[00:00:00] Phil: is warehouse native just a feature? Or is it a fundamental rethinking of how we build MarTech? [00:00:07] Istvan: From the point of view of the user, the end user, it is not really a new category. They are unable to do the same type of analytics as they would do in the traditional third party analytics tools. But from point of like. The people or the team that builds these, like marketing, um, infrastructure, essentially data infrastructure for them is a completely new approach. [00:00:31] It's a totally, it's a completely new category as well. [00:00:33] ​ [00:01:00] In This Episode --- [00:01:00] Phil: What's up everyone? Today we have the pleasure of sitting down with Ish v Ros, founder and CEO of mitsu.io ish. V started his career as a software developer at cern, the largest particle physics lab in the world. He also led data analytics and engineering teams at Skyscanner and later Shaper 3D. And today he's building a warehouse native product. [00:01:21] And marketing analytics platform. In this episode, we explore merging web and product analytics with a zero copy architecture feature or new category. What warehouse native really means for marketers, how decoupling storage and compute lowers analytics costs, how seed based pricing works in warehouse native analytics, and what a data warehouse does that your CRM never will. [00:01:45] All that and a bunch more stuff afterwards. Super quick word from one of our awesome partners. [00:01:50] ​ [00:02:57] Phil: Istvan, thank you so much for your time today. I'm really excited to [00:03:00] chat. [00:03:01] Istvan: It was a pleasure to be here. Thank you so much. [00:03:03] Phil: Uh, where are you calling in from today? [00:03:06] Istvan: Uh, that's, uh, Budapest Hungary from Europe. [00:03:10] Phil: Very cool. Yeah. Uh, most of the time a lot of folks are calling in from, uh, bay Area or like West Coast and. Uh, usually my co-host can, can join us because it's early morning for his time. Right now it's like, uh, 6:00 AM for him. He is got a young newborn, so I wasn't able to to make it for, for this episode, but I'm pumped to chat with you. [00:03:28] Uh, I first discovered you when we had, uh, the folks from castle.io on the show a couple years ago now, and Arun introduced me to, um, warehouse Native Marketing Analytics as well. [00:03:39] 1. How Warehouse Native Analytics Works --- [00:03:39] Phil: We talked a lot about warehouse native MarTech on the show. A lot of the discussions have been around CDPs and composability. [00:03:46] Uh, for the folks that are unfamiliar with the term warehouse native in the context of marketing analytics, how would you explain Warehouse native analytics to the marketing operations folks that are listening? [00:03:57] Istvan: Alright, so it's a big topic. [00:04:00] Uh, definitely, definitely I would say it's, um, compared to like traditional marketing analytics tools, the main difference with warehouse native, the warehouse approach is that the data that you're using for is owned by you in your centralized data warehouse or data lake, uh, in your insider company. [00:04:17] And, uh, basically this is like the technological difference, uh, that is, that is basically covered by the term warehouse native. It's hard to explain like in layman terms, but the biggest difference is in the technology. Yeah. [00:04:33] Phil: Gotcha. Wait, so most of the folks listening are. Um, a bit more on the, on the technical side. Like they, they support marketers, they support tail sales teams. We do have a lot of data engineers, a couple of data professionals that, that do listen. So, um, I don't, I don't think you're gonna scare too many folks off by like, uh, talking a bit more at technical terms here, but. [00:04:54] Like a lot of folks are used to Google Analytics and Amplitude and, and Mixpanel, and having all of those tools [00:05:00] be an extra tool in the stack that stores an extra set of users. You have to figure out user IDs there. You have to stitch different devices together from different browsers. You're doing this a lot differently from like a billing standpoint, but also just like building on top of the warehouse where the data team is doing a lot of that work already. [00:05:21] Right? Like unpack that a bit. [00:05:23] Istvan: exactly. So. If you think about how, like a typical company, let's focus on like B2C company, uh, does analyze data in, in, let's say this, let's say imagine like a startup, maybe 50 people. What they would do is collect all the data to these various tools they use for marketing, maybe messaging, uh, maybe analytics, maybe all kind of like deal and, uh, customer, um, um, customer success as well in the warehouse, native space, it's a bit different. [00:05:54] What you do is centralize first everything, uh, into the data warehouse. You can use [00:06:00] snowflake, Databricks, big query, all, whatever you, you can imagine all there is like a maybe 10, 15 cloud data warehouse types currently out in the market. All of them are mature, or of them are very, very easy to start. But anyway, use first, centralize your data and you do most of the thing in the data warehouse, uh, for all the things that you, uh, you can, you can do for marketing, essentially. Um, yeah, [00:06:25] Phil: Gotcha. Okay. So the first time I heard about, uh, Mitsu or how, what, what's the best way to pronounce that? Mitsu. [00:06:32] Istvan: that's Mitchell. Yes, MI. [00:06:33] Phil: Mitsu. Yeah. It, it was, um, I think like self-service bi, right? Like you guys were launching on, on product hunt, uh, made a lot of splash about, uh, the announcements and yeah, self-service BI for SaaS. [00:06:46] And maybe we can riff on this a a little bit here, folks who might be curious about. The, the differing opinions and philosophies on like, [00:06:54] 2. BI vs Analytics vs Measurement vs Attribution --- [00:06:54] Phil: what is the overlap of bi business intelligence with analytics and measurement, and there's like a different layer of attribution on top of that too. [00:07:04] Um, the way I've thought about this and, and unpacked as poke holes on, on this here, but like, this is the kind of like hierarchy that I see it, like the base layer is measurement. [00:07:13] We're collecting raw data. For analysis. And then the layer on top of that is like the processing layer. So we have analytics there. Analytics is kind of a loose way of defining it as like examining data, finding patterns and insights within the raw data that you collected. [00:07:28] And then anything on top of that is like you have your decision layer, which could be like. [00:07:33] Bi, you're putting data visualization on top of that. It's got a lot of different business reports, but then you could have like specialized applications, like attribution. You could have ML prediction. Is that kind of fair? Is that how you see it? What's the difference between BI and analytics for you? [00:07:47] Istvan: It is roughly the way I see it as well. I think you, you discovered very well, uh, in my head. It's like the capture earlier. Exactly. You, you move the data to some someplace. In our case, it's warehouse native, so you move it into the. Data warehouse, [00:08:00] analytics and bi. Definitely there is a difference in my head, BI is mostly static dashboards that somebody, a data expert created. [00:08:08] And it is like, you know, it is, it is proven to be the fact about the business. You shouldn't touch it. Nobody should touch it. Everybody's just read it. If there is a problem, contact me. I'll fix it. In the analytics space, I would, uh, separate the two, two categories. Analytics done by data specialists, uh, and analytics done by non-data specialists. [00:08:34] And the cell service bi, what we used to call ourself, uh, was essentially focusing on the analytics by non-data specialists. So you don't have to be a, a data expert, data analyst, data engineer to create, uh, insights from your data. We code it bi business intelligence because we thought it resonate more than analytics, uh, as a word. [00:08:57] Um, but to be fair, uh, since [00:09:00] then we changed our, like, value proposition, our H one on our website. It, um, we became product and marketing analytics, uh, a tool for large data sets, um, which is actually the product itself didn't change a lot, but the, this is what we've seen from market. This is the, the biggest use case for self-service analytics that people do product and marketing analytics with our tool essentially. [00:09:24] So, yeah. [00:09:26] 3. Merging Web and Product Analytics With a Zero-Copy Architecture --- [00:09:26] Phil: When you say marketing analytics, most folks will immediately think of Google Analytics, GA GA four. And I feel like that's changing a little bit because a lot of folks are trying to find ways to. Move off of D four, but when you mentioned product analytics, