journey and it's going to be inserting these ads into places, different spots in the customer journey than it was before. It's finding new places for that. And so This ad may not be the best top of funnel, you know, high spender, but it may work great mid-funnel when somebody's comparison shopping. And that's when Meta might call on it, right? To come out and do that job for you.
Kurt Elster
All right, now I after you know talking about this for a while, w are meta-ads still the go-to? You know, if I'm I've got a product, I've got an online store, I want to sell it, I need to drive awareness atop a funnel traffic. Is Meta still the place it should be starting in 2025?
Kurt Bullock
I think that it is. Yeah. When I look across, we've we've tried, you know, all the platforms and Nothing has the scale or the data that Meta has. They've been, you know, they've been collecting data for so long and they've got such an advantage in that department. You know, every e-commerce store has the one of the a MetaPixel on there. So they're collecting data from everybody. Um, and so I really haven't found another platform where I can go in and um and scale. So that's usually the first place I go. And then as I find scale in Meta, then I start adding it to other platforms. you know, I'll do Meta and Google, right, at the same time. Meta's kind of throwing the ball and Google's catching the ball. Um But uh in terms of like TikTok and Pinterest and th and those other platforms, they're great sort of ancillary platforms um in general, but I think if you look in 95% of ad accounts or or uh brands, it's gonna be heavily meta weighted.
Kurt Elster
For sure that's that's what we see.
Kurt Bullock
Yeah.
Kurt Elster
It's gonna be like your qualified traffic is gonna be, you know, Google organic traffic, meta ads, uh Google paid ads and email marketing. And you know, it's no one thing. It's all of those working together. But if you take one, you know, they're all going to take credit, of course. But you know, you you take one out and suddenly the whole system does not perform as well.
Kurt Bullock
Yeah.
Kurt Elster
So if you had one key takeaway, you know, you wanted, you know, people are running Facebook ads and this episode's coming to a close, what do you want them to go do?
Kurt Bullock
I think that it's about coming up with, you know, trying to find ways to get into different neighborhoods, right? Um, with our creative. And so it's It's just you can go and do a br you know, I'm doing brainstorming sessions with my team in a way that I wasn't before where we're sit we're just, you know, I've got you probably can't see in this image, but I've got sticky notes all over my wall here from uh brainstorming sessions that we're doing. And so
Kurt Elster
Instead of you just trying to figure out different positioning.
Kurt Bullock
Exactly. I'm just thinking more about positioning and more about diversity with my with my creative because I know now That it actually matters, you know, where Meta places me. I'm gonna tap into new audiences that way. And so that's one way to scale it, right? Um, you find an audience that's working, great, sort of multiply and conquer that neighborhood as much as you can, but There's gonna become a limit, right? There's a limit to that, to how many people you can reach in that neighborhood. And if you keep on putting, you know, together ads that look the same, Meta actually send it sees them as duplicate ads, basically. Um and so you may not be making the progress you think you are. Like great, we came up with 15, 20 new ads this week. They all look really similar. Well Meta's going to place them all in the same spot and it doesn't want to serve duplicate ads. You might actually get less, you know, your ads are going to be served less if they all look the same. Um so I would just really stress creative diversity. Um, all uh, you know, include a resource that has a bunch of questions that you can go through that we're using in our uh in our brainstorming sessions. And I would just try and get creative about that.
Kurt Elster
I got you know you talked a lot about hey the creative diversity and you know thinking through it and then like some of these questions and having you know custom GPTs, chat GPT review it. I what's the resource here? Give me something. You always have great resources for us.
Kurt Bullock
Yeah, so I'll I'll put together I've got a um a PDF that has the questions that we got uh you know that we have put into our GPT. And so you could put those into your GPT as well and upload it or just run through it yourselves as just sort of a brainstorming session. So that I'll include that. It's it gonna be at produce D-E-P-T dot co. uh slash usp for unofficial shopify podcast. So produce d ep t dot co slash USP and um it'll have two things in there. It'll have those questions that we go through and then I'll also include just a map of our testing our campaign testing structure.
Kurt Elster
Beautiful. I will also be grabbing that list myself and then turning it into a custom GPT. I would not mind playing with that. Kurt Pollock, thank you so much. If people wanted to learn more about you, where can they connect with you? Where should they go?
Kurt Bullock
You can find me on Twitter at Kurt Bullock, K-U-R-T-B-U-L-L-O-C-K, or just at the at our uh agency website, producede-ep-p-t. co.
Kurt Elster
Kurt Bullock, Produce Department. Thank you so much.
Kurt Bullock
Thanks a lot, Kurt.
Kurt Elster
Crowdfunding campaigns are great. You can add social proof and urgency to your product pre-orders while reducing risk of failure But with traditional crowdfunding platforms, you're paying high fees and giving away control all while your campaign is lost in a sea of similar offers. It can be frustrating. That's why we built Crowdfunder. The Shopify app that turns your Shopify product pages into your own independent crowdfunding campaigns. We originally created Crowdfunder for our private clients. and it was so successful we turned it into an app that anyone can use. Today, merchants using Crowdfunder have raised millions collectively. With Crowdfund You'll enjoy real-time tracking, full campaign control, and direct customer engagement. And it's part of the Built for Shopify program, so you know it's easy to use. So say goodbye to high fees, and hello to successful store-based crowdfunding. Start your free trial and transform your Shopify store into a pre-order powerhouse today. Search Crowdfunder in the Shopify App Store to get started.