If you want people to make data driven decisions, they need to trust the data. And I don't mean this from, like, the the robots hallucinate. I mean more like, they need to see consistency in how this is being reported, and they need to feel this level of, like, trust that the recommendations. Right? If it gives you recommendations that are generic because it doesn't understand the industry, it doesn't have the context, this is how people go back to, like, I'm just gonna win this. We're not answering questions. We're acting on this data. Okay. Fast forward to twenty thirty. What does crowd data look like? I don't know how you even, like, run a business by twenty thirty. How you build software is different. I think that most businesses by twenty thirty are Welcome to AI twenty thirty. Keith Jensen, president at CADRE AI. We have Alex Lurzman here, founder and CEO of Corral Data. Welcome. Thank you. Thanks for having me, Keith. We've known each other a while, so this should be a fun one. Yes. Minus the cameras being in the room. Minus that. Minus that. Okay. I think one of the things that's most interesting with you is you had Ready, Set, Rocket. Right? Agency, New York, crazy clients. We also had an LA office. LA office. Okay. Crazy clients. NBA. Right? NBA, Michael Kors, JP Morgan, like, Sweetgreen, Strava. Like, in terms of team, it was a small team. Right? Yeah. It was a small team with Mighty. Yeah. Mighty. We just worked a lot. It's unbelievable. So and you and we went to UCF together. We did. So you went up to New York. How did this even how did you get to a point where you had this level of clients? I don't think I was qualified for a corporate job. That's probably it. So my first job out of college, like, graduate, you go to you go to a real job, and you sit in the cubicle, and you're like, yeah. I don't wanna do this. So I I don't think it was just I I think it was self selecting. Right? Like, just like if you have if you have ADHD, you probably cannot make it in, like, a super corporate world. So my first job was at Prudential Securities. My second job was at JPMorgan Chase, and they paid for me to go get my MBA. And then during that process, I realized I probably can't do this long term. I can do this short term, a year or two, but Did you guys muscle I mean, think of how many small agencies exist where they're never touching that level of client. Like, there's gotta be is there anything you could like, anything you guys did unique that you would give, know, advice you would give to any other any company starting in this? I mean, talent. Right? Like, I think that's really it. They're just, like, talent and drive. Like and I know everyone talks about culture, that's sort of, like, a weird word because you can't, like almost, like, can't fake it, but treat your employees with a ton of respect. Make sure that they're well compensated. Make sure that you're able to actually, like, deliver really good product and obsess over the delivery of the product at the end of the day. Right? Like and if you can sort of, like I don't know. If you can create a culture of obsessing over delivery. Right? How do you provide the most, like, outstanding, not just customer experience, but also, like, outstanding results? Right? And I think that's how we did it. I think we also just in the beginning, we took on work that like, we took on massive brands for very like, small parts of really massive brands so we can look much bigger than we are. And I think everyone's sorta, like, fake it until you make it, but, like, the work we took on was literally it was like, somebody would pay us twenty grand, and we would spend thirty grand on freelancers. Like, we were not making any money. We would barely break even, but we were able to grow really quickly, lose a lot of money, almost like a VC model. We were we were growing really quickly. We were taking on the really tiny parts of massive brands so we could get our logo boards up. And then eventually, when you get in, you it's land and expand because at that time, I think we also really got I think we also got really lucky. So when the agency was started at sort of this time where digital was just like this is before, like like, Facebook didn't really like, we're just starting to come into the scene. We made the first Facebook app for it was for emergency, the vitamin drink mix for another agency that we partnered with. And they're just, like, these kind of, like, really first cultural moments at that time were also something that, like, you can share with, like, corporate customers. And they're like, look at all the cool stuff that they're doing on Facebook. This new young this new young thing, they understand social media. So I think, like, at the end of the day, just, like, getting in no matter how small the opportunity is at places where you could scale up pretty aggressively and then over delivering. Right? Like, somebody is saying, we need you to get me here. You overshoot that. Somebody says, can you deliver this by Friday? You deliver it by Thursday. Right? So I think our sort of, like, approach was always that. We started to do a really good job after a while of just not taking on even if the opportunity was big, not taking on smaller brands with big opportunities, we would take on big brands with small opportunities. Get that logo count up. Get the logo count up. Get that brand board up, and and it feeds everything else, and then it's land and expand. Right? Okay. So then you've worked with marketers were your stakeholder. Right? Because you were working with the marketing teams. Now you leave you exit Race at Rocket, and you go found Corral Data, data analytics, marketing analytics platform, data analytics platform mainly for marketers, but as you know, I've used it for even plenty of other departments. It's could be used across the board. I'd say, like, is it