Meta’s Black Box Is Ruining Your Creative Strategy

[00:00:00] Taylor Holiday: And I think that's why, if I'm meta. What I like, you know, people call this a black box, but I think in many ways what they're trying to do is actually like, I call it like stupidity defense, which is like, they're actually trying to eliminate you from doing things that are counterproductive to yourself.

[00:00:15] And, and so to me, this creative thing is an example where if they could actually use the actual signals about the predictive value of whatever intent or lead you called them leading indicators, actually predict purchase and then use that to reinform the creative ideation. That's the actual dream workflow, but it's actually happening using real computation and real mathematical principle, not like human interpretation and extrapolation in very slow and weird ways.

[00:00:45]

[00:01:22] Taylor Holiday: Welcome back to another episode of the E-Commerce Playbook podcast. Today I'm joined by longtime internet friend of mine. I don't think re have we ever met in person?

[00:01:31] Reza: We have not. We have not. We

[00:01:32] should change that.

[00:01:33] Taylor Holiday: still just internet friends. We haven't crossed over into the real world, but our DM history goes back many years.

[00:01:39] We've we've been rapping and jamming about all things D two c e-comm, and now focused on creative together. But Reza Kajabi, the CEO and founder of Motion, everybody's favorite reporting and strategy tool that has taken over the industry. And today we're gonna jam on a bunch of things.

[00:01:58] So excited to have you here, Reza. Thanks for joining.

[00:02:01] Reza: Yeah. Thanks for having me. Excited to chat.

[00:02:03] Taylor Holiday: And I give it, give the people. I think one of my initial experience actually with you predates the creative endeavor that you're currently in. So why don't you tell us a little bit about your journey in the D two C space for those folks that don't know you yet, and then we'll get into sort of more of what we're doing now and talk a little bit about creative and what we believe about it.

[00:02:23] Reza: Yeah, sure. It's, it's kind of wild 'cause me, me and my two co-founders have been working around this problem for actually a very long time. We started a company prior to motion called Shoelace. Shoelace later evolved to becoming a marketing agency, but how it got started was we were, we were one of the first apps in the Shopify app store at this time.

[00:02:43] This was probably 2015, 2016. Shopify had, I think something like 50,000 merchants on the platform. So like extremely early days. Um, their app store was also just getting off the ground and it was a really interesting moment where it felt like the. Facebook was starting to go a lot more direct to advertiser.

[00:03:03] There. There was a time where like people were still buying ads on Facebook through like third party partners. I think it was like the, that time period was where the Facebook pixel was starting to become really important and they were just doing a lot to allow advertisers to advertise more directly on their platform. And I think everybody who've seen the direct to consumer boom over the last decade, it, it's, it's impossible for that to have happened if it wasn't for like this perfect distribution channel of meta that like, kind of arrived at almost the exact same time that that Shopify was, was, was growing. And so we found ourselves in those early days trying to, we understood that like every advertiser that was going to every merchant on on Shopify that wanted to advertise, needed to really wrap their head around meta as a platform. And at the time there's a lot of just like technical setups, so like people needed to install their. Pixels on the website, they needed to integrate their product catalog and so on. And so like the early version of Shoelace, like automated some of that. And then later on we tried to automate the role of a media buyer.

[00:04:05] And this is kind of pre LLMs, and at the time it was a lot of audience targeting was actually a really big part of, of advertising on meta. Like I remember the, the times of like lookalike audiences and like the types of things that people would do there. And it, it used to be that the, the, the obviously creative was always really important, but in those, in those, in those early days distribution and like audience targeting was like V lever on meta.

[00:04:31] If you were like a yoga brand and you found a way to get in front of yoga enthusiasts, like you just unlocked crazy revenue potential. And so I think the, the, the, the challenge with what we were building at the time was that we were helping people automate all of these like audience targeting structures, but meanwhile, like the algorithms just getting like better and better over time and doing that. For you. So over time, I think we saw this difference probably around 2020. I feel like that was like the turning point, year 2020, early 2021, where suddenly kind of flipped. No one was talking about audience targeting anymore. Everybody started talking about creative. Maybe iOS 14 played a role in that. And, and so anyways, towards the end of 2019 shoelace became a marketing agency.

[00:05:17] I knew I still wanted to build software and so I went off with our engineering team in search of what's next, which eventually became motion. We spun out motion into a separate business and like had been working on that ever since. And it's been a fun kind of transition. And I, I remember looking at one of our, our slide decks in the early days of Shoelace where we were trying to automate the role of this like human marketer.

[00:05:39] And we're kind of coming full circle a little bit around that problem now with, with, with

[00:05:43] AI and LLMs. And so I kind of feel like we've been

[00:05:46] working on roughly the same problem for Uh, a very, very long time.

[00:05:50] Taylor Holiday: that you, say that. So one of the things I like to do when I get in these conversations is I like to go back and like look through the DM history and feel like, okay, what have Res and I have been talking back and forth about for five, six years and like, so our DM history starts in 2019.

[00:06:03] And it's funny, so it's like about a content exchange around, 'cause Shoelace was publishing a bunch of cool content and we were sort of like doing a collaborative thing around that. But then, funny enough, so September 9th, 2020, okay, I am, or in August, I'm like, Hey, I'd love to see a demo of some of the newer stuff you're working on.

[00:06:18] And you're like, cool. Let me send you something over and you send me this YouTube video and it looks like it's matte. And Nick Shackleford actually like going through the early days of motion, it's all about this idea of like manual labeling around attributes of creative Crazy, right? it's crazy this, so this is like September of 2020.

[00:06:37] Reza: Yeah. years ago you were working on this same sort of conceptual idea that like, but in this case, I think it was like humans would like tag things, right? Like as different attributes of creative and then be able to group them and report on them accordingly. And of course me, like in my usual fashion, I'm like super cool.

[00:06:54] Taylor Holiday: I'm a little skeptical of manual labeling around attributes of creative, but to be then full circle now and have this evolution of AI where now it's a similar problem that you're after, which is like, I think you've always been deeply connected to the idea that people want this intimate working knowledge of something qualitative by nature but is connected to a quantitative assessment of performance.

[00:07:16] And they're trying to tie together these ideas that are informing ideation and concept and then trying to put the data layer on top of that. But that goes back five years of the same problem, man. Yeah.

[00:07:27] Reza: to totally does. I, I think we've, we've, started to see like the things that we've been excited about all these years are just starting to accelerate. And I think a lot of it is around this idea of the collapsing of, of roles and like specialization starting to become a, a bit more, these kind of full stack generalist type individuals.

[00:07:47] And I think everybody, it, it used to be that everybody would aspire to some kind of greatness around that. So if you've ever met someone who's like deeply analytical and deeply creative at the same time, like we, we cherish that, we're like, oh my God, what a gem of a human. Like, if you're able to be both of those things at the same time, that's like a unicorn type of a profile. And I think we've always been really interested about trying to bridge those two worlds because I think increasingly it was like, I think there was a time where you could day trader your way into media buying. You know, like if you found a good, like good audience targeting hacks, you could put like whatever creative on it.

