12 Years of Agency Knowledge in One Document

[00:00:00] Taylor: That was a good, that was a good callback right there. Um, no, I'm, I'm just gonna talk. I don't, I don't need the mic. Um, I lost to Aaron. That's the joke. Aaron was the pro homecoming king. So that's, uh, that's where the runner-up was. Just, not just runner-up, runner-up to Aaron in that, uh, in that race. I appreciate you all being here.

[00:00:19] This is cool. Um, good amount of folks, uh, in store. You noticed something on your seat today that we're gonna talk a little bit about, um, called the Canon. That is a very expensive print job, so please treat it with care. Um, rush delivery of full bleed print, not, not ideal. But we'll get out in front of it next year, right?

[00:00:37] Louder? Louder. Louder. You want a mic? Okay, fine. I was... I wanted to just Bieber at Coachella this thing and just sit here and type on this thing, but that's fine. Um, okay, today what I wanna talk a little bit about is who we are and how we work on your behalf, and this is something that has evolved a ton.

[00:00:56] I'd say the last eight months of my experience of CTC has been more dynamic than any of the 11 years before it. In, in a lot of diff- for a bunch of reasons. Um, but the biggest of which is what AI is doing to our business and our capacity to service you, um, and also the demands and expectations that it places on us as a service provider to continue to evolve our capacity to meet what you yourselves, um, are capable of as well.

[00:01:24] What I'm experiencing is actually something really exciting, which has been our dream all along. So an agency ultimately has a promise that it makes to you, which is that there's some set of collective knowledge that we've gained over many years and many repetitions and many efforts with brands and failures and wins that have somehow been institutionalized and then get deployed through every one of these people, right?

[00:01:47] CTCers, wave. Look at all of them over there, happy and smiling. Okay? But here's the problem with that as an idea. Um, let's see. Nick, where are you? Are you there? Nick, how long have you been at CTC? Okay, so Nick has been here since October. So if I told you Nick possesses 12 years of experience at CTC deployed on his client's behalf what you guys literally do to us is you go to LinkedIn and you go, "Nick's only been here six months.

[00:02:17] That's not possible. He's not experienced enough. He doesn't, couldn't possibly possess that knowledge." And that's fair because the reality is none of us do, not even me. I wasn't on every client call for 12 years. I didn't work in every account for 12 years. So despite the fact that I've worked at the company for that long, I don't possess the unique knowledge of everything that's occurred over the history of CTC Um, and so that presents a problem for us in terms of our capacity to deliver against the ideas that we sold you on when you showed up.

[00:02:47] Whatever you clicked on, on the content pieces and came to the website, there was some promise that was made that we would deliver some methodology on your behalf, and we have to figure out how to do that. And we have to do it through new people all the time. Um, and so that's what I wanna talk a little bit about, uh, today and how we're continuing to evolve that, and how it's gonna show up for you tomorrow in your one-on-ones with your team, and how it's going to change the way we continue to work together.

[00:03:09] But first I have to, uh, admit that I am a little emotionally distraught, um, and a little down. Yesterday I didn't, wasn't at the Angels game with you guys. I was, uh, at my own Little League game, uh, with my children, my 12-year-old boys. I have twin boys. And we lost in the semifinals of the, the TOC tournament four to three.

[00:03:28] Uh, George came out and supported us, uh, which was great. I, I, I wanted to invite all of you to our game. To be honest, I think it would've been more entertaining than the Angels. But, um, it would've been a bit selfish if you all had received an invite to the Huntington Beach Little League game, so that would've been a little odd.

[00:03:43] But, um, that whole experience of coaching really occupies my brain in ways that, uh, are equally as competitive with CTC. And because baseball is this, like, perfect arena of being able to combine strategy and data, uh, and information on the behalf of 12-year-olds, um, it really is something that I think, uh, parlays a lot into what I'm watching happen inside of our own business.

[00:04:07] So I wanna tell you a little bit of a story of how I behave in this arena, which is a little bit obsessive and neurotic, uh, not too dissimilar to most areas of how I approach my life. But, um, so three weeks ago my son, one of them, made the all-star team. Always a devastating experience for twins when one of them has successes that the other doesn't, but, uh, we can talk about that over a beer tonight if you've got kids.

[00:04:27] Um, but, but when that happened, I was also, uh, able t- to become the coach of the team. And like any excessively competitive neurotic dad, I started thinking about what edge I could create on their behalf using the tools available to me. Um, and so I, uh, I had this repertoire of experience. So I played baseball for 20 years.

[00:04:46] I got a chance to play professional baseball in the minor leagues for the Yankees, so I had a sense of what it looked like to prepare at an elite level for performance. Um, and my kids and I went, um, one of my... My college roommate is now the head coach at UC Irvine here, and so he lets my kids come and sit in on the pre-game meetings, and hang out in the dugout, and watch them play.

[00:05:03] And so before those meetings, they all gather about two hours before the game, and they pass out these scouting reports. These are images that I took of those scouting reports, of everything the opposing team's about to do. Because at the college level, now every single play of every team in the country is recorded and databased on video, and you can search it.

[00:05:20] And so that allows them to build these incredible repositories of information from heat maps, to what pitches guys swing out on two-two counts, to do they struggle with off-speed, or change-ups, or fast balls? And so they sit as a team and they go through this information. Um, well, of course this doesn't exist at the 12-year-old level, um, because nobody is that insane yet.

[00:05:40] Um, but the information technically does exist. Um, and that's where I was like, ooh, I have an idea. Um, I'm not UCI and I don't have the resources of a college team, but I do have three things that I think illustrate the evolution of what is evolving in the world to create capacity for people to perform at the level of an institution, even at a Little League team.

[00:06:01] So the three things are, one, I have experience. I have two decades of playing the game. I know the information that is important to look for. The second thing, and this is really what is important, is that I know the data source to get the information that I want. So in the youth sports world, how many of you have kids that play sports?

[00:06:17] Okay. So we're all familiar with GameChanger then. Okay, so GameChanger is the sports center. Uh, it is the hub. It's the ESPN app for parents of kids. Every game, every box score, live streaming. It's a multi-billion dollar app platform that is... produces more media content per year than the entire history of the ESPN library.

[00:06:39] Um, it is an unbelievable platform, and what it does is parents every game log in and play by play, they record what's happening in their kids' games. They keep a box score, and they keep a video database of everything that happens. Well, um, because I'm aware of that, um, I was able to create for myself, um, a user that just happened to present like they were a member of every opposing Little League I was going to play.

[00:07:06] So I was able to join the GameChangers for every one of the surrounding teams' Little Leagues, which meant I could scrape the play by play of every game these kids have ever played. So these 12-year-olds, um, I now know more about them than anyone. Um, and what that meant was I could feed this combination of the knowledge that I wanted with the correct data access and my friend Claude, uh, to build a scouting report to help us understand exactly what we wanted to do for each team.

[00:07:36] So what you'll see here is my, uh, uh, my scouting report for the 12-year-old All Stars. Okay? So at the top I've got all the leagues we're playing, Costa Mesa, Ocean View, Fountain Valley, Seaview, Huntington Valley and Huntington West. And then I have an output of the team's performance across a series of metrics that I care about.

[00:07:52] Do they swing at the first pitch? How often, um, are they walking? What is their contact rate? Things that I care about. Every aggregate stat and analysis of who each kid is, the ones that are gonna be the most important, every individual stat. And then if I wanna go into the individual player pages, I have spray charts of where they hit the ball and everything for every kid in the city, okay?

[00:08:12] And I can make this in 45 minutes. I can now in forty-five minutes know everything that every twelve-year-old did across the city for multiple years. And some of them who play travel ball, I pull their travel ball stats in too. So I can have hundreds of plate appearances for every twelve-year-old that I'm about to play in two weeks and know exactly where they're gonna hit the ball, what pitches they struggle with, on and on, and on, and on.

[00:08:35] Um, and when I sent this to one of the other coaches who works for Scott Boras, who's one of the biggest agencies, uh, if you know Scott Boras, he's sort of a legendary baseball agent. Uh, and he asked me if I had hacked the Boras database to gather this information on these twelve-year-olds, but the answer's no.

[00:08:49] Actually, it was very simple. It's not complicated. It actually was just the capacity to understand what I was looking for, what is the root source of the data, and then how could I compile that into information that would be useful? And off Claude runs while I went and made a sandwich, scrapes every play that had ever been run, and presents back the information instantaneously And so when I think about that, what I see is what the opportunity that's in front of all of us, which is that if you can combine some level of unique knowledge with the correct data access, plus AI, you create a competitive advantage.

[00:09:24] Now, does that move our chances of winning up 2%, 4%, 5%? At the end of the day, some little kid has to get on the mound, and if I ca- tell him to throw a curve ball for a strike, he's still gotta execute it. Nothing in my data makes that happen. But I know that kid's gonna struggle if he does, 'cause I have that information, so I create a slight edge.

[00:09:41] And that's really what I feel like our responsibility is on your behalf. Um, we've talked before, and I got in a little bit of trouble at this last meeting because I sort of, uh, deferred the responsibility of growth back to some of you. I said, "At the end of the day, there's only so much we as your partner can do.

[00:09:59] Um, at the end of the day, it's your product, it's your brand, it's your vision for what you're creating that's gonna define how the market responds to what's happening." But as I think about that, I think that our job within the context of that is to provide you a competitive advantage in that process, so that if you're competing against other people with similar options, that there's some way in which you are advantaged in that process.

[00:10:22] Um, and that's where I began this process of thinking through our institutional knowledge. Um, and so what you have sort of sitting on your seat is an example, an illustration of what we consider to be the source code of CTC, the knowledge aggregate of those 12 years that is being deployed on your behalf.