a competitor to, like, a Tableau and all that? Personally, I think it's better, but like it I appreciate that. Yeah. There's you you are up against businesses using Power BI, Tableau, these these massive, you know, platforms in their business, and you're trying to go, hey, marketers, you shouldn't be using that. You should be using ours. Now I guess my question of this is, what have you seen? What has been, like, the adoption issues you've seen with marketers? Because I I can't believe a marketer is gonna look at Corral data and not go, **** ****, I need that level of granularity, and it looks beautiful, and I can ask my ask questions of my data, build my own graphs. Like, it's cool. It's really cool what it can do. What's the sticking point? Like, what are you seeing marketers like, why are they resisting, if any? Like, look, I don't think we have any issues converting marketers. I think marketer so, like, the three areas where we convert really well, marketing, operations, and finance. This is like a slam dunk sale. Right? Like, this is not I've never like, I don't remember the last time we were blocked, like, marketing. It's like, I'm gonna stick to, like, building reports on Looker or Power BI or Tableau. Like, that just, like maybe it happens, and I just haven't like, that's not a that's not a challenge. I think sort of, like, the fundamental challenge is, like, the change management when you're like, hey. We've invested all of this time in our Power BI, and our IT team has built out sort of, like, the infrastructure for it, and we have the headcount and like, the world is changing. Right? Like, so when you're saying it was like, hey. We could literally instead of you having to manage a data warehouse, pipe your data or ETL your data over into a data warehouse, then build widgets and reports one by one, and we could just you could just connect your data, and then it'll give you all the reports, and then you could ask questions and then build anything you want, and then you can say, how should I run my marketing? And they will tell you. Like, I think that's could sound really cool to a marketer, and I think to, like, somebody that's like, oh, what used to take me months is now, like, minutes. That is, like, sort of, like, a terrifying, like, proposition. Right? Like, in this day and age when, like, there is a job security issue. That's right. And the marketer is not gonna have the the authority to be like, well, let's go fire fifteen people on the, you know, analytics team. Yeah. And that's not and I think that's not our goal. I don't I actually sort of, like, don't think and so we have since we started, I think we're at a hundred and twenty brands now. And I haven't seen sort of, like, maybe there's some scenarios that I just don't know them, but I I don't know. Like, someone's like, hey. We went and we fired a bunch of people. I think what ends up happening right now is just like, oh, Corral Data is doing this. Now those people aren't answering marketers' emails of, like, how did our campaign perform? There or operations, which is, how do we sort of optimize our scheduling? Or I think what they're mostly focused on right now they don't have to answer these, like, questions that they were annoyed with in the first place. Now they can focus on, like, deeper level analysis, and now they're using Corral. So I think, like, the big shift has been when we started out, the platform was really, really built for, like, marketers or sort of, like, business users that needed access to their analytics. And then we added operations people really started using this, and I would actually say we probably have as more operations people than marketers on the platform, and then we started getting a lot more finance people on the platform. And recently, I would say within the past I don't know. Like, since Claude became, like, sort of, since everyone migrated from Chad GPT to Claude, like, we've also seen this, like, massive, like, analytic, like, analysts, the people that were essentially, like not the same people, but, like, sort of, like, that cohort of people that would have blocked something like this are are probably biggest advocates and customers because they are the ones that are now using Corral to get everybody set up with this is a govern because Corral data now runs the governance layer. We do sort of, like, all of the security around the how people interact with their large language model of choice. Usually, it's Claude. Right? So how do they actually interact with this inside of Claude? How is that data governed? How do we make sure that the person that is asking a question in Claude about their data only sees the response that is, one, is relevant to the business and has the semantic layer in it to, like, really understand that nuance, and then the second piece is they they have governed data, so they only see what they're allowed to see because they are like, everybody's asking them in their organization. It's like, we want a cloud license, but the and we wanna we wanna do this. They're, like, almost, like, afraid to give it, and now we're giving them. It's like, oh, hey. We got you. And is that role based permissions, right, does that get inherited through the platform that the data came from? Yes. Exactly. So they set all the role level. They set up all of their reporting or or they do all their connectors inside of Corral. They set up all the permissioning inside of Corral. They can use Corral, like, with any of the three large language models main large language models. They can have all their reporting. But, also, if they wanna use Claude, they can then use Claude or or Chad GPT to also ask any questions, and that inherits everything that they've set up inside of Corral using the Corral data MCP, and it makes sure that the data that is passed like, we work with a lot of health care brands, for