Shownotes
In this Retail Technology Spotlight episode, Armen Mkrtchyan, CEO and co-founder of Extuitive, joins Omni Talk to tackle one of the oldest problems in advertising and to show how AI is finally cracking it.
John Wanamaker famously said, "Half of my ads don't work. I just don't know which half." Nearly 200 years later, that problem hasn't gone away. It's actually gotten worse. Armen breaks down how Extuitive's AD Intelligence Engine uses a fusion of real consumer panel data (150,000 people, complete with purchase receipts) and brand-specific platform history to predict which ads will perform before you spend a single dollar launching them.
If you're a retailer or brand marketer trying to make your ad spend work harder in a world drowning in AI-generated content, this episode is for you.
🔑 Topics covered:
Why AI-generated personas push everyone toward the mean, and why that's a problem
How Extuitive's fusion model combines real consumer data with brand-specific ad history
The pre-validation approach: submit 200 ads, launch only the 65 that will actually work
How to optimize for awareness vs. conversion depending on your campaign goals
What you need (6 months of ad history, ~250 ads) to get started
🎧 Don't forget to like, comment, and subscribe for more retail tech insights!
#retailtech #digitaladvertising #retailmarketing #adspend #predictiveanalytics #omnitalk #retailai #contentmarketing #ROAS #retailpodcast #brandmarketing #Extuitive #metatiktokads #retailtechnology
Music by hooksounds.com
*Sponsored Content*
This podcast uses the following third-party services for analysis:
Podcorn - https://podcorn.com/privacy
Transcripts
This Retail Technology Spotlight series podcast is brought to you by the Omnitalk retail Podcast network.
Speaker B:Half of my ads don't work.
Speaker B:I just don't know which half.
Speaker B:Many of the platforms today push advertisers to just create a whole bunch of content.
Speaker B:It only helps probably one stakeholder, the platform itself.
Speaker B:We say, listen, put your 200 ads through the platform.
Speaker B:These are the 65 that you should launch.
Speaker B:And out of 65, probably 50 will do very well.
Speaker A:Hello, everyone.
Speaker A:I am Chris Walton, your host for today's interview.
Speaker A:An interview in which we will explore the question, how should retailers and brands plan for a world in which we will all be fighting for eyeballs amid a plethora of AI driven content?
Speaker A:That question is particularly interesting to me because as you all likely know, I too, make my living producing authentic content.
Speaker A:And this question really honestly keeps me up at night.
Speaker A:So I brought in an expert who has some really interesting ideas on how to stand out in this coming future.
Speaker A:Armin McCurchin, the CEO and co founder of Xtuitive.
Speaker A:Thank you for joining me today, Armin.
Speaker A:How you doing?
Speaker B:Chris?
Speaker B:Awesome being with you.
Speaker B:Thanks for pronouncing my last name as I.
Speaker B:As I told you earlier, sometimes I'm not able to do it either.
Speaker B:It's awesome being with you.
Speaker B:Look forward to our conversation.
Speaker A:Yeah, it's great.
Speaker A:It's great to have you too.
Speaker A:Yeah, no, I was joking with you before.
Speaker A:Like, you.
Speaker A:You definitely win the Continent award with one vowel in that entire last name.
Speaker A:That's a new one.
Speaker A:It's always the toughest part of the job, too.
Speaker A:So thank you for calling that out when I get it right.
Speaker A:Because, you know, it is the toughest part of the job sometimes getting everyone's names correct.
Speaker A:But.
Speaker A:But.
Speaker A:Yeah, so.
Speaker A:So you've got an interesting background.
Speaker A:We were joking around before we got started.
Speaker A:You actually went to school in North Dakota, Armin.
Speaker A:Tell us about that.
Speaker B:I did.
Speaker B:I went to University of North Dakota.
Speaker B:Started for three years engineering, electrical engineering, mostly.
Speaker B:I loved it.
Speaker B:I. I ended up in North Dakota directly from Armenia, where I grew up was a bit of a culture shock, but I met some of the nicest people in my life, I bet.
Speaker B:And.
Speaker A:And did you take to flying like everyone else that goes to school in North Dakota?