[00:10:43] And this, this document does a bunch of things. One, um, is it creates accountability for us to you so that you know how we intend to work and why. Um, I think that any agency should be forced to contend with the presentation of their ideas into operational form for you. They should be able to tell you why they're doing what they're doing and how.

[00:11:08] And for us, it forces us to clarify that. Um, a lot of these things exist in tribal knowledge format, in reality, in most organizations. That there's a story about a thing that happened once with a client that you heard in a meeting that becomes the way in which you're supposed to interpret exactly then how to behave in another ad account for a different client in a different instance.

[00:11:32] And there's an expectation that that knowledge is sort of diffused throughout the organization, and the new people pick it up from the old people, and off we go like we are a tribe in the early, uh, you know, 200 AD, sort of passing it down knowledge about life. And actually in a, uh, environment where you're in person, that's actually fairly effective.

[00:11:53] Um, it's, it's, it's actually hard to do better than in-person communication and tribal knowledge being passed down from listening and being around people constantly. But that isn't our reality. We're a remote company with people from all over the world, literally. We've got people in here today from Japan, from all over the place.

[00:12:12] And so we can't depend on that as an idea. It just fails. It fails to meet the necessary ability for people to deliver the service at that quality. And so the canon is something that we've been working on. It exists in many different form factors, but it exists for you today, uh, as a print, as an accountability to what we believe and why.

[00:12:33] Um, and my hope is that you pressure test it, is that you read it, and that you argue with me for the next 36 hours about why I'm wrong about all of it, about how your knowledge or experience could help to shape it and transform it and improve it. Um, and my promise to you is that this edition that you see today will never be the same again, that the purpose of it is that printed today is the present state of methodology as deployed by CTC, and it will never be like that again because we are committed to its constant improvement.

[00:13:05] Um, and we are also committed to using it in a way that ensures that it's deployed as often as possible on your behalf. We call this idea organizational integrity. Uh, how often do we do the things we say we're going to do? And this is actually really hard, um, in part because you all don't always agree, and you all have high levels of influence on our behavior as well, and sometimes we capitulate too often to your ideas.

[00:13:34] Sometimes your ideas are better. Sometimes there's ways in which we just failed to implement it upfront. There's all sorts of reasons why organizational integrity can degrade, but what I've gotten so excited about with AI is it gives us the capacity to see it. So in the process of building this, what we're able to do is to actually turn it from prose, which is what this is, a narrative structure about methodology, into something that looks a lot more like literal source code.

[00:14:02] So, um- In there is a letter from me, like a letter from the editor, and it sort of reiterates what I said, which is that the promise of an agency is that it can operationalize its ideology. What we believe about how to grow a brand has to be reliably translated into action on a Tuesday morning by someone who was not in the room when the ideology was first formed.

[00:14:22] Um, and in this right now are seven chapters of material that we've wrestled over. We've analyzed data to present back why we're in the position that we're in. But the seven s- uh, the seven sections that exist today are what we call the technology stack. So it gives you a sense of how this is built, how we think about the organization of the information.

[00:14:42] And one of the things you're gonna notice, just like my story about the scouting report, is that the number th- one thing I would challenge all of you to think more about inside of your own business is, how is your information organized in a clean and accessible way to make your capacity to leverage these tools even better?

[00:14:57] So we talk a lot about that. We think a lot about the starting point at which we're going to build the stack of information that we're building on. We go through forecasting and modeling, marketing measurement, meta advertising, Google advertising, email strategy, and creative strategy. And then we're currently writing new chapters on TikTok Shops, Amazon, and App Eleven that will be in the next edition.

[00:15:16] Our goal is hopefully to update every quarter with an update to the canon. So if we have a test that we're running right now on incremental attribution that McKay is leading of like, okay, we have a number of brands where we're trying to decide, is incremental attribution optimization a default part of how we're going to operate going forward?

[00:15:34] Well, we need enough information to feel confident about that. And so we build a hypothesis. We begin to run it in a number of places. We analyze those results. And in the event that we think it's accretive to the process, we'll add it into the default structure. But the same way, we're going through that with TikTok Shops and Amazon and App Eleven, but you'll see those sections soon.

[00:15:52] Now, this is the important part. Okay, so the canon is written twice. It's once for you and once for the machine. Um, this is written, uh, again, in a narrative format so that you could read and interpret in the way that humans communicate, um, in our language. But the reality is, we have to turn that into something that allows the AI to actually operationalize it, to pressure test it against real-world situations.

[00:16:13] And so that, uh, you may have heard the phrase markdown files, dot MD files. These are often references for the kinds of information systems that AI tools are using to process and analyze and make decisions. And a lot of times what that involves is, like when I would create it, we would write the meta version We would s- talk with our, uh, Claude instance or OpenCar, whatever anybody's using, and it would ask situational questions.