[00:08:25] It was such early days in the platform that you could make it work. But I think we were always very convinced that it was going to be both skill sets that really matter. So if you're just a creative person and you can't kind of build a feedback loop around what's working and what's not, then you're just gonna be guessing all the time and it won't actually work that well. And if you're just, just analytical and just looking at data over and over again, like that's not gonna give you the kind of creative inspiration to drive meaningful interactions with customers. And so I think that of all the things that we've been working on and all the various like implementations of trying to solve that, I think that's the thing that we're trying to get after is how do we, how do we help more people become balanced in terms of their creative skills and analytical skills? And for the people who are very originated on the creative side, we try to really introduce them to the data side and say like, look, these numbers actually are not that scary. They mean something. And think of it as like a scorecard to your work and so on. And then for the people that were coming from more of the data side, it's like trying to get them. More familiar with, with, with the creative part of things. So I think Motion as a platform has always tried to bridge those two worlds and as balanced the way as possible. 'cause I, 'cause like our view has always been, um, you need both of those things to do really well. I,

[00:09:45] Taylor Holiday: Yeah. Super interesting. And what about specifically the idea of labels? Like let, let, and I think. This term is just, I think, a categorization of you would maybe visually describe an ad to be about or to contain or the defining attributes of it. In a way that is language that is like, I think ambiguous.

[00:10:11] it's degrees of closeness everywhere, but they're always like slight variations of

[00:10:16] Reza: hmm.

[00:10:16] Taylor Holiday: refer to things as. But there's this, we all could look at, let's just say a creator UGC ad and we could describe the attributes of it in ways that would bucket it into a certain type of content. doing that like sort of consistently across a large diverse set of people where we would all choose the same language all the time is challenging.

[00:10:38] Reza: Totally. this debate I had like, yes, last night I was like arguing with Shireen just 'cause that's fun for me. 'cause I think she's great and it's fun to argue with her about like, what is a media buyer, right? Like this is an example where there's a word. That has almost no shared meaning anymore,

[00:10:52] Sure.

[00:10:52] Taylor Holiday: It's like one person will be like a media buyer does all these things, including creative and then like including financial planning and inventory discussions. And then other places it's like, no, they just work in the like, so these words get very convoluted and so language and numbers sometimes are at odds with one another in their specificity I

[00:11:11] Reza: Mm-hmm. Mm-hmm.

[00:11:13] Taylor Holiday: And so I'm curious about what you think AI has the potential to help us with in terms of this sort of challenge of trying to add a qualitative or maybe a semantic layer on top of the data layer. How does this maybe more possible now than it was back in 2020 when we were trading messages?

[00:11:33] Reza: Yeah, yeah, yeah. So I, I feel like we saw since the early days, the best teams had some kind of taxonomy around how they viewed their creatives. And it seems like the, the, the taxonomy, the purpose of it was to kind of map to the hypothesis of the teams, right? So like, the only reason to label something is that it is a variable in your hypothesis.

[00:11:58] So if I am, if I'm like creating an asset and I want to put a UGC label on this, it's because I have an assumption that UGC might work better than this other format or this other format. And so I think, and, and, and the fundamental purpose of, of these like naming conventions, were a map to our testing strategy. And so often we've, like, for years when we try to help people get onboarded into motion or just generally doing good creative strategy, having like a, a taxonomy system internally was a really good starting point because I think the problem that we're trying to solve, like everyone, all, all teams who are trying to produce more winning creatives is like, I need a kind of a map of the world.

[00:12:41] What are we testing? What do we think is driving performance? What do we think is gonna work versus not? And that could be across like messaging angles, visual side. People do a lot of this around hooks. And, and so that process, it can be very manual if people wanna kind of think about how to, how to organize it, but like critical, because otherwise you can't go back and look at the data on any meaningful dimension. And I think the thing that becomes interesting about AI is a couple things. So we, we released our AI tagging system a month ago and like we're really thrilled about it and like we've been spent a lot of time and money on, on getting it right. And there's something really interesting about it because one, it obviously automates a lot of human work.

[00:13:24] Like LLMs are, are trained on words and they're good at labels and they're good at words, obviously gotta teach them a little bit how to look at things, but like it's very core to their nature is to kinda deal with words and labels. So like, it, it automates a lot of the process for having to like manually label things.

[00:13:41] I think it does become important. What do we want to label? I remember in, in even the early days of, of motion, there were like pre LLM AI companies that made this claim that, hey, we can use computer vision to like tag certain variables of your creatives. The problem was that none of the variables mattered. Right? Like we can detect that there's a dog in this video. It's like, okay, great, thanks. Like, how's that gonna help? Because we're not necessarily testing whether having a dog or not having a dog is going to help. But I think LLMs allow you to label based on the real testing variables that you might think about.

[00:14:17] So like, here's this idea of a messaging angle. Here's what I mean when I say messaging angle. And so if it fits in like x, Y, or Z category, I want you to label it in this particular way. So like AI allow, I think we're now at a point with LLMs where you can label things with AI in the same way that like a world class creative strategy team might look at their, their creative. But what becomes interesting is that when humans are doing it, we had this interesting, we, we looked at our own ads and how our AI had like labeled them. We have this category called intended audience and it had labeled everything creative strategist. And at first we're like, wait, this has gotta be wrong.

[00:14:54] 'cause like there's no way that. We're just talking to creative strategists like our, we wanted to be talking to like media buyers, heads of growth and so on. And so at first we're like the, like I, I think, you know, something was wrong with the tagging, but then we actually went and looked at a lot of the creatives we're like, it turns out that we were really overindexed talking to creative strategists and that was not intentional. And so we thought we found a bug in our LLM system, but what we actually found is that we were talking way too heavily to creative strategists and like that spawned off in, in our internal team, this race now that we're trying to create creative to make it so that the AI is going to pick up labels to the other intended audiences.

[00:15:36] And so I think that's like a new thing that how AI sees the ad is potentially more powerful because you don't have your own biases. Like I could label something that, like we're talking to media buyers doesn't mean that we are,

[00:15:47] doesn't mean we did a

[00:15:48] good job at doing that.

[00:15:49] And so that's like a really interesting, more powerful than the automation of the labeling comes out.

[00:15:54]

[00:16:37] Taylor Holiday: For sure. It's really interesting to think about the idea that, so, so one of the things that's fascinated me, I, I've always been like really compelled by how well distributed protocols are in like, educational systems. So as an example, like when I was growing up, like we all wrote in MLA format.

[00:16:58] Reza: Right?

[00:16:59] Taylor Holiday: Right.

[00:16:59] There's this structure that got like incredibly widely distributed and adopted into a very broad scale of schooling to where we all wrote papers with one inch margins, double spaced with 12 point funk,

[00:17:13] Reza: Yeah. Yeah.

[00:17:14] Taylor Holiday: it wove its way in where everybody adopted a similar structure of behavior. And it allowed for like this like transference of knowledge and communication style really broadly in a way that I always found super compelling.