example. All of it is also, like, from my PHI perspective, we're also not passing any PHI back in there. I'm interest I I I'm scared to go down some rabbit holes with AI, but I think it's really interesting of, like, the okay. So you've got the data stored at the data warehouse level. Is it vectorized? Yeah. Okay. So you you now have and there's probably a Rag architecture. Okay. So how are you doing contextual memory? Meaning, like, if somebody has Claude, right, and they are connected via MCP to Corral Data's MCP server. Yeah. Infrastructure. So now they've got you've got all the data on Google Ads, finance data, all of that. Each of those each piece of that data came from a different platform, which has some level of role based permission. This person has access to this Google Ads account, etcetera. That gets inherited Yep. Down at the okay. Well, no. Actually, so usually an admin would set it up so we get everything. Right? And then they will sort of say, it's like, okay. This person's allowed access to this entire sort of, like, connector. So QuickBooks, like, usually people don't have it. And then you could go a step further and say, like, this person has access to this entire, like, EMR or EHR system even, like, you wanna get really complicated, but they're not allowed to see any locations that are not tagged to them inside of Corral. So you could go as fine grained in terms of your role based security as you want, and then when you're asking this question whether it's in Corral or in Claw or Chai GPT, it inherits all of those permissions down funnel to be able to then answer answer those questions. And so you're able to essentially be the data lake Yeah. Providing the the the context, and there's and they could create skills, let's say, in Claude Yep. That would then run against the data that's coming from the data lake. Yeah. And yeah. And in this case, it's a data warehouse because the data is already it's, like, nicely stored for you. Right? The pieces that we're also adding to this is this entire semantic layer. Right? So, in every single industry, there's sort of, like, their own definition of how things are defined, and in every company, you define things differently. Those definitions also come through. So the the way that we do that is it's not that it's just hitting a data warehouse. It's hitting it's hitting Corral Data, which has already normalized all of this data, transformed it, made it accessible to Claude or Chad GPT. And what we're also doing in that instance, we are also saying these are all the definitions of these things. So when you say it's like, how has my rebooking how has my rebooking rate changed over time? We know your definition of rebooking rate when you say LTV. What is my lifetime value of a customer? Every single customer we work with has a different definition of lifetime value, and if you go into Claude and you just give it, like, a random dump of data and say, what is my LTV? It doesn't know how your company thinks about LTV or how your company thinks about CAC. So you're storing it like a markdown file or something that's it's that it's referencing that doesn't see are referencing because we've built all of the reporting within Corral that gets automatically built, and we have all of the the entire semantic layer that is built across everything that a customer just pulled in. That's the second piece of it. And then third piece is because we work with in some of the industries we work with, there's a commonality of how things are defined. So there's this level of alignment that happens already. Speaking of the industries you work with, you've gotten into health care, and specifically, multi location health care, which is very fond, to me because that's that's a world that I come from. I have found incredible results when being able to do something that I was never able to do before. Actually, I was doing it via Excel sheets and a lot of, you know, manual scripts, but being able to actually connect, let's say, a spend of a Google ad to actual revenue that comes in. And for anybody who's not in health care, they're like, yeah. Of course, you can do that. Right? Because it's going through Salesforce or this and that. But when you're talking about these legacy EHRs, right, and these these platforms that health care companies are using, it's incredibly difficult to get data out of there. And very it's very unlikely that they are, you know, pulling in Google Click IDs natively or anything like that. But we've been using Cryo. I've been able to do it multiple times, and it is transformative to the business. Are you finding marketers in multi location health care when you actually show them that this is possible? Is is a light bulb going off, or are they skeptical because of, like, compliance issues? No. I think that any marketer that sees it, like, they this is one of those things that, like, gets integrated, like, next day. There's an email that goes out or a text message that goes out immediately to someone who's like, hey. Can we can we get this set up? Because on our end, all we're really doing is we're matching the Google Click ID with what if sometimes we have to help a marketer out and say, it's like, hey. Make sure in your Google Tag Manager or whatever, like, you're passing the Google Click ID to a form field. But then all we're really doing is we're seeing in the form what was the phone number or the email address and what is that phone number and email address in the EMR system. We're matching it together, and then we're able to say, based on this based on this campaign or this keyword, what did this person actually end up showing up, and then what is their lifetime value? And, obviously, I'm simplifying this a little bit, but that is essentially that is the heart of it. The challenging part of this is being able to do this in a HIPAA compliant