Speaker B:I. I didn't study flying, but I had so many friends who are flying that I actually went with them and started flying.
Speaker B:They allowed me.
Speaker B:And then eventually I actually started working towards my private pilot license as well.
Speaker A:Wow, that's awesome.
Speaker B:Yeah.
Speaker A:For those that are listening to the Podcast like for those that are maybe not from the Midwest, North Dakota is a known pilot school.
Speaker A:So I was joking with Arvin beforehand, I was like, oh, you're not an actual airline pilot.
Speaker A:So that's the, that's the origin of that conversation.
Speaker A:But, but.
Speaker A:All right, Armand, so your company, you know, it's built on the idea, it's really built on the idea that there is a right and a wrong way to answer the question that I, that I posed at the outset of this podcast.
Speaker A:And so set the landscape for us 30,000 foot view.
Speaker A:What are the issues facing retailers and brands when it comes to promoting the right content?
Speaker A:What are the issues that they're having to tackle right now?
Speaker B:In a way, the issue is probably millennia old, at least centuries old.
Speaker B:In fact, if we go back to about 200 years, a guy named John Wanamaker who literally pioneered the concept of the department store in the U.S. i think he had this saying where he said half of my ads don't work, I just don't know which half.
Speaker B: And this was in like: Speaker B:So in a way that problem has not disappeared.
Speaker B:In fact, if anything, I think it has been exacerbated, especially in the digital world by the amount of content that people see every day and trying to actually get attention of your user, of your potentially subscriber, of your buyer.
Speaker B:So what extuitive is arguing is that there is potentially a way of figuring out what's going to work before you push content, especially in our case digital ads out there in the wild and letting people see it.
Speaker B:So that's almost like a 40, 50,000ft view.
Speaker B:Now if I come a bit lower, what happens daily is all of us make if we're launching content and we do also launch content for extuitive as well.
Speaker B:We every day make gut based decisions.
Speaker B:You and I will look at a content and say, you know what, this feels right or this doesn't feel right or I kind of like this.
Speaker B:Maybe orangish color is the background.
Speaker B:Let me launch that a bit more.
Speaker B:Some of it we kind of have a gut feeling from what has worked in the past.
Speaker B:So we'll say let's go launch this and let's not go launch this for example.
Speaker B:But that's kind of what we do.
Speaker B:Like if you think about it or what we will do a lot of the time is we'll, we'll launch and we'll say listen, let, let the kind of large platforms optimize it.
Speaker B: whole bunch of content, maybe: Speaker B:In fact, some probably per week launch several hundred.
Speaker B:And they'll say we're going to let our, our platform partner can be meta, for example, optimize which ones are going to work and they are basically going to kind of increase the spend on those and they are not going to increase the spend on the ones that don't work.
Speaker B:The challenge with that is you are basically spending most importantly time in figuring out what's going to work before it works.
Speaker B:Second, you are spending money in testing things that actually aren't going to work.
Speaker B:But you are at least, you have to give them at least an equal shot in the beginning.
Speaker B:So what extuitive is saying, can we make that process whole bunch easier?
Speaker B:So when you are launching something, we are almost pre guaranteeing that that content is going to work and it's going to be in your top quartile, top 15 percentile of, of all the content that you've ever launched before you even launch anything.
Speaker A:First of all, I love, I love the retail historian in you.
Speaker A:I didn't know, I didn't know you had that in your background too.
Speaker A:First North Dakota, now retail history.
Speaker A:So that's great.
Speaker A:It's great reference point.
Speaker A:You know, I love that whenever we get that on the show.
Speaker A:But, but yeah, I mean what you're saying, what you're saying makes sense.
Speaker A:I mean I even, I think about it for myself, like I was looking at it yesterday.
Speaker A:I think I produce 450 pieces of content in the last month for LinkedIn, you know, and it's really just my gut feel on what I think is going to work.
Speaker A:You know, me and a couple other, you know, a few other people on my team trying to decide what that is.
Speaker A:But the, the question I have too is like no, put it and put it in like the, you know, the 20 so now.
Speaker A:But you know, you went back 200 years.
Speaker A: We're in: Speaker A:Which does that also mean there's going to be just a premium placed on getting it right?