[00:17:25] I think this is also fairly common in software engineering where there's like language and protocols around how to do certain things, but in our space in eCommerce, I find that these things are like actually so fragmented. it becomes incredibly challenging as an agency. You get to see this like a lot because you start interacting with all these different teams

[00:17:44] Reza: Mm-hmm.

[00:17:45] Taylor Holiday: just like how disparate the behavior and patterns of everything from like asset management is a good example where this is like chaotic to like naming structures, naming conventions, like all these things are so fragmented and disparate that I've always wished that there was like a sort of open source protocol that gets published that allowed for the adoption of these kinds of things.

[00:18:07] And I'm wondering like, and 'cause you and I had a conversation about this, about, well what happens if Meta does this differently than you? What if there identification system ends up different than yours and now I've got sort of categorizing and identifying things as being diverse or not, and then you giving me a different one and now I'm competing with both of that.

[00:18:25] Like, you ever thought about trying to create a universal protocol that gets broadly adopted by these systems?

[00:18:32] Reza: To totally thought about it. I think the question becomes the question becomes how, because we, we can release the taxonomy and like that, that the taxonomy won't be that hard for us to like put out. But the way that AI would bucket things and categorize things like that becomes a little bit harder to just, to just open source.

[00:18:51] But I think the, the way that we've been thinking about it as it relates to meta and just in general, the, the. The, the way that I think as people try to solve for creative diversity, there's levels to it, right? Like the, the, the part that we've found is if you think about a slightly higher level of abstraction, then you know that like downstream of that there is variety. So what I mean by that is like we, we introduced a, a label for hooks and most people are looking at their hooks verbatim, right? So whenever we looked at people analyzing their hooks, they would like have these spreadsheets with the long string of text of whatever it is they're saying is their hook. And they tried to see like which hook is working better.

[00:19:32] Okay, great. That's like one way to do it. The way our tagging mechanism works for hooks is we actually call it hook tactic. So we're grouping together the various tactics that you might use for hooks. So it could be like curiosity or bold claim, or shocking statement or like, relatability or question or these types of things.

[00:19:52] And then what's interesting is you could see that. If your percentage of spend on hook tactic is all shocking statement, well, okay, we know we're not being di diverse enough then like we can, we can be

[00:20:05] diverse at the level of the hook tactic. Or same thing with visual. Like

[00:20:10] I see a lot of

[00:20:11] people that, like the graphs that we've seen where people are trying to reverse engineer Andromeda is like, they'll look at two creatives and they're somewhat similar. And then like, are these two creatives that are like roughly similar? Is Andromeda gonna group that as similar? Like the easy answer is like, yes, they are. If like, if it looks similar, they're gonna group it similar. Like it's, it's actually not that hard of an equation.

[00:20:31] The question becomes like when we think about visual, the, the place that we're obsessed about right now is this idea of like a visual format. And so a lot of people, when they're doing creative research, they're thinking about things in, in terms of a format. So what's an example of a format you could say, for example, comment, response as a format or us versus them as a format or like. Founder stories as a, as a format. And these are like horizontally diverse enough so that every derivative of like an us versus them, you might have questions around like whether these few iterations within us versus them are visually diverse enough for meta to kind of create a separate entity idea.

[00:21:16] And I think we're all gonna be guessing at that. But similar if a, a company is like, Hey, when I look at my percentage of spend, I am way too over indexed on us versus them. And that, that tends to be what happens with most people. Like they're over indexed at like one level even higher of abstraction. And then of course all of their creative is going to be visually

[00:21:36] similar.

[00:21:37] 'cause they're just kind of leaning into like one messaging and one, one visual format that has worked for them and they're just kind of iterating their way through that past winter. Whereas we want to break out to somewhat higher levels of obstruction. So I think it kind of comes back to. Just fundamental principles of which

[00:21:56] altitude level are people

[00:21:58] the issue that we've had in this as an industry is like way too focused on the

[00:22:02] lowest level of

[00:22:04] Taylor Holiday: analyze,

[00:22:05] Reza: executing a single

[00:22:06] But it's like, what does my

[00:22:07] portfolio look like? And there's multiple levels to that.

[00:22:10] Taylor Holiday: Yeah, I, I think that, that we've sort of, the challenge is like, like hook is an interesting cons thing where like, I think for a very long time it was a very normal creative strategy and probably even a best practice to say that like a single video with variations of the hook is like a good practice.

[00:22:29] Reza: Sure,

[00:22:29] Taylor Holiday: And to me it's now not clear at all that that's actually anything different to the system in any way, shape or form. right. let's say you have a founder story and you just, you have this like 30 minute interview and you're just cutting variations of the first thing that that person says,

[00:22:45] Reza: Totally.

[00:22:45] Taylor Holiday: not actually clear to me at all that meta is recognizing those as different assets at all.

[00:22:50] Reza: There. I. I know for a fact they're not.

[00:22:52] That is that you're right. yeah, that, that, that, that takes like a fundamental like practice around hook as an exam hook tactic as a concept and sort of dismisses as it as a valuable strategic initiative at all.

[00:23:05] So,

[00:23:05] so here's, here's the interesting thing. The, the thing that we, we, we went on this like rabbit hole of really trying to understand Andromeda and spoke to a bunch of different people at, at meta about it. And you know, when people say, does meta count this as the same creative or not? That's an interesting question.

[00:23:23] But then the question after that is like, so

[00:23:25] what happens when they do?

[00:23:27] Right? And what happens when they do is that they just, they just pass along all of the past learnings of that creative into that same creative. I think I lost your audio for a second. I dunno if you

[00:23:39] Taylor Holiday: Yeah. They, they assign all the priors

[00:23:41] Reza: they assault all the priors

[00:23:42] and so it might be, it might be, and this is not, I'm not saying this is like as a guarantee, but it might be. If all the priors to that creative were really good, and you can keep pushing it along with minor iterations, like that might actually be fine. What, what, what you don't wanna do is like, if you think you're making a small iteration and you're hoping that this becomes like net new learnings on metas platform, like it's, it's certainly will not do that.

[00:24:08] so

[00:24:08] Taylor Holiday: my question is like, let's, let's, let's stay on this 'cause I think this is an important topic. So I think we could agree then that in the case that something's bad and you think you're gonna iterate your way to improvement, no go, not gonna happen based on that because the new iteration is gonna be assigned the negative priors of the same ad concept.

[00:24:26] Reza: it's a

[00:24:26] great way to think about it.

[00:24:27] Taylor Holiday: is the, in the event that it's good, so let's say I have an ad that's man on the street and it works really well and I'm like, Ooh, this is working. In the event that I introduced a new ad, same priors, why would it get delivery over the existing ad? So

[00:24:42] Reza: Yeah, it, it, I, I think, you

[00:24:43] know, there's some meta that no one can even quite

[00:24:46] explain anymore, where like the

[00:24:47] same. You turn it off and then like three months later you turn it back on and it keeps flying again. Like

[00:24:52] there's some of these,

[00:24:52] Taylor Holiday: what I would expect to happen, but

[00:24:54] Reza: you know, but, but I think the, the principle to me is if there's something that is working and you do actually wanna ride the learning on that, great. But, but to try to do that as, as visually distinct as possible, I think is always a good, is always a good idea. So it's like you take the same thing that worked and then you wanna make an iteration on it, fine. 'cause you know, there's like, this concept is roughly working. I think the way you get the best of all worlds is you try to iterate in as visually

[00:25:27] diverse as possible.