way without downloading Sheets and not being in breach of HIPAA is really hard. The most interesting part about this entire piece there's two interesting pieces. One is our more advanced marketers, what they're also doing is we're not just showing them what happened. They can send this data back to the ad platform. So instead of optimizing against optimizing against a book, the somebody books an appointment online on some form, they actually showed up and this is how much they spent, and now we're gonna say that this Google click ID through an offline conversion event actually generated this amount of money in lifetime value over usually, we're looking at ninety days. Right? Like, so what did they spend in in the last ninety days? Right? And that's I think that's also, like, the the big unlock more than just looking at the data. That's massive because any marketer that's thinking about this, right, they're thinking about how am I gonna attribute back so like you said, click happens, form fill happens. That's the immediate trigger to say to Google, yeah. This campaign did a good job. Maybe you should spend more money on this campaign. Maybe especially if it's doing, like, automated bidding or Pmax or anything like that. Right? It's just gonna start spending more money. What if those leads suck? I literally had a conversation today with, with a customer of ours, and, they have a new CMO, And we're talking about sort of like, hey. Our upper funnel, like, traffic, our upper funnel leads have really fallen, and she was, like, very concerned about what's going on, and she was just asking me for advice. She was like, hey. Alex, can we hop on a call? Can we can you help me sort of understand what's been happening? I'm trying to understand sort of, like, the the landscape of what's been going on. And, like, I mean, I I don't I don't know what's going on, so I'm just like, what happened to their leads? And so I'm asking Corral, what happened to their leads? And the question was like, they turned off Pmax. I was about to hop on the call, and I was just like, you turned off Pmax. What are the leads? It's like, are you and then I'm like, what if she wouldn't have turned off Pmax? And so it's like, she would have had more leads, but those leads would and so it went a step further. It's listening. But those leads would have not generated additional revenue for them because the way the because it is optimized towards it's just optimized towards towards a lead. It's not optimized towards somebody showing up. So now Facebook and Google well, Google Ads in this case has an incentive not to drive revenue. It has an incentive to drive leads no matter what. Right. Not the form fill. It doesn't care about the form fill. It's gonna care about, is this the type of person that is actually gonna show up and pay? It's the same reason that back in the day, like, you like, I think the statistic was, like, over eighty percent of all clicks happen by the same people. Right? Like so, like, yes, it's really easy to drive up your click through rate. Show your show your ads to people that are clickers. Yeah. What do you think is the thing that marketers are doing wrong most like, most commonly? In terms of, like, you get on calls with prospects and you're like, hold my beer. Let me show you what's going on here. Like, what's the thing that you're seeing them where you're like, ah, if I show them what we can do here, something's gonna go off. Is it the attribution mostly? So when you first asked, what are what are marketers doing wrong? And, somebody that was a marketer is just going into marketing in the first place. I don't know. I think sort of technology has changed these things so much. I think if you're if you don't have I think today, at least, if you don't have if you haven't figured out how to automate this entire process as a marketer so you could really scale this, that's, like, probably the first part. Like, forget about, like yes. We could talk about attribution, but I could build a me like, our customers build media mix models themselves inside of Corral just like, build me a media mix model based on my current performance and spend. Even if attribution is, yes, we wanna have perfect attribution. Sometimes that's not possible. Let's say somebody didn't tag something perfectly before. There's never, like, sort of, like, an excuse anymore of just, like, I don't have the data or I don't have the skills or I don't have the experience. I think it's more like, I don't have the intellectual curiosity to figure it out or to, like like, spend the time to solve problems that, like, I I keep on seeing. Yeah. Okay. So you've got I mean, first off, let me say the ability to do media mix models inside of a platform like this is pretty wild because I remember us talking two years ago, and I was at a company that was trying to figure out how we were gonna do our media mix models. And I remember you being like, yeah, that's not something we're gonna **** with at this point. And now, it's like, yeah, we could just ask to get it I mean, so it all comes at do because we don't have to build it. The robots build it for us. So, like, we didn't add any new functionality outside of, like, the ability for for the ability for our platform to have data that is essentially to sort of, like, have our agents understand this data at, like, every single possible level through multiple lenses and to give it the data that is absolutely sort of, like, easy to work with. Yeah. And those that semantic layer, right, that that helping it understand, like, what LTV means for this client, that's gotta be probably the most important part, I would think, to, like, it being able to actually go deeper to understand here we have all the data. We understand what LTV is and all of that. Now if you're trying to to determine, well, did this person did this action happen