[00:16:38] Okay, so I've interpreted the information here. Now, what if this happens? What would I do? And you can just do that for hours and hours and hours and hours on end to create what is functionally a complex decision tree that covers as much surface area as possible so that as you encounter the problem, you build the ability to make a decision through the system, right?

[00:17:03] We wanna defer decision-making from Nick's individual bias, not because he's not super subjective, but because he's been here for six months and doesn't possess the years and years of institutional knowledge into systemic decisions. How many times can we do that? How many times can we take and reduce the dependency on any single individual to solve every individual nuanced problem and use the systemic information to do so?

[00:17:28] Um, so what that practically turns into, um, are skills, uh, in Claude. So, um, for those of you that are familiar, um- A skill is just a markdown file that you can give to Claude that will reference every time you talk about a specific task. So if I take the meta section of the canon and I create a markdown file, what happens in our corporate instance of Claude is every time someone asks a question about meta or references an account, the response first passes through the markdown file.

[00:18:05] So if I ask, "Hey, can you analyze or use your meta ad skill to analyze Riddle's meta account via the MCP? Make sure you cross-reference any changes to the canon structure. Again, Zoom and Slack communications with the client to call out any specific requests." It's going to give me a breakdown, a full written format, that if you were to gut check against the information in the canon, you would find consistency.

[00:18:26] So now what I have is a team of media buyers that are not dependent on their own capacity to read this, learn it, and deploy it perfectly in every instance. Instead, they're an editor. They're a filter. They get to gut check. They get to read, resume, understand, provide additional context. "Oh, hey, no, we actually made a decision with Riddle that they wanted to break out this specific subset of creative because it was really important to them because they spent a gajillion dollars on it," or whatever the reason may be.

[00:18:54] That's why we're violating the four campaign structure, blah, blah, blah, blah. They can provide that additional context. But for the most part, the burden of them on filtering the question or answer is just reduced dramatically through the systemic implementation of the idea. And what's great about that is we can take this idea...

[00:19:11] I can take it to Ben. Ben, wave over there. Ben's from Meta, okay? Um, and I can say, "Ben, I want Meta to critique this methodology. I wanna know that we have a POV from you that this is something that you think represents consistency with what your implementation. If not, why not?" Right? So we can sort of source from not just our own experience, but where is the institutions that so often we end up on a call and it's us and it's Brad and it's the other Meta rep for Sunday Read, and it's our rep, and we're all arguing about the structure and why it should be that way, and everybody's got a different point of view.

[00:19:43] And what you experience from them is, oh, that's just a Meta rep with their own individual opinions too. That's just more people with individual opinions about what we're doing and why, and we're all arguing over it to no end. Um, so how many times can we defer that? And so we do this. For everything in the canon, there's a corresponding skill that lives inside a cloud that allows us to reference and deploy on your behalf.

[00:20:04] So I, I realize as we started talking about the profit engineer- And this idea that we were collapsing roles into one, right? Like, one of the things that's happening right now in the industry is there's obviously this huge conversation about is AI about job destruction, job creation, what's happening? And our message, as I reflect on our own marketing over the first quarter, I think it was reductive in a way that we presented it as if this person was able to simply by themselves replicate the work of entire teams on brands.

[00:20:36] And that's not intended to be the idea at all. In fact, um, raise your hand, guys over there, if you work on the data science and Statlas team for CTC. Devs, too. Bo, guys, come on, raise your hand. Ikra, come on. Okay. There are- That's... They're... Get them way up there, guys. Come on. Don't be afraid. There you go. Okay.

[00:20:58] This group of people, independent service responsibility for any individual brand, spends every day trying to analyze the information set and improve the underlying information models and reporting. So the reason the profit engineer can exist is because the canon exists, because the team of supporting resources exists to enable them to deploy the work effectively.

[00:21:21] And so the idea that that's just a person is reductive to the truth of what it requires to get there. Um, but it is powerful when you don't have to spend countless hours analyzing an ad account through this lens yourself. It saves you an immense amount of time and energy. Oh, sorry. Oh, we're gonna FaceTime.

[00:21:49] Um, so I, I give you an example. So this is what a skill practically looks like. It's literally just, um, a set of written rules that's a really long file that goes through a bunch of if-then style scenarios. And that's, again, another thing that we're constantly updating. Meta is a very dynamic platform.

[00:22:05] There are things that we're wrestling with all the time that also have to come into conflict with what I call the cultural Overton window of expectation, which is this idea of like, do you want AI enhancements on? Are you okay if there's rock music played on your ad for a mailbox? I don't know. That's a decision that in many ways you all have to hold for us that's sort of disparate from is it effective.

[00:22:30] And so there's this balance all the time between the way in which the default structure lives and maybe the evolution of the skill that we could develop for Adorn's specific meta ad skill that has expectations of do's and don'ts and words that are said or not said or images that aren't presented certain ways.