[00:25:28] So like, this is the,

[00:25:30] the piece about Andromeda that seems to be pretty clear is that messaging is

[00:25:34] not actually a really big part of their evaluation system. It's very over indexed on visual. Doesn't mean like messaging is really important still 'cause like that's what makes your customers convert. But as part of meta's algo visual is very important. And so why not try to be as visually diverse as possible with iterations? Then you kind of, you

[00:25:54] can iterate on a winning concept and you can do it in a way that's

[00:25:57] like.

[00:25:58] as Andromeda friendly as possible to like extend that concept as far as possible. Just I think, and there are ways, like you can have a founder ad it's, and it's happening one time in a car, then another time in a, in, in an office.

[00:26:13] And like you change the kind of visual setting of that concept. Otherwise it's like roughly the same, roughly the same thing. You're just changing a few of the kind of visual elements of it. And I think that is a good way to successfully iterate.

[00:26:28] Taylor Holiday: I guess the thing I don't really understand is why doesn't meta just publish entity id?

[00:26:33] Reza: Yeah.

[00:26:33] Taylor Holiday: to understand the, like the, the, the value to them of withholding the transparency of when this is occurring and when it's not. I, it

[00:26:42] Reza: Yeah, that's a good question.

[00:26:43] Taylor Holiday: inspire the right behavior, which is to give us the structure to build the thing that they're asking us to do, versus it feels like we're trying to intermediate this information in

[00:26:53] Reza: Yeah.

[00:26:54] Taylor Holiday: is unclear.

[00:26:55] And I think that's, that's the part I don't totally understand. The, the, the like deep push and like presence and marketing distribution of the story and this narrative, and then this very clear way in which they say, we do this in the backend, but we're not gonna tell you when it occurs or not. Like, what,

[00:27:11] Reza: Yeah.

[00:27:11] Taylor Holiday: Why is that? Do you think?

[00:27:13] Reza: I, it, it, it seems to me like they, meta's dream scenario is probably more along, whatever's going on with flex ads is probably like a window into like what meta probably wants. My guess of what they want is they even want to take out performance metrics. Out of the creative level. Like I don't think they actually want to provide detailed performance metrics on like this creative versus that creative.

[00:27:41] I know it's not in their system yet, but they talk a lot about this idea of like the future of Andromeda being able to show ads in a sequence, right? So like when you see an ad, you can then it'll then informs like an ad that might be relevant for a further stage down the funnel and so on. And so I think they want to go like full on black box, you know, and

[00:28:04] Taylor Holiday: I agree and I, I actually think that this is, like this in this inherent lies, this like my fear. Of a system that reports, that like builds a behavioral pattern in our industry around looking at creative ad performance.

[00:28:20] Reza: sure.

[00:28:21] Taylor Holiday: so like, how do you think about that? Like if that's the direction they're moving, how, what does Motion, how does Motion fit into that movement?

[00:28:28] Reza: You know, to, we, we, we are really fired up about that direction. Because what that means is that we all go back to like fundamental principles of advertising, which is, you know, what, what messaging are you saying? And we kind of think about it as like pyramid of, you have some kind of messaging anchor that could be around an au an audience, a pain point, a persona, a product, some combination of these.

[00:28:53] But like, these are kind of the, they're, they're in the land of like the business strategy, right? And then it's like an anchor. And then what the, what the advertising team needs to do is like take that thing and then create lots of content around that content that is like, okay, this pain point. But if you are solution aware versus solution unaware or problem aware versus problem unaware. And so what you want to do is like take those anchors and really just like fill in the blanks of have I created content that is a tell, like telling the anchor message here in horizontally as broad as possible from a, from like a visual execution standpoint. And doing that with messaging that talks to people at various stages of the funnel and so on. And I think a lot of where we see motion's responsibility is actually trying to help people check the boxes on the leading indicators on like, what am I putting into the system? So if, if we're gonna go more black box, then the question becomes less about do I have precision on which individual creative is gonna, is gonna perform well or not?

[00:30:07] But it's more like, am I, am I balanced enough in what I'm feeding into, into the machine, basically? And so you

[00:30:15] might look back and say in your, in your performance spend. What is the, what is the spend allocation for this messaging theme, for example, and is that good or bad? Or what is our spend allocation for this stage of awareness? Right. And I think

[00:30:30] those are the

[00:30:31] inputs that really matter. And I think, I think we, we are, don't think it's a good idea for people to be obsessed about performance at an individual ad level because you have to put out, even like in this interim period where it's not full on black box, it's like semi black box.

[00:30:47] And we're, and we're getting there, the volume, and we've talked about this before too, but the volume of creative that people need to put out is a lot. Like in order to deploy any meaningful spend, you have to produce a lot of creative. And the, the, the question that we wanna help people answer is, how do I think about that?

[00:31:03] If I'm, if I'm producing a ton of creative, how do I decide what to create? And I think the the answers are, are somewhat based on performance. That's obviously important to see, like what's resonating in the world. But it's also as much about is, are the inputs that we're feeding into from a creative standpoint, do they hit all of the different parts of the equation that we think is important from like, messaging to like diverse enough visual formats and so on. Um, and you know, on on the reporting side, like I've actually long believed that the, the current approach to analyzing the performance of creative is like completely flawed because except for one day click attribution, it's very unlikely that someone saw just one creative, right? So it's, it's, it's actually more like a person probably saw multiple creatives before they converted and then that the lower funnel, static ad like soaks up all of the row asset.

[00:32:00] It's like, okay, great. This one seems to be performing not quite, like they're kind of working together as a, as a group. and so.

[00:32:07] Taylor Holiday: you you've created that though. Like in some ways. Like, I, I, I, I think, and, and I feel this all the time too, 'cause I think there's this like tension between when you create software, it's like creating the thing you believe versus the thing that the person keeps asking you for.

[00:32:20] And, and it feels like though, like the, the primary use case is like, best performing ads last 30 days sorted by ROAS and spend that one is the best one. Put that in a deck. Show that to everybody. Use that to inform some sort of next step strategy of X, Y, Z is the like default workflow in most Okay. So, so here, here's the interesting thing. I actually think, I don't, I don't believe ROAS and performance is a, is a really great indicator. I do think spend is a really good one though. Like, if, if something is collecting spend, there's signal there. And it's

[00:32:52] Reza: almost like if, if if you gotta make sure that the spend is not like, you know, lighting your, your, your business on fire,

[00:32:58] but it's still, it's, it's not nothing for something to like get meaningful spend and get traction from a spend standpoint.

[00:33:04] Like that's not nothing. And so, like to me. That, that's the same thing as like an ad doing well sorry, a social post doing well on, on the organic side, trying to figure out what is going to get blessed by the algos. Even just from like a spend takeoff standpoint. 'cause like so many people create a bunch of stuff, they put their heart and soul into it and they put it in the ad account and it's just like sitting there.