because of other actions that also happened? A click here, a website visit there, how you know, that's I mean, ultimately, the media mix model, you're trying to figure out if I turn this down, is something else gonna Yeah. Well, like, sort of like, what is that incrementality that is like from that additional spend? Right? And I think on media mix models, it's incredibly important. I think it's just it's important for answering any basic question too. Right? Just like like something as simple back to, like, that question. It's like, what is what is my rebooking rate? You ask, what is my rebooking rate to Claude or Chad GPT without, like, context? Without it's this is sort of like it will give you a different answer every single time, by the way. So I think, like, this level of just, like, what is your definition of rebooking rate? Like, how should we be looking at this? And then how should I manage my business around this? And then I tell, like, the other piece of just, like, if you want people to make data driven decisions, they need to trust the data. And I don't mean this from, like, the the robots hallucinate. I mean more, like, they need to see consistency in how this is being reported, and they need to feel this level of, like, trust that the recommendations right? So it's one thing to get, like, answers. I think they also need to trust the recommendations, which is sort of this next level of, hey. Look. My lead I'm responsible for I'm a new CMO at this organization. I'm responsible for upper funnel. Help me understand how I should reverse the trends that I'm having. And if it gives you really good, solid recommendations and you start acting acting on those recommendations, like, you will have trust. If it gives you recommendations that are generic because it doesn't understand the industry, it doesn't have the context, this is how people go back to, like, I'm just gonna wing this. Yeah. Okay. Fast forward to twenty thirty. Right? What does Crowddata look like? And what are how are marketers using it? The way that we used to do and we're look. We're a software company, sort of. We we talk we're talking about marketing and operations, but data, but, like, we're a software company. I would say about four months ago, the way that we developed software at CorelData was somebody a product person had a massive backlog of tickets, and everyone was throwing stuff at them. And they were just, like, grooming. And then they would jump on some customer calls to understand the vision of where this is going, and then I would throw some wrench into something. It's like, we need to build this. So I would say within the last four months, and we're not the only ones, we've moved to sort of, like, this model of anybody can put a ticket in. There's no more friction. There's no more, like, barriers to putting a ticket in so something actually gets built, a new new feature, an improvement, whatever it is. Right? Nowadays, when you put a feature a request in, it gets interrogated by our AI. Our AI interrogates sort of like that request. And then once that request is actually interrogated and vetted and made sure that this sort of aligns with the principles and the focus of the company and everything else, it'll actually go out, build out fully function like, full acceptance criteria, testing everything else, and the testing isn't like binary testing of just like, hey. Will this work, not work? It's just like, what is the goal? What are we trying to do? How will this solve these problems, all of this? Right? And you don't want binary, and we could talk about that stuff, but you don't want, like, really binary yes or no tests. But it's able to essentially now build out these features, and our engineering team is now able to either accept this pull request from an AI, or it's able to say, it's like, no. That's that's that doesn't make any sense. Right? So I think it's really hard to answer this kind of question because in four months, we've come from sort of, like, a stand point of, like, how you build software is different. I don't know how you even, like how you run a business by twenty thirty. I I I don't wanna be, like, gloomy about this kind of stuff, but, like, I think that most businesses by twenty thirty are much more focused on how do we provide like, their moat is essentially their data. How do they understand whatever nuances substantially better than anybody else in whatever industry or whatever person that they're serving because the software moat completely goes away? So all you have is you have, like, a you have proprietary data, and you have service that is substantially better than substantially better than somebody else's service. Right? And I think, like, by my assumption is and this is this feels like a lifetime from now. Right? My assumption is that by that point, we have entered a lot more of these sort of industries where we have this massive data moat where we could provide not just sort of, like, we're not answering questions. We're acting on this data. And an example of this is right now, we don't run, nobody runs our, ad campaigns for us. Nobody runs a lot like, all of this is automated already because inside of Corral, we have our the Google Ads MCP and the Meta Ads MCP, so and the LinkedIn Ads MCP. So our ads are being run by agents based on performance. So it's no longer just sending, a pixel. It's going into campaigns and going into Google Ads or Meta Ads and actually changing things. And I think, like, where the value of, like, what we're providing and what I think any organization should be providing is just like, you have, like, your brand, you have your experience, and you have the data. Those are the those are the three pieces, and that's what I think, like, twenty thirty really looks like for us. Awesome. Alex, thank you so much for coming by today. Thank you. Appreciate the time.