[00:22:47] And so that base foundation begins to be a representation of what we can build on top of with you. It's to understand what matters to you uniquely or individually, things that you want to try or override or whatever you want in your individual capacity. But it's a foundation that moves us forward a lot faster.

[00:23:04] Um, so that unique knowledge is what's contained in this book. It's, this is, uh, the starting point at least of trying to write down what we've learned over the last decade of doing this on your behalf. And my hope is that every time you come, it'll be a little bigger and a little more evolved, uh, and that we're gonna continue to possess it.

[00:23:22] We have a monthly meeting for the, uh, builders of the canon that where we sit down and we're gonna argue over premises that we think should become default part of the material. Um, and then our job is, and my expectation going forward is that you begin to ask us why we aren't doing the thing we said we're going to do.

[00:23:41] Uh, and that we provide better oversight to the process of deploying it on our behalf. And, um, some of you won't like it. You won't like the structure. You don't believe in the methodology, and that's awesome. I, I, I love the idea that we could contend with ideas, um, and that we could come up with better ways to improve our own work together because there's so many of you in here that are so, uh, incredibly talented and experienced yourselves that we would benefit a ton.

[00:24:06] So that's the unique knowledge part, right? Similar to the way twenty years of playing baseball allowed me to understand what data points to look for. Um, but the second part I wanna talk about is the actual set of well-structured data that this sits on top of. Um, and more and more, my experience of our capacity to do the things that we want to do, especially when you start talking about delivering against financial outcomes, um- E-commerce has a problem still, I think I said this last year on this stage, which is that the underlying cost structures and information systems for our industry are very, very fragmented.

[00:24:46] Uh, in the SMB world, generally speaking, there's not database as a cost expectation that is part of how you think about building and organizing your businesses. And so what you end up with are lots of different data sources, an ERP system, QuickBooks, Meta, Google Analytics. Maybe you have a Domo instance where three of those things are connected.

[00:25:09] Then you have a different financial system, and then your POs exist over here in this spreadsheet, and it's all sort of fragmented in a way that makes it very difficult to navigate with. And we've, we've dealt with this just simply trying to integrate into Statlas as, uh, an entity, where we still have lots of challenges getting to things like real-time view of cost of goods, which as an idea is generally difficult to accomplish, uh, for a myriad of reasons.

[00:25:34] But what I would say is that we are more sure than ever that the brands that are going to take advantage of the tools that exist are gonna be the ones that sit on the best structured data, is that your ability to organize this information, um, is what is going to enable these tools to work. And we wanna play a role in that.

[00:25:53] Um, and if we think about some of that information, so you obviously all the account data, the things that I just referenced, your source data, we want to continue to build more and more integrations into that capacity to organize that information on, on your behalf. The other things that we feel like we possess that are novel that we can add into that context for you to enrich that data are the D2C index, which are contextual data that lives across all of our database for categories and industries and experiences, and you'll see some of that in your H2 presentation that helps you understand how your performance sits in context to the broader market.

[00:26:29] Conversation signal, um, Zoom call transcripts, Slack, uh, interactions, email threads, strategic notes from our people. As of two weeks ago, now every conversation we have is part of your MCP instance in your database. So sitting alongside your Meta performance is your call where you said, "I want Meta to do X, Y, and Z.

[00:26:51] We wanna test A, B, and C," so that that information can exist again perpetually in a way that historically degrades immensely as we move from every end node of interaction in the system. If I go from... Again, I'm just picking on Nick 'cause, uh, he's standing over there, I can see him. If Nick has a call, and then he goes to his manager and explains what happens on that call, and then that manager goes to his director and explains what happens on that call, and then they come to me and explain what happens on the call.

[00:27:20] Then on the side, the client Slacks me I have two completely different stories. I literally am like, "I don't know what is going on." This perception and this perception are wildly different. And the beauty of it now is that I can take them and I can just cross-reference against the initial interaction. And I can go, "Okay, I understand w- how we got here."

[00:27:38] But that game of telephone is a degradation of information that we can help to avoid. Um, test results. So, uh, one of the things that we obviously have done, I think we're about almost 200 incrementality tests in now after about a year running it. So we have an aggregate growing database of information, not just your individual test results, so every time you've run a test, that becomes part of your historical context, but also every test that exists in every channel as we go.

[00:28:01] That helps us to identify new opportunities for you to be able to say, "Hey, we're seeing app web and represent a really high incrementality factor. That could be a next channel to test," or, "Hey, this channel is not that case. We can avoid it." And then the growing model outputs every time we update your new customer acquisition model, your spending power model, your retention model.

[00:28:18] Every instance is a learning that continues to grow. And so that structure on the database side is something we're spending a lot of time thinking about. Um, and it comes to life in the form of the MCP. So we, the Model Context Protocol, this is a connection between the Statlas database and whatever AI instance you're using.