[00:33:26] It doesn't move. But the ones that start to move from a spend standpoint, I think that starts to sell, tell a, tell a meaningful story even before we look at any of the performance metrics.

[00:33:37] And so like, I think,

[00:33:40] you know, people say like, does click through rates or watch times, like, do these things kind of translate back to bro oas?

[00:33:46] And like, I think, actually we've debated this in the past too, and I, we hosted a debate. I think you and some one other person that was like talking about this, I remember. But I think there's something about an ad taking off from a, like a distribution standpoint, that is an interesting signal back into the machine on, okay, there's something there and then you kind of triangulate a few other things.

[00:34:08] But the, the, the, the spend signal to me is a, is a pretty good one.

[00:34:13] Taylor Holiday: Yeah, I think, I think in the event that you control for the human budget that's set on it

[00:34:20] Reza: Hmm.

[00:34:20] Taylor Holiday: I set a campaign with a $10,000 budget and another with a hundred, and then compare the spend at an ad level. The, the, I'm not gonna get, and one's on a cost control and one's on highest volume. Like they, they just tell you very different things relative to the, the bidding and structural setup.

[00:34:32] But I, I would agree, I, I prefer spend as a signal of at least a relative comparison to the options within the structure of that campaign as a signal by meta for sure. But, but the question of whether that spend is an indication that it will perform at my desire to anticipated level, I think is like, may or may not be true.

[00:34:49] It just might be the best of the bad bunch. You know, you never know. But I, I do

[00:34:52] Reza: Hold on.

[00:34:52] Taylor Holiday: a different leading indicator thing that I am really compelled by. Did you hear, do you follow Eric Seifert at all?

[00:34:58] Reza: I see. I see him

[00:34:58] on on X now

[00:35:00] and then. So, he just did an interview with. I'm gonna, I'm, I'm gonna misquote what university it was, but one of, one of the Ivy League schools published a paper on using an LLM to predict human preference for advertising.

[00:35:15] Mm-hmm.

[00:35:15] Taylor Holiday: they basically would create a a survey of people on a one to five Likert scale of like, and they weren't actually visual ads, they were briefs, product descriptions, headlines, value propositions, et cetera.

[00:35:26] Reza: Mm-hmm.

[00:35:27] Taylor Holiday: And then they would ask the humans how, like their, I, I, I don't know the exact question, but whether they liked it or not on a scale of one to five.

[00:35:33] And then they would have the LLM view, the same brief and predict the human answer, Nice.

[00:35:38] And they got to a place where they could get a really high correlation between. The LLMs capacity to predict what the humans would have preference for.

[00:35:45] Reza: Sure.

[00:35:46] Taylor Holiday: And there's something really interesting there.

[00:35:50] 'cause I think what we're trying to do all the time is some sort of version of that, right? Where we're trying to analyze some set of data to use to then inform a strategy to then make an asset that we believe is likely to produce the desired outcome. And I just think that the, like all of the work around this is basically hogwash, it's almost utter and complete nonsense in terms of what humans are interpreting and attempting to do and then jamming into the ad account and then demanding that it gets spent in order to validate the idea.

[00:36:20] And, but I think this, this, this premise that there's an intermediary step before it ends up in the ad account that actually where the LLM feedback loop exists is on the qualification of the viability of the thing. And I think that's where the why, if I'm meta. What I like, you know, people call this a black box, but like, I, I think in many ways what they're trying to do is actually like, I call it like stupidity defense, which is like, they're actually trying to eliminate you from doing things that are counterproductive to yourself.

[00:36:49] And, and so to me, this creative thing is an example where if they could actually use the actual signals about the predictive value of whatever intent or lead you called them leading indicators, actually predict purchase and then use that to reinform the creative ideation. That's the actual dream workflow, but it's actually happening using real computation and real mathematical principle, not like human interpretation and extrapolation in very slow and weird ways.

[00:37:20] So, so like, should we want that? Like, do, do you think that's like hopeful? Do you think it would be better?

[00:37:26] Reza: I, I do think it would be better. I think the reason is that, so I think the era of growth hacking is starting to like fade away. And I think the, the, the fundamental principles are coming back and the fundamental principles, like, I have a product, I'm selling it to someone. Why should they buy it? What pain am I solving for them?

[00:37:47] I think, I think like these are questions that in the, in the like last decade of Facebook ads hype, I've not seen people like talk about that much, right? It's kind of like, how can I hack the system? Like what, what can I do on like the lookalike side? Whereas now I think as we start to lose more and more controls on that side, because AI will just do a better job at, at those kind of execution details, the question just goes back to like, okay, I'm a company and I'm selling this product.

[00:38:15] I'm trying to sell it to someone. Why should they buy it? Do I know that? Do I have, I, you know, we, we, we, I think about teams that are doing really well on the creative strategy side versus not, is that they need to be able to articulate internally their messaging and who they're selling it to. And like, you know, it, it kind of all starts there.

[00:38:35] And, and those are, those are like timeless principles that we're almost being like forced, like people are being forced to really identify those so that they can. Create assets that are, that are resonating on that. So I think to me, the, the, the two aspects are like, do you really understand from a messaging standpoint what you're saying?

[00:38:55] Like why, why people should care? Why should they, should they buy your product? And then on the other side, I think this is where a lot of people, and I agree with this talk about like looking at organic social. Barry Hot has a really good example of this, where like his whole method is if you've understood who your customer is and who you wanna sell to, how much time have you spent as an advertiser consuming the content that they consume? Understanding just like on, on the organic side, what is the content that they consume that has done really well? So then you get a bit of a sense of like, the way that they like to receive content, right? The way that the, the kind of content that they find engaging. And so I think those two pieces, you start to combining them and produce assets that like, have an understanding of the way that customer, the, the types of like content visually that.

[00:39:44] The, the, the customer's used to receiving information on and have a ton of clarity on what you're trying to say to them, that's the right level. So I, I, I think basically what's happening is that we're, we're moving up into like higher levels of abstraction around how to think about the problem and not so, like, in the weeds of

[00:40:02] Taylor Holiday: Targets

[00:40:03] Reza: this attribution setting versus that

[00:40:04] attribution setting.

[00:40:05] And like, I think all of those things become a lot more black

[00:40:07] box. The, you know, there, there's some challenges with that, but I, I think it's inevitable because AI will just do a much better job than humans can on like

[00:40:17] the, the execution delivery.

[00:40:19] Taylor Holiday: you don't think they'll do better at the extraction part? Crushing

[00:40:22] Reza: You

[00:40:22] know, this is where I am.

[00:40:25] I am

[00:40:25] bullish on humans here

[00:40:27] because the abstraction part is not fixed. You know, and like we have control over this. So I, I, we can decide as a company tomorrow that we want to talk to retention marketers, right. We might start, we might come up with this idea that if we go and talk to retention marketers and say, Hey, as a retention marketer, if you start understanding what's going on in the paid social side, that will help you, right?