Speaker A:Because it's just going to be.
Speaker A:So it's going to become prohibitively expensive to do this, you know, in a way that, you know, is if you don't know, if you don't know what you're doing, it's going to be difficult for you to succeed.
Speaker B:I mean we kind of are seeing it already.
Speaker B:Chris, what's happening is many of the platforms today are not naming any, but they would push you, they would push advertisers to just create whole bunch of content.
Speaker B:They would say create diversity, just launch more and more and more and more and more.
Speaker B:And if you think about it, it only helps probably mostly one stakeholder, one player, which is the platform itself, because you have to pay for all of the content that you are launching for the user who is consuming it.
Speaker B:They just basically get bombarded by all kinds of messages and all kinds of content.
Speaker B:So they got to make a decision on what is more appealing.
Speaker B:As we are scrolling through their feedback and for the advertiser, for the merchant, for the retailer, they are trying to figure out, it's like, how much should I even create to start with?
Speaker B:How what is diversity mean?
Speaker B:Literally, it's like, does it mean like I just go and create for the same product, 50 different ads?
Speaker B:Do I go create five different ads?
Speaker B:Do I change the Personas that I'm using in my ads?
Speaker B:So that has, in fact, if you look at that, and we have done a whole bunch of experiments ourselves, and if you look at the CPA for, for many of the ads over time, they keep creeping up and that is not generally sustainable.
Speaker B:So there is a, we got to figure out a way of understanding how do you at least bring some predictive power of knowing what type of diversity you should create and what type of ads you should even launch before you for per ad spend the type of CPA amounts that you are generally going to be spending.
Speaker A:Right.
Speaker A:Yeah, that's really interesting, especially when you bring in the advertising element into it too.
Speaker A:Like that content I'm putting out is not typically, you know, I don't typically advertise it.
Speaker A:And so, yeah, if you're advertising that content too.
Speaker A:Yeah.
Speaker A:Then there are potentially some, I never thought about that before.
Speaker A:Some potentially misaligned incentives with the platforms on which you're advertising who want to get you to put out as much content for that reason, because they know you're going to advertise it or put.
Speaker B:Well, I'll give you another, I'll give you another data point, Chris.
Speaker B:We have, again, this is kind of our internal analysis, but we have looked at, for some of the larger platforms, at the type of ads that get promoted to basically be the top ad for money to be put behind.
Speaker B:And what we have found very consistently is that the very top one or two ads aren't the ones that get promoted by the large platforms.
Speaker B:It's generally the ones that are from, again, from our analysis, it's generally are the ones that are in Your maybe second quartile that get promoted.
Speaker B:And that kind of makes sense if you look at the incentive of the platform.
Speaker B:The incentive of a platform is to basically figure out what is going to work, potentially work well, not necessarily extra, extra well, but work well enough that if you launch an ad, they'll generate the returns, but they'll also generate the inflow of, of cash for themselves as well.
Speaker A:That's kind of a mic drop right in the beginning here.
Speaker A:Armin.
Speaker A:So, okay, so basically, so if I recap, you know, everything we just talked about, we said like, you know, there's a, there's a, there's a cost of producing ads, like there just is.
Speaker A:And so you want to be thoughtful about how many that you're or they're producing content, I should say, whether you advertise it or not.
Speaker A:But there's a cost of producing content that you want to get right and you know, everyone wants to control their costs.
Speaker A:And, and then there's also some information asymmetry that's going on through this process in terms of how that content goes out into the world.
Speaker A:So, so you're, and your theory is that you can be more predictive about what content is going to work well.
Speaker A:So, you know, so I get that theory is one thing, I've heard this theory before.
Speaker A:But how do you, if I put your feet to the fire, how do you actually increase the probability of being right?
Speaker B:It's a great question.
Speaker B:We built what we call at extuitive an ad intelligence engine, which is our predictive engine in figuring out what's going to work, what's not going to work.
Speaker B:But probably even as importantly, it also helps you figure out who is it going to work for and who is it not going to work for.
Speaker B:It's both the probability of working, but also the audience who it may work for and may not work for the way we do it.