[00:28:35] You have heard us talk about this. Um, and what it enables is not just our team, so our- the MCP is connected to all of our individual employees, uh, but also, um, you have the capacity to connect yourselves directly into it. And one of our core focuses for Qua and team is to every month ask ourselves, "How could we enrich the underlying data that sits in there?"

[00:28:54] So this quarter, as an example, we've integrated with Omnisend and Northbeam. Where's the Omnisend team? I think they're in here. There. What's up, guys? So we now have direct integrations for Omnisend, for Northbeam we added in. We have, uh, product-level data. So one of the biggest actual structural database issues, like, if you wanna grow your database by ten million rows, the way to do it is to pull your history of every individual SKU sale over time for the history of your brand.

[00:29:17] It takes a lot of effort, and so that's something that this quarter we've now enabled. And not only can we see all of the SKU-level sales and make inference about it, but we can also connect it back to ad spend, so we can begin to identify the relationship between where we're spending money and what units we're selling.

[00:29:33] We're in the process right now of integrating into systems for real-time inventory view and TikTok shops on the cost side in particular. TikTok shops revenue a lot of times comes in through the Shopify channel side. But you get the idea that our capacity by default of what we're gonna be able to have view of and then analyze and present information back, uh, is growing constantly, and that's an intention.

[00:29:53] Um, and then the other thing we're committed to is that we want to understand what information you uniquely want in there. So as an example, the Normal brand team who we're working with on the MCP side has their own retail stores, okay, that are run on Shopify POS. So we're working with them right now on a direct integration to ensure that we have all of their retail store sales data included into their MCP.

[00:30:16] So if they wanna analyze meta performance, we can ask a question about how meta impact relates in geographic regions, where there's store performance, and how store performance relates to spend in those areas. Inventory is another big one. But we wanna know, what are the things that matter most to you? Is it customer service tickets?

[00:30:33] Is it subscription churn? Is it retail sell-through? Whatever's currently stuck in spreadsheets or other systems, we would want to help you structure, and we're gonna deploy sort of the, the FDE model, forward deployed engineers. Joe, who's not here, he just left today, is leading that team on our side to come and sit in your organization and go, "What information can we help contextualize on your behalf?"

[00:30:52] Um, and the good news is, is that- After trying to think about this as a presentation of is this a product we want to sell, and part of that was gatekeeping the path to get there, we made a decision that we're just gonna give it to you. Um, so as of today, all of you that are here will have access to the Statlas MCP, and if you paid for it, you're getting your money back.

[00:31:11] So don't worry about that. But we just- there's too much value, and my perception is that organizationally, it's not intuitive yet that this is a thing you should be paying for. So even though I think you're all insane for not doing it yet, I don't care because I don't think we understand yet culturally where we're all headed, and so I want to get this in your hands.

[00:31:30] I want you to have access to it and use it and play with it and see what you're querying and wonder, "Hey, what questions are you asking? What information can I gather and bring you? What other stuff can we put together on your behalf? How can we organize your database structure for you?" So after this week, your team will be reaching out to help that get...

[00:31:47] Whatever, whether you're on Codex, whether you're on, uh, Claude, whatever instance you're using, we can simply build a very simple connector that will give you access to all of the database of your information for free. So that is good news because I just... We can't live in this world where we're dealing with different disparate information sets.

[00:32:05] The other thing that's gonna help is that I watch in Slack all the time, um, this experience happen where you guys will ask us at two PM on a Tuesday, "How are things going today?" Okay? And I just want you to know it's a terrible question to ask a person. It's a terrible question to ask a person. And I get it.

[00:32:24] That's what we all did. That's what we were here for. But, but here's what I promise. You don't actually want it to be true that your growth strategy is sitting in Slack going, "I can't wait till they ask," and then, like, fires off and runs off to do that. They're in a meeting. They're doing something else. So the idea that it would be an hour before they respond to a question where you wanna know what's happening right now would actually be, like, best in class service.

[00:32:46] But that's a horrible response time to that question. So instead, what I wanna do is I wanna give you the resources to get you instantaneous access to all of that information right now. And whether that happens through Slack, and we build sort of an iterative version of Scout or whatever it might be, or you just do it in your own instance, I don't care.

[00:33:01] I want you to have that because that's not the work we're really here to do for you. Like, that's not the most valuable thing we can provide you, is an answer to what's happening right now on a Tuesday, but it's really important for you, and it's important that we know in context of the actions we're gonna take.

[00:33:16] For sure, we need it if it's a big sale moment. That's very different. But I'm just saying, that information is stuff that you guys should have available to you any time you want it. Any question that you have bouncing around in your head about, again, like I said, is, does meta affect store sales? Like, the idea that you have to go, "Hey, you know, Nick, could you run this analysis for me?"