[00:40:55] That's a very subjective decision. And we might decide to do that. We might decide not to do that. And I think there's like, is no limit to this type of directions that people go and it's like pure subjective creativity on. And you know, it could be like the level of abstraction. And actually this is where I think things like early days of motion, we talked about the feedback loop between the, the media buying side and the creative side.

[00:41:21] And that was like a lower level of abstraction. How I think about it now is a feedback loop between the advertising and the business. So like, I think on the advertising side should inform the next, like product launches, right? You're learning signals from from the market. And then you might decide that I wanna launch a new product that goes into this direction. And it's kind of subjective to me how I would like to position that in the context of my competitors. Like if you have a third party system deciding how to position everyone's products against each other, that just like that, that's the end state where I, I, I refuse to be like, you know, maybe we get there, but at that point it's all over anyways and who cares,

[00:42:02] right?

[00:42:03] Taylor Holiday: Well, I think the interesting thing is like third party system is just like, it could be another phrase for a human, like, I like my, the funny thing I find about this is that like, let's product development's such a good example. Okay. Let's take motion or CTC, let's speak about our own businesses.

[00:42:19] I have ideas about where we should go next with our service and product. Yep. And I I think I have like, be really hard for an LLM to replicate my context because so much of it exists in a non-structured data format that it can't consume and read necessarily. but,

[00:42:38] Reza: can it decide for you. you.

[00:42:40] know what I mean? Okay.

[00:42:41] Taylor Holiday: but I believe that if I were to ask GBT, because I interact with it a lot about what market opportunity we should move into next from a service provision,

[00:42:53] Reza: Totally.

[00:42:53] Taylor Holiday: that the quality of the

[00:42:54] Reza: I,

[00:42:55] Taylor Holiday: would be really, really hard for me to do better than even with as much context as I have about my unique business and space.

[00:43:04] And I think it's like. It would be, it's up there with what, as much as any human could have about, about it. I consume a lot of information. I've been doing this specific thing for a long time, and yet I find all the time I can't actually project better ideas that effectively.

[00:43:22] Reza: I think that's true. I think that's true that the, the piece becomes like at some point you have to put a few of those ideas together and decide something that is still uniquely human. Right. And I, I don't, I don't imagine a world where. I don't see that happening anytime soon. Where the thing that happens after that chat GPT example, is that then on the execution side, C, DC now also has this like new service offering, right?

[00:43:51] Like that has to go through a process where you decide that that's right for your business. And I think because that is subjective, that's where we have some to create value. So like it, these are, I think these are not, there, there isn't necessarily a right or wrong answer here. 'cause you might decide, for example, you, you have a perspective on the world regarding this, like Sean's thing, right?

[00:44:17] Like I think it's a great perspective. You might decide to act on that perspective. Or you might say like, ah, you know what? I think this happens in 15 years. Right

[00:44:25] now

[00:44:25] it's a waste of time. I wanna do this instead. Like, that's a choice.

[00:44:30] Taylor Holiday: So I think, this is the thing that I believe is the most human thing that still remains, is that it's whether I act on it or not, is really just about the reality that I am accountable

[00:44:43] Reza: Yeah.

[00:44:44] Taylor Holiday: is that if, if whatever I decide to do, it's not necessarily that, like I shouldn't use chat GBT to make the decision.

[00:44:52] Reza: Hmm.

[00:44:52] Taylor Holiday: that the reality is chat GBT doesn't get fired. I do if I fail.

[00:44:56] Reza: Mm-hmm.

[00:44:57] Taylor Holiday: the most human thing is actually the accountability. And I think the question I'm wondering is in light of that responsibility that I have, what is actually the best mechanism to make the decision? And what I'm finding more and more is that in a lot of cases, I.

[00:45:18] With the full acceptance of the responsibility for myself, the right way to make the decision is in partnership with chat. GPT

[00:45:24] Reza: Totally.

[00:45:25] Taylor Holiday: to do it absent that. And so what I see is just like at some point do the people like, like let's just imaginary, there's a governance authority, well wait a second. What, what is the actual value of Taylor in this interaction other than we can fire him And he becomes accountable and kind of cover creates CYA if he is just using these things to source the information.

[00:45:52] And that's like a obviously like sort of an extreme example, but I see this in sort of every role I have this as a CEO, I have this experience with a media buyer. I have this Yep. with a creative strategist where I go. Hey, we're, your job is to come up with the next set of ideas for Brand X around issue Y.

[00:46:10] And I go, okay, perform that task for me.

[00:46:13] Reza: Sure.

[00:46:13] Taylor Holiday: and then they run off and do it and whatever. And then I start to go, wait a second. I noticed the format of this thing you sent back to me. I, I can see the emojis and the m dashes and the, the structure of it. Like, why do I need you to in be an intermediary prompt person for me?

[00:46:29] Like, I don't know that I need you to do that. Like at some point you start to see this like collapsing of what's occurring at the human level, sort of over and over again.

[00:46:37] Reza: The way I see it, like back to the idea of a pyramid

[00:46:41] is that you, a typical org chart forever has been this like leadership layer

[00:46:47] and then very long tale of like execution work on, on like the organization. I think what starts to happen is like that bottom part of that pyramid starts to go up so that no one actually needs IC execution work anymore.

[00:47:03] That just gets like done Like that. Just that's now everybody has that and

[00:47:08] I think what hopefully what starts to happen is. There's just more organizations that get created and more, more people. I, I do. I'm generally, I think that we're headed towards abundance and not scarcity. And so I do think the amount of productivity increase will mean a lot more companies, a lot more, a lot more teams.

[00:47:28] But I think they'll just be smaller. But I think the, the, the role of leadership has always basically been this level, right? It's like how do you take stuff that, that, a team of people who are like executing and how do you point that in the right direction? And I think more and more that team of execution is gonna be ai and even at the level of leadership, you're getting the assistance of, of ai. But for any task that could like end to end, be done with AI and more and more of the tasks can be done end to end with ai, all you need is a

[00:48:01] Human.

[00:48:01] who's gonna like, supervise that and say, go this way or go that way. But I think, you know,

[00:48:08] every human,

[00:48:10] Always wants to rise up and no one wants to be doing like grunt work, right?

[00:48:14] Things that they don't, they're too repetitive.

[00:48:16] So I think

[00:48:16] that the hopeful thing in my

[00:48:18] mind is,

[00:48:19] everyone just needs to accelerate how fast they rise on going up on in terms of like layers of abstraction. I think as long as people do that, they will be fine.

[00:48:29] Taylor Holiday: yeah. I think the other thing that I've noticed with what AI does. In a way that humans are really bad if I think about like a distribution of outcomes, right? I think one of the things that's under explored is like the most extreme versions of things as a mechanism for reaching some of the like tail distribution outcomes for in creative.

[00:48:49] I think we have this problem a lot where we're sort of stuck in this, like we repetitive copy cycle of doing basically what everyone else has done. And so like I think about like the true classic grandma ad, did you like get into this drama? You see this? They

[00:49:01] Reza: The one where Facebook just like swapped it out

[00:49:03] to,

[00:49:04] Taylor Holiday: like made a grandma wearing the clothes in

[00:49:06] Reza: yeah, that's fine.