Speaker B:Chris, to get to your question, we started about a couple years ago collecting Data from about 150,000 people, mostly in the U.S. demographic data, very detailed, where they leave, obviously no names, so everything is deanonymized but where they live, how old they are, how many kids they have potentially, what kind of car they drive.
Speaker B:And we asked them for about 30 minutes all kinds of questions and showing different ads and seeing their preference on what they like, what they don't like.
Speaker B:At the same time, we collected also receipts from these people in figuring out what they buy daily, what are they gravitating towards?
Speaker B:They can tell you and me that they like a certain yogurt brand.
Speaker B:But if they go and buy something else every day that you can see on their receipts, then it's a data point at the very least that I need to use when we're doing modeling.
Speaker B:So what we did was we took the data from this 150,000 people, we created digital Personas modeled after these people.
Speaker B:Again, we don't know generally who these people are in a very specific way.
Speaker B:We don't have their names, but we can call them Chris and Armin and Joe and Bob Armin living in Boston now, Chris lives in Minnesota.
Speaker B:And we've covered all of the states with the same way.
Speaker B:The U.S. census basically would cover very representative census data.
Speaker B:And we created these agents that Personas that represent US population.
Speaker B:So we can ask them questions, we can show them ads and say how would you react to this?
Speaker B:We can also give them different messages that would go with the ad and say, hey, if the ad had this text versus this other text, how would the Personas react?
Speaker B:And we can cut them any way we want to.
Speaker B:We can say let's test only moms in Minnesota, for example, for a very specific product that we are targeting.
Speaker B:Or we can test bikers in Massachusetts for another product that we are targeting, for example.
Speaker B:So that's one part of the model that we use.
Speaker B:The other part of the model is very brand specific data that we get from large platforms.
Speaker B:When people work with us and connect their account to extuitive, we get very specific brand level information that we analyze in terms of what people had clicked before, how they had clicked, what are all of the metrics from top of the funnel, from all the way from ctr, CVR all the way down to potential CPA and roas.
Speaker B:We take the collection of the Persona based model, the brand specific model, and we create a bespoke fusion model for every single brand that we work with.
Speaker B:So it's a fusion model that combines the two parts and every single time it's bespoke, it's custom created for every brand.
Speaker B:And that's what we will roll out to brand that we are working with.
Speaker A:So to increase the probability of getting, getting your content right more often.
Speaker A:You're saying there's foundationally two building blocks.
Speaker A:I heard of one is, one is you've got to have a database of real people that you can understand their inclinations, their proclivities in terms of how they view advertising, how they view products, how they actually go about making purchases.
Speaker A:And then you also integrate then the other key piece, the second step that I heard is that you also integrate with the brand or the advertisers or the retailers, you know, social media data to understand how their past performance of what they've advertised has actually performed in the marketplace.
Speaker A:And you're fusing those two things together as the foundational building blocks of kind of creating a predictive modeling exercise, so to speak, to help the brands be right more often.
Speaker A:Did I summarize that correctly?
Speaker B:Armin, you recap it much better than I could.
Speaker B:Chris?
Speaker A:Well, no, no.
Speaker A:Well, thank you for that.
Speaker A:No, thank you for that.
Speaker A:But, but you know, the curious, the question I have is okay, because I've seen a lot of, especially in AI.
Speaker A: hy I made the point of saying: Speaker A: you know, let's put this in a: Speaker A:I've seen a lot of pitches on stages, at a lot of conferences where they're, they're trying to use, you know, kind of AI Personas to approximate step one of that or foundation pillar number one of what you described.
Speaker A:What is, what is right or wrong about that approach relative to using real people.
Speaker B:So in a way, if just not to get too technical, but it's what, what you and I can do today, or anyone, any, any of your listeners can go to any large language model provider.
Speaker B:Let's, let's assume ChatGPT or OpenAI and prompt ChatGPT and say, hey, create me a Persona that represents Chris.
Speaker B:And Chris is this person who lives in the zip code, does the following things, for example, and it'll create something for you and then you can say now ask this Persona whatever you are going to like, a specific cup that I'm trying to sell, for example, and it'll give you a ranking.
Speaker B:And we have done that exercise as well.
Speaker B:And what have we found out?