[00:33:34] Like, he's literally gonna copy and paste your statement into Claude himself and then just send you back what he just got. Like, that's just the truth of it. So I don't wanna exist in a charade where our perceived value to you is a middleman instance between a chatbot It's like a bad version of my future is to act like that's value.

[00:33:54] At some point, you will all go, "Nah, it's not a very good thing to pay somebody for." And so what I wanna do is I wanna be off building and coming up with net new ideas and making sure things are deployed correctly and actioning on your behalf, not querying on your behalf. And so I think this is really important and we wanna really interact with you.

[00:34:12] So I would challenge you to, to s- create space to try it, to interact with it, to make use of it, to point out the issues. There's gonna be errors. Point out where the information's wrong. How can we get it right? What can we do better on your behalf? Um, and the hope is that if we all do this, this is the opportunity and why we're in this together a bit, is that if everybody participates, the system gets smarter every time.

[00:34:35] Every time one of you goes, "What if we added X?" Then we can deploy it across every end node of the system. Every time somebody goes, "Ooh, what if we did an analysis of A, B, and C?" We take that analysis, and we can deploy it across the end node of the system. And so the reason an agency will always outperform from a knowledge basis an internal team is because of that compounding effect, is because every day those questions and answers happen across all of our two hundred plus customers, and as they gather, then they get deployed broadly, and you can contribute to that in your own way.

[00:35:08] So the MCP is a great way. What's beautiful about it is that, again, every time you ask that question, the more we can understand the questions that are being asked, the more it develops our own capacity to service you. So if you think about the same way you guys interact with customer service tickets, right?

[00:35:22] You could run a query and say, "What are the top ten questions asked in our customer service database?" And then you go and proactively bring forward information on the PDP to answer those questions ahead of time. The same opportunity exists here. And so whether that's the real-world results of an incrementality test, of incremental attribution optimization, or it's a constant querying of, "Hey, what is the difference between how Shopify defines discounts versus how we do?"

[00:35:47] Like, whatever the constant ask is, the more of that we surface, the more better, the more effective we'll all become. Um, and in the middle of this, again, is still this person, this engineer that is going to provide judgment and thoughtfulness on top of what the system remembers. And so when we talk about the idea of one operator, one accountable outcome, what we want you to understand is that you do have a person.

[00:36:11] You have a Nick. You have somebody who's there to be your advocate, to be available, to respond and be responsible on your behalf. But they aren't themselves on an island. They are surrounded by a lot of these people in support of the efforts that they're doing and the system that they're after. But they are here to take those skills then and deploy on your behalf.

[00:36:28] And so what you're gonna experience here, um, both in the form of the presentations later today from our other members, to specifically tomorrow in your one-on-ones, is you're going to see that system at work for you in what we're calling the H2 plan. Um, the H2 plan is a skill that Ally... Where's Ally? Are you in here?

[00:36:46] She's here somewhere. There she is. Oh, pointing back there. Ally led organizing the information of the canon into a skill that could be deployed across every brand that was attending to put together the H2 plan. So it's a combination of the canon, the skills, and your data into a plan for every one of you of how we think the second half can come to life.

[00:37:05] And specifically, it involves the forecasting methodology, the creative strategy, your specific data across a number of, uh, individual instances, the skill that I described, uh, into a deck that is automated and built, again, automated as an instance, review, analyze, contextualize, improve, iterate, present to all of you.

[00:37:25] Um, and it's going to be these specific chapters. Your, uh, H2 readiness score, so how prepared we think you are to accomplish the things you intend to accomplish in the second half. A 17-month trajectory, your pacing versus the data set. Uh, model grounded and daily refresh. How are you performing against the expectation in your models?

[00:37:45] How are you performing against the broader data set? What we call the nine peaks of H2, looking at the second half of the moments that are critical to take advantage of. Your specific forecast, your financial expectation for the second half. The ad math, one of the big things that's part of the canon is about the, a number of ads that need to be produced relative to the goals that you have.

[00:38:05] Um, this is probably one of the more contentious ideas that we run into all the time. Um, I would love to wrestle with anybody about this. Um, I, uh, there's nothing I wish was less true than the reality of creative performance. Um, I want you to know that I don't have, uh, I, I don't have some dog in the fight around the idea of creative volume.

[00:38:25] I really don't. If I could make three ads and have them work, I would love to do so. Um, I just have watched this play out in the most specific, detailed, uh, data-driven way that I've ever seen, is that the reality of the performance across creative sets are just highly unpredictable and highly concentrated into a small subset of outcomes.

[00:38:43] Um, and we wanna live in that reality and present it with you, and then partner with you to decide, okay, in light of that reality, what's possible? What do we have the production capacity for? How do we do our best to improve the expectation of outcomes as we go? But we want to be clear about that. We're gonna combine that with what we call the surround sound grid.