[00:49:06] Taylor Holiday: everyone freaked out and thought it was the worst thing ever.

[00:49:09] But what, what I recognize. In that is, and I've thought this a lot about the math of creative, is that if we think about this attempt at what everyone says, they want net new reach, unlocking new customer acquisition, and you gave a system an imperative against these objectives, like the kinds of things that it would produce outward, would likely be pretty large deviations from the present.

[00:49:35] Reality like that, from a first principle's idea makes sense to me. Like

[00:49:38] Reza: Sure.

[00:49:38] Taylor Holiday: do that. But what happens when we see it is it's, it's like it's stark. It like, it, it it, it startles us almost. Its the rock music on the static image ad that you're like, whoa, that doesn't fit at all.

[00:49:52] Reza: Yeah.

[00:49:52] Taylor Holiday: get resistant to it.

[00:49:53] And it's, it's that actual like natural tribalism and like safety mechanisms that I think are very human, that actually like, I think, prohibit us in some way that the, the system like partnered with, I think doesn't, isn't governed by any of that like preconceived sense of what should be like or what is acceptable.

[00:50:12] You know, it's like it

[00:50:13] Reza: You y

[00:50:14] Taylor Holiday: beyond that

[00:50:15] Reza: it, it does, but there's always a boundary. And I think we are all, we, we will

[00:50:19] have the ability,

[00:50:21] Taylor Holiday: into it. Right. Like, like,

[00:50:23] Reza: sure. But everyone will program some boundary. So someone might say like, use as like much vulgar language as you want. All good. Go explore that. See what you find. Someone

[00:50:33] goes like, no, no, no, sorry. Like I just, I don't care about what you might find if you explore that.

[00:50:38] I'm not interested in that. Here's the guardrail.

[00:50:41] And so someone might say, you know, and I, I, I think ultimately we will set those boundaries and that's kind of where the subjectivity comes into it. And like we have. Human agency to decide, like, for my business, here's what I think makes sense for my business.

[00:50:57] Here's what I think makes sense.

[00:50:58] Taylor Holiday: you think, like, are we into like a question of like, what, what is

[00:51:02] Reza: I don't,

[00:51:02] Taylor Holiday: is the appropriate way to set those boundaries?

[00:51:05] Reza: I I I think it, you know, the, the, the, the, the buzzword language around this would be taste, I think. So like, somebody creates a brand and a brand in some sense. It's like a, you know, you, you obviously wanna make money, but people are doing that with some level of taste that they're putting out into the world. And I think people will care about that. I don't, I don't think we're gonna throw that out out the window anytime soon. And so people who start companies I think are gonna be people who have a lot of taste and have a lot of perspective on what they want to create into the world. And I think it's that taste and perspective that will mostly be the thing that's like creating that guardrails of like,

[00:51:47] within this worldview.

[00:51:48] Taylor Holiday: if you, had to, okay, so let's this word, I love this word taste. If you had to like apply a formulaic assessment of whether or not someone has it, what, what would you do? How would you assess that?

[00:52:00] Reza: I don't think, I don't think, I don't think you can assess it alone. I think if somebody has good taste in a specific domain, probably that domain has mechanisms to measure it,

[00:52:15] right? So for example,

[00:52:16] Taylor Holiday: like,

[00:52:17] Reza: okay, here's an example I think. I think Mr. Jess Bachman has phenomenal taste, right?

[00:52:23] I could say that subjective.

[00:52:24] And when I

[00:52:24] talk to him like this is a person who's got good taste when it comes to creative, and I think he's got the track record to like back up that taste. So I think in our industry it

[00:52:33] would be

[00:52:34] Taylor Holiday: So, what, what is that? Yeah, yeah. Here we go. Okay.

[00:52:36] Reza: like consistently producing ad creatives that have performed really well at a large scale.

[00:52:42] Taylor Holiday: So, okay, so then taste is just strictly a financial measure and there's just no way, there's just no way that under an objective outcome like that, that you're going to outperform a system that can iterate

[00:52:54] Reza: Well, you, you, I think, I think in that, in that and I, I'm not saying it's possible that like a, a system can do it. I'm not saying it's, I'm not saying it's not possible, but in, when Jess works with his clients, he's negotiating some parameters around his exploration areas, right? He must be doing that to some degree.

[00:53:15] Okay, here's how wild I'm gonna go. Here are the things that might try, are you okay with that? You're not okay with that. It's like some level of negotiation. But there's still something that happens there around like

[00:53:26] a negotiation of boundaries.

[00:53:27] Taylor Holiday: right? Like they have a brand standards.

[00:53:29] Reza: Do they though? But I, I think someone like Jess, for example, would like push that and say, oh, but have you thought about this?

[00:53:34] And have you maybe consider expanding? That, and that could certainly happen by a system. I, I think that could happen by a system to say like, I think your perspective is too narrow right now. If you were to expand it in this way or that way, I think it would produce really great results for us. But like it goes back to that accountability thing. Somebody on the other side has to be like, okay, I like this. And so I think what ends up happening is like in the systems future, Jess is on the other side of that and he's got all these crazy systems that are telling him like, Hey, this perspective, that perspective, I like this. I don't like this move this, don't move that.

[00:54:11] Right? Like the final accountability piece on how to set the parameters.

[00:54:17] But someone's gotta set 'em.

[00:54:20] Taylor Holiday: yeah. I, I think, I think the boundaries, the what, the problem with the idea is that if it's anything other than performance, which I think is like what businesses primarily exist to be

[00:54:31] Reza: No,

[00:54:32] Taylor Holiday: to, at the end of the day, like there's just

[00:54:34] Reza: that's not.

[00:54:35] Taylor Holiday: at the highest levels, like I think about fashion brands.

[00:54:38] Reza: Sure.

[00:54:39] Taylor Holiday: great CEOs get fired if their business doesn't sell product.

[00:54:45] Reza: Totally,

[00:54:46] totally. No, for sure.

[00:54:48] But that, that's a, I think

[00:54:49] that's the secondary thing, like

[00:54:51] primary thing is like delivering something to the world of value, right?

[00:54:55] And

[00:54:56] Taylor Holiday: value is a financial word. Re That's like

[00:54:58] Reza: it is, it is. It, it, it is. But the way to determine what creates value happens before the financial measure. So you might say like, I think if I did this for this person, they would find that valuable.

[00:55:13] Okay, let me go see if

[00:55:15] you did a good job of that. But basically,

[00:55:18] Taylor Holiday: and measured by

[00:55:19] Reza: the finances are the, you know, what's a good way to think about it? The finances are a lagging indicator of value, not a leading indicator.

[00:55:28] Taylor Holiday: They're

[00:55:28] Reza: You need.

[00:55:29] Taylor Holiday: indicator of value. It's like

[00:55:30] Reza: Uh,

[00:55:31] Taylor Holiday: of the word.

[00:55:32] Reza: yes, but la lagging, like you, you, you gotta start somewhere. Let's say you're starting a company from scratch.

[00:55:37] There are no financials, right? There are no indicators whatsoever. So you need to form some something somewhere that if I did this, it would create value. If I'm right, I'll see it in the numbers,

[00:55:48] right? And so I think you, you, you can't, like, you can't build strategy off of la lagging indicators. Like it just, especially in an age of ai, because everyone's lagging indicators are gonna be like a solved problem, then all we have are untapped creativity on the leading indicator side of like, what, what might I do?

[00:56:10] Like what, what, what could I do that is unique or different or interesting? like I, I just, I.

[00:56:17] Taylor Holiday: I just don't know what they are. Like I, and if they, if they exist and are real, then they'll produce financial outcome, and in which case the feedback loop, I, I don't think they're that lagging, like the idea that like. Money is lagging. It's like, well, sort of, except our entire industry is predicated on

[00:56:31] Reza: What, put it this way, put it this way, if, if, if a company got up and put an ad and said, we don't care

[00:56:37] about you and we don't care about the product we're selling, all we care about is like how much money we make,

[00:56:42] the customers wouldn't appreciate that very much. Right? Like, and so either that's a dishonest experience, which I don't think so.

[00:56:48] Like I, I, I don't think everyone's

[00:56:50] just being dishonest. Like people

[00:56:51] want to

[00:56:53] do something mo like I think the best companies wanna do something that customers appreciate along the

[00:56:58] way. Their profit maximizing capitalism's great, no problem with that. But I don't think there, there is something higher than that, that I

[00:57:07] think people do care about and they get fired up if they can do that thing and make

[00:57:11] a lot of money.

[00:57:12] But I I don't believe.

[00:57:13] Taylor Holiday: think it's higher because I don't think you get fired if one is true and the other isn't. I think that's just the simple is that like, if the first is true, if you are crushing it on financial value, but you're below the line on like taste, like I don't, I don't think anybody loses their job over that, but they

[00:57:27] Reza: You,

[00:57:28] Taylor Holiday: on the the other one all the time.

[00:57:29] Reza: you might lose the trust of your comp of your customers. So like people who maximize finances in the short term could

[00:57:36] lose all of their financial

[00:57:38] returns in the long term. In the long term. 'cause they burn

[00:57:40] trust. Like you can burn all

[00:57:42] customer trust in the short term, the spreadsheet will look terrific,

[00:57:46] right?

[00:57:46] But the spreadsheet after, that's gonna look terrible. And

[00:57:48] so that takes, I

[00:57:49] Taylor Holiday: a horizon question. That's like a of, of the assessment of the measure, but that doesn't go back to taste like to me, like I

[00:57:56] Reza: sure it does.

[00:57:56] Taylor Holiday: the the, problem is just when we use this phrase, we like often re re rescind the obligation.

[00:58:05] the reason the business exists and the efficacy of it relative to that measure. And it becomes dangerous as a tool of manipulation over all sorts of things, including process and communication and in ways that like people will exert the idea or the presence of this thing. It's in the room with us here, taste somewhere and I have it and I have it, and you don't.

[00:58:30] But there's no way to point to the reality that it exists in any way, shape, or form relative to the business result we're trying to create.

[00:58:38] Reza: Yeah. I, I think it comes back to the example that you mentioned about delegating something to someone, and

[00:58:44] if you delegate That could be

[00:58:45] done by ai,

[00:58:47] Taylor Holiday: verbally, are

[00:58:47] Reza: then that that individual has not done anything of

[00:58:51] of value. So the

[00:58:52] question becomes

[00:58:53] like.

[00:58:54] Taylor Holiday: and

[00:58:54] Reza: Where is there value left to do? Because like, you know, things like modeling and financial planning and like a lot of these things, they're still not perfectly done by, by ai, but like they will happen, right?

[00:59:05] Like the, it's just gonna get so easy to, to do all of that. The question becomes like, then what? It's either

[00:59:12] the system does

[00:59:13] everything. So like how, how far does it actually go? Or there's something

[00:59:18] Taylor Holiday: we

[00:59:18] Reza: at the very top that's left for humans to do. And,

[00:59:22] and even that, I

[00:59:23] would say it's like

[00:59:23] some time period, a after a long enough time period, I actually do believe

[00:59:27] that

[00:59:28] Taylor Holiday: gonna honor that scope

[00:59:29] Reza: our entire world is

[00:59:30] gonna look very different. But

[00:59:32] the, my time

[00:59:34] horizon on that is like,

[00:59:35] Taylor Holiday: then. Yeah. Like, it sounds like maybe we agree then like I don't,

[00:59:38] Reza: I I think, I think it's something like this. Humans have got like a, probably a good decade of rising to a high level of abstraction to create value with these AI systems. And then after

[00:59:51] that I have no idea. What happens, but like for some period we still

[00:59:56] And that role to play is somewhere in like the fuzzy land of subjective decision making that hopefully is measured in

[01:00:04] real financial returns. Right.

[01:00:06] But I think it's, there's an art to it.

[01:00:08] Taylor Holiday: next round.

[01:00:09] Well, in the meantime, you'll keep helping us analyze the data of today's performance of that creative in an attempt to turn it into action. So, I'll look forward to seeing the taste metric inside of motion that we could try to, to capture and bottle up. But in the meantime, I I I am hopeful that there is a lot of good information out there that we can feed into our LLMs and teeth just to, to make more cool things.

[01:00:32] And I'll leave the taste to Jess. So, Reza, I appreciate you man. Where can people follow you and what's the best way to interact?

[01:00:39] Reza: Yeah. I'm on, I'm on, I'm

[01:00:40] on Twitter. Reza Kaja at Twitter. You can visit motion@motionapp.com. We have a newsletter that goes

[01:00:47] out every Sunday called

[01:00:48] Thumbs Stop, and our team puts a

[01:00:50] lot of effort into that. Uh, so

[01:00:52] if people wanna stay up to date

[01:00:53] on how Motion thinks about the world, that's a

[01:00:55] good, it's a good newsletter.

[01:00:56] Sign up for.

[01:00:57] Taylor Holiday: And hey meta, if you're listening, partner with this guide to create a universal categorization that we can all interact with. Because I think you're both doing cool stuff. And it would be really helpful for me, just a lowly agency guy trying to sit amidst a bunch of customers and get them to agree on the definition of words.

[01:01:14] That would be really helpful. 'cause I think the, the idea that there is a classification and categorization for this stuff would be so helpful to all of us Reveal entity. Id let a, let Motion pump it into their system so they can tell us if we're creating diverse, creative or not. That can help us on the creative planning side.

[01:01:30] I hope to keep doing some cool stuff with you. Reza, I think you guys are killing it. You have a lot of the respect of the industry in ways that clearly people have just seen you as the the definitive source of truth for what is, what is happening in their creative landscape. So keep killing it and looking forward to hopefully revealing some of the cool stuff we you're hoping to do together in the next year,

[01:01:48] Reza: I love

[01:01:49] it. Appreciate it. Thanks for having me.

[01:01:50] Taylor Holiday: Thanks buddy. Take care.

[01:01:51]