Speaker B:What we have found out is that these models tend to predict basically mostly the mean, the average of what a Persona would do.
Speaker B:So every single time it's a, yeah, it's like you have a middle aged person living here, that's probably, this is what they're gonna like generally, although you are on the younger side, Chris.
Speaker B:And it'll basically give you an answer to the question and every time generally the same answer.
Speaker B:So what you lose is the diversity of views that you want to capture.
Speaker B:And that's what the real data allows us to do.
Speaker B:That's one thing.
Speaker B:The second, because of our real data is also very much infused with almost receipt level information, what we are able to do is not just Figure out what is the person's intent, meaning what are they going to say they are going to do but also what they really have done in the real world.
Speaker B:Which is a much better signal to use to capture whether a person is going to buy something.
Speaker B:Because ultimately most of the retailers, most of your listeners, when they are launching a product, when they are putting up an ad, they don't just want clicks, they want conversions, they want sales at the end of the day.
Speaker B:And that's what we are trying to predict with the real data that we are collecting from folks as well.
Speaker A:Right, right.
Speaker A:Yeah, that's, that's funny.
Speaker A:Flattery will get you everywhere too, Arvin.
Speaker A:I'm also, I'm always amazed at how people, young people sometimes think I am but I'm like I'm pushing 50.
Speaker A:But, but you know that which is, which is.
Speaker A:So thank you for that.
Speaker A:But.
Speaker A:But you know, that's interesting point too because like that's actually, you know, I think what a lot.
Speaker A:I'm not the only one that thinks this too.
Speaker A:But like to hear you say like the, the AI just kind of pushes everyone towards the mean.
Speaker A:I feel like that's actually what's happening in the AI generated content sphere too.
Speaker A:So.
Speaker A:So it makes sense.
Speaker A:And yet you're right to do this.
Speaker A:Well, in theory you would need some correlation with actual purchase behavior and data which you're not going to get through strictly using AI models.
Speaker A:You have to use traditional sampling and researching.
Speaker B:And the other thing that happens is, and the reason we have also we have expanded to include brand level data is even if two brands are selling very similar, let's just make it up.
Speaker B:Potato chips.
Speaker B:There are enough differences between the brands generally that unless we capture very brand specific information, the much larger Persona based model that we have will give you good recommendations, but it's not going to be as great as we would want it to be.
Speaker B:Hence our push to also incorporate as much brand specific data as we can to create this kind of fusion model that we call to be able to say for your specific brand, your demographic, even though it still might be middle aged men, but you are actually grab.
Speaker B:These are the types of people that are gravitating towards your brand a lot more than this other type of people.
Speaker A:Right.
Speaker A:Then you probably even know, right?
Speaker A:Yeah, yeah.
Speaker A:I'm going back to business school.
Speaker A:I'm thinking like there's like massive conjoint analysis going on in the background.
Speaker A:But you know, but yeah, the actual, the inclusion of the real data in terms of how the past ads have Performed has got to be a key ingredient in this.
Speaker A:Okay, so then, so if those are the foundational building blocks, what happens?
Speaker A:Like what happens next?
Speaker A:Like does.
Speaker A:So are you bringing generative AI into the equation to predict and test, like how you think the ads that the retailers or the brands are thinking about putting into the space are going to perform?
Speaker A:Like, how do you actually help them in that regard?
Speaker B:So we help them in two main ways.
Speaker B:So one is figuring out who to launch products for or who to target.
Speaker B:So that's almost figuring out the right Persona to go after in a very, very specific way.
Speaker B:Like I said, gave an example of saying moms in Minnesota.
Speaker B:It can be also extremely specific.
Speaker B:You can say moms in Minnesota who are driving the X types of cars, for example, because you may actually need to target them with very specific product for that car.
Speaker B:Because they are buying car seat, for example, for their babies, like very soon, for example, as they are growing up, like a different car seat, for example.
Speaker A:Oh, right, okay.
Speaker A:Yeah.
Speaker B:And so one is just kind of figuring out what is, what is the right Persona.
Speaker B:Once we figure that out, then we can help them create the content as well with generative AI.
Speaker B:Now, in this case, when content is created, the content is created exactly to match the Persona that you are targeting.
Speaker B:To your earlier point that a lot of the AI generated content, there is no way to get around it.
Speaker B:It's kind of AI slop.
Speaker B:You just create a whole bunch of stuff and a lot of it is towards the mean.
Speaker B:But if we know who we are targeting, we can very specifically create the content for that specific Persona.
Speaker B:So that's one thing that we do.
Speaker B:And then we basically help our partners that work with us to launch this content on their accounts.
Speaker B:The second way that we help folks is to help them literally pre validate before they launch and pre validate.
Speaker B:Exactly.
Speaker B:So if someone is launching hundreds of ads per week, so making exclusive part of their stack, so before they go and put it up on a platform and spend a few days to figure out what's going to work, what's not going to work, we say, listen, put your, let's say 200 ads through the platform.
Speaker B:These are the 65 that you should launch, for example.
Speaker B:And they go ahead and launch the 65, they don't launch the other 135, for example.
Speaker B:The 65 is only the ones that they launch.
Speaker B:And out of 65, probably 50 will do very well.
Speaker B:Rather than launching 200 and hoping that still the same 50 are going to do well.
Speaker A:That's where the Rubber meets the road, then is the predictive side of this, because that's what's going to increase your return on your ad spend, right?
Speaker A:Because you're, you're saying like, yeah, you might have created these, but you know, you're telling, you're basically telling the brand saying, based on your modeling, saying, look, this is these, these 35.
Speaker A:I think you said these 35.
Speaker A:I wouldn't advertise these because they're not going to, they're not going to make the grade for you in the way that the other 65 would.
Speaker A:And therefore you're going to get a bigger return on it.
Speaker A:That's, that's the sauce, right?
Speaker A:That's the secret sauce.
Speaker B:It exactly is.
Speaker B:And you could, you could use that to try to optimize for different things.
Speaker B:One, you could optimize for top of the funnel.
Speaker B:If there are brands that want to generate more awareness, so they are more interested in just people initially getting to know their brand.
Speaker B:So you can optimize your content and your predictive power towards figuring out what is the content that will generate more eyeballs.
Speaker B:Or you can optimize for.
Speaker B:Which is mostly what our partners would do, or optimize for sales.
Speaker B:And in that case, folks will probably optimize for lower cpa, partly because we are spending money to try to get the product to convert basically to get sales.
Speaker B:So that's up to almost the objective of the campaign that the retailer wants to launch, and we can optimize for any.
Speaker A:Got it.
Speaker A:So if you're a marketer listening to this, there's really two things you should be thinking about if you're going to try to take a similar approach to this, whether it's with you guys or anybody else, which is like, okay, you should be able to predict which of the ads that you have in your stable are going to perform the best.
Speaker A:And then the other part of it, too, which you also said, which I want to come back to, is you can also help them refine that to some degree, using analytics to say, like, you know, okay, you know, this image may need to change or this, I'm guessing this color might need to change or something to make it better.
Speaker A:Where does, where does the, where does the, the change or the adaptation to the ads begin and end?
Speaker A:Armen, like, are you going into the AI video creat?
Speaker A:Is it static?
Speaker A:Like, how do you, how all.
Speaker A:What all do you, how do you think about that question?
Speaker B:Yeah, yeah, I'll answer that in a second, Chris.
Speaker B:And I'll also mention there's a third part as well, which is figuring out who are the people, potentially a new group of people that you have never targeted that actually you should be targeting through the product because you maybe you never thought that they're actually your right customers, except they probably are.
Speaker B:So that's the third part of the platform as well that helps marketers.
Speaker B:Now to your question, in terms of where we start, we do static images in terms of creation, optimization and pre validation.
Speaker B:Again, doesn't have to be in the net sequence.
Speaker B:If someone has enough creatives that they already have a team developing them, we can just do pre validation, basically saying don't launch this and launch this.
Speaker B:But if folks don't have it, we can also just create for the right audience and then rank them, pre validate them before they launch.
Speaker B:On the video side, we do pre validation, we do the scoring, but we don't do creation yet.
Speaker B:That will come to the platform in probably about three or four weeks, but we wouldn't do creation yet.
Speaker B:We will tell you, if you run your video through our platform, we'll tell you how well it's going to do before you launch it.
Speaker A:Got it, Got it.
Speaker A:Yeah.
Speaker A:And given all the, given all the recent smoke in the, you know, the AI video space, particularly with Aries announcement this past week of, you know, just refusing to use AI generated models in any of their advertising, like I would, I think actually as a marketer I'd probably be leaning more into the static imagery in terms of taking the value of generative AI and what it can do, you know, to, to the point that we're having in this conversation, which is just like if your goal is to make your ad spend more productive, you know, that's probably where I would start.
Speaker A:All right, so Armin, I've, you know, I've, I've been doing this for eight years.
Speaker A:I've seen a lot of people pitch this, but is there a pitch, you know, something similar in vain?
Speaker A:Is there an example you can share in practice of, you know, how this works, where you can put your money where your mouth is?
Speaker B:Yeah, I'll give you, in fact, let me start with a more recent example of a well known supplements brand that we have been working with without necessarily mentioning their name.
Speaker B:So I'll tell you the situation.
Speaker B:Before extuitive and after extuitive just to kind of make that contrast perfect.
Speaker B:Before extuitive, they're launching dozens of ads per week.
Speaker B:Let's again, I'll change the numbers just a bit.
Speaker B:But they're spending.
Speaker B:The order is still Very accurate.
Speaker B:From what I've mentioned, they're spending in the order of about $50,000 per week on ads, which is not insignificant.
Speaker B:There are partners that we have to spend a lot more, but that's a significant number to spend per week for many brands.
Speaker B:And this is on a very specific platform.
Speaker B:And after three or four days, they find out that from about 50 to 70 ads that they launch per week, Chris, probably about seven or eight perform very well.
Speaker B:The rest don't.
Speaker B:But at that point, they've already spent probably half of their $50,000 budget.
Speaker B:And about 20% of the half of their budget went on ads that didn't perform at all.
Speaker B:Maybe they actually resulted in almost either no conversions or extremely low conversion.
Speaker B:So let's just say that's the baseline to start with.
Speaker B:So what we did first with them, we said, let's figure out who we are launching these ads for.
Speaker B:So in fact, we went and explored with our agents the whole space of different Personas that would be buying their products.
Speaker B:And we came up with two new Personas that we said they had never tried it before.
Speaker B:We said, listen, let's go and target these folks on some of the major platforms.
Speaker B:We went and created the images as well.
Speaker B:We launched the campaign.
Speaker B:I'll give you the stats.
Speaker B:There is about 2x of the AOVA that we accomplished.
Speaker B:Ruas increased by about 2.7, 2.7 x CTR by about 70%, CVR by about 60%.
Speaker B:And that's for just this specific brand.
Speaker B:Now, we may not see the very same numbers, obviously every single campaign, but we have run now multiple campaigns with them and consistently both in figuring out what audiences to launch campaigns and ad sets and ads for, and figuring out how do you bring the CPA down and increase the.
Speaker B:We have been able to do it compared to the baseline that they started with.
Speaker B:So that's a, that's a very specific example of a, of a, of a partner we work with today.
Speaker A:The thing I love about this conversation, Armen, is, you know, you've provided a really good framework for.
Speaker A:If I'm a marketer sitting here listening to this, like, okay, these are the types of questions I need to be asking myself in terms of how am I going to do this better?
Speaker A:So before I let you go, like, one thing I want to ask you is, like, how hard is this to do?
Speaker A:Like, what, what, what?
Speaker A:Like, how long does it take you to set this up if there's a retailer brand interested, like, what does that process look like?
Speaker A:Like, is it complicated?
Speaker A:Is it Simple, Is it straightforward?
Speaker A:What does it look like?
Speaker B:So the process, the technical process in the background is pretty sophisticated.
Speaker B:We spent probably more than a year trying to refine it to actually get to a point where we are.
Speaker B:But the process for a partner to work with us, we have made it as simple as we could.
Speaker B:In fact, that's probably half of our research went into focusing on making it a seamless process for our partners.
Speaker B:So the way we start, Chris, when a partner wants to work with us, generally we'll actually work with the right partners and say that there is whether people want to call it a 30 day trial period or I call it calibration period, partly because I want the partner to calibrate with us what we do and recalibrate their process as well.