[00:38:58] This is a way to think about what those ads should look, sound, feel, and say. Uh, we're gonna talk about your marketing calendar mix, account structure, demand capture, your measurement roadmap. Right now, and this is o- our responsibility, we are testing way too little. Every one of you should have an expectation that there are one to two tests, geo holdout, incrementality studies running every single month.

[00:39:20] That is my goal, is to ramp up substantially the amount of testing that we're doing, not just at the channel level, at the tactic level, in as many different ways as possible. And so we're gonna present a measurement roadmap to you of where we think we should go that would provide you the best context to understanding.

[00:39:35] And then very specifically, six things we're committed to doing by July 1st. So you're gonna walk out of here with a big picture view of how I'm gonna go after all of half two, and then you're gonna get six things that in the next twenty days we're committed to doing on your behalf. So we wanna go big picture, and then we wanna get as specific as we can, and then the scope of how, what we think it will take to get there Um, and every one of you will get that.

[00:39:58] Uh, and again, it's this output of a methodology that we created a skill, that we ran across all of your brands that contextualize everything that existed and deployed this on your behalf. And the hope with this information is that we can get to faster diagnosis, that as issues arise, we can begin to build alert systems and recognition where that information...

[00:40:16] Uh, you guys, anyone been reading about loops lately? Okay. So loops are this idea that rather than prompting, that we're gonna build systems that prompt themselves. So, so imagine like this idea of what a profit engineer does every day at CTC. The idea is you wake up, and you ask yourself the question: how did performance land yesterday to expectation?

[00:40:37] What am I going to do about that? Uh, or what does that mean, and what am I gonna do about that? Okay, well, that is functionally a loop, right? It's this idea that there's a prompt predefined and a systemic set of behaviors that should happen on an ongoing basis. And so this idea that we could become faster at diagnosing this information means that I don't have to depend on someone's actual biological clock of when they wake up to ask the question, that we can progressively build these things and answers faster and faster.

[00:41:01] Um, second is recommendations you can audit. Um, I want to be accountable to doing what we said we're going to do. Um, it matters to me, and I know that we have room to improve. But you can just, "Hey, Taylor, on page twenty-seven, you said that you built structures like this. This is what my audit count looks like.

[00:41:18] What's up?" And we'll talk through it, and if we need to fix it, we will. But I wanna be accountable to the recommendations that we made. Um, defensible bets, um, that we can look out into the future and feel like we are giving you a view of reality that we can stand behind with good information and good thought process.

[00:41:36] And then ultimately, we want you to be able to ask the questions yourselves. Um, there's this other thing that happens a lot at CTC, which is that we end up in a dialogue which is really two instances of Claude talking to each other with our clients. So it's like you guys will do an analysis, you'll send it to us, and then we put it in our Claude, and it does an analysis, and then you send it back, and we just kinda go in this loop back and forth.

[00:41:58] And what I wanna recognize is that that is a reality of the future, is that you have access to the information too, and our job, again, is not to gatekeep access to things that we have, but it's to contend with the idea that we want to provide you great information. We want to be clear and accountable to what we say we're going to do so that we are all processing the information in a similar direction.

[00:42:21] The hardest thing we deal with is that somebody came to CTC, heard our marketing, got the sales pitch, signed up, and then on the first sales call is like, "Yeah, I don't really wanna talk about contribution margin." It's like, what, why, what are we doing here? Why are you here? What, what, what did you come for?

[00:42:38] Um, and so the idea that we would have this shared idea to say that if you chose us, you chose this, and so in light of that, we should contend with it, and we should decide what it is. Uh, and then we should give you access to the information 'cause we know you're not trying to gotcha us into some alternative view of definition of success, that we've agreed on it, we're processing in the same information, and we're all pulling on the same end of the rope.

[00:43:00] Um, and if we do that, then I think we get to a place where knowledge plus data plus AI, it just used to be the math of institutions only, but today it's the math of my all-star team, and it can be the math of your business. Um, and the canon on your seat, it's 12 years of CTC written down once. The MCP is yours now.

[00:43:15] It's free going forward for every brand in this room. Um, and in your breakout, you're gonna get the H2 deck that your engineer built, uh, defended against this canon. So please, as you get time today, whether it's on the boat tonight, uh, with a beer sitting on the deck, it'd be a nice time to read the canon or later at lunch or sometime.

[00:43:32] Please, I know it's a lot of pages, but contend with it. Tell me why I'm wrong. Bring it up. Help us to improve it. Um, it matters to us, and we wanna do a good job, and we care about doing the best that we can on your behalf and continue to improve as an organization and help you get where you wanna go. Uh, and then tomorrow in your meetings, pressure test it.

[00:43:48] Ask questions. Understand where it came from. What is the source? Why are we doing this? And let's build the best plan that we can to have the best second half that we can. Um, and we're really excited to work with all of you on that goal. So appreciate you all being here. Thanks for continuing to be awesome partners and excited to keep this thing going.