Speaker B:And we wouldn't charge anyone if they are working with us for the 30 days for them to actually see the value before they decide to continue.
Speaker B:They connect their Meta or TikTok account with us to our platform.
Speaker B:It takes probably a day for us to build the bespoke model once.
Speaker B:Yeah, exactly.
Speaker B:So that's, we went down, Chris, from about close to two weeks down to a day now where we build the model in a day, we spend the next day doing the quality assurance to make sure that the model is going to perform.
Speaker B:And basically after 48 hours that's rolled out to the partner where we'll walk them through how it works, we'll walk them through the metrics and we'll continue using that platform with them and observing for the, for the, at least for a month, almost every day to make sure that it delivers.
Speaker A:So if I'm contemplating taking this approach, like does it matter how big I am to get the benefit of this?
Speaker A:That's kind of my, my final question in closing out Armin is like, you know, does this is, does this approach work for everyone or who does it work best for?
Speaker B:It's in terms of the size of the company, what we work with, folks who are and kind of top line revenue are 5 million up to several hundred million, for example, and some that we are now in conversations with that are in tens of billions.
Speaker B:So it doesn't really matter.
Speaker B:Chris There are features that matter though, and those features are the partner has to be launching enough ads per week for us to be able to actually make predictions that are substantial for them.
Speaker B:And if a partner is launching one or two ads per week, even if you guessed it right or didn't guess it right, that probably wasn't going to make that much of a difference.
Speaker B:So what we require our partners to have is about six months of history, whether it's on Meta or TikTok.
Speaker B:Those are two platforms that we support these days.
Speaker B:Six months of ad history and generally about 250 ads that they have launched in the lifetime of their account.
Speaker B:That's the requirement basically, to get started.
Speaker A:Yeah, I mean, that makes sense.
Speaker A:Yeah, fundamentally, like, yeah, because there's probably some brands that are listening to this, being, getting really excited.
Speaker A:But you've got to also make sure that you've got the sample size of data ready and ready to fit the modeling that you're going to put everything through, essentially, is what you're saying.
Speaker B:Exactly.
Speaker B:Because, I mean, it's like, again, some of the partners we work with, they launch 250 ads per week.
Speaker B:So for some brands, that number would be just so low.
Speaker B:But there might be also brands that are just growing and they are just trying to figure out how to even launch on Meta or TikTok.
Speaker B:And those may not be the right partners for us or we may not be the right partners for them at this stage, partly because they need to get enough traction so we can also figure out who gravitates more towards their brands to build this fusion custom model in a way that also performs pretty well.
Speaker A:Yeah, yeah.
Speaker A: y as we come back TO it's now: Speaker A:You know, there's a lot of new things that people can put towards this and your competition is going to, going to take this approach or similar approaches to beat you to the market in the advertising game with technology, you know, at the technology like we have today at their fingertips.
Speaker A:So, man, Armin, this was great.
Speaker A:I really, I, I really enjoyed.
Speaker A:I love how you laid it out too.
Speaker A:Like I said, like, anytime I can have a conversation where I come away with, you know, a framework for the building blocks of how to think about something and then the, the, the steps I need to take to execute it correctly.
Speaker A:I know it's been a really successful podcast conversations, so thank you for that.
Speaker A:If people want to get in touch with you, what's, what's the best way for them to do that, you or anyone at extuitive.
Speaker B:Yeah.
Speaker B:First of all, Chris, been such a pleasure, so thanks for having me on would love to continue the conversation in the months and years to come.
Speaker B:If people are interested, I would encourage them to go to ext.com which is with ex.
Speaker B:Then Twitter.
Speaker B:Or they can go there, find, find the team, reach out to us.
Speaker B:But you can also email me directly at armen armenextuitive.com as well.
Speaker A:Awesome.
Speaker A:Well, Armin, again, thank you so much for joining us.
Speaker A:That wraps up today's conversation.
Speaker A:This podcast was produced, of course, with the help and support of our fabulous producer, Ella Sirjord, and on behalf of Ella, myself and everyone at Omnitalk Retail.
Speaker A:As always, be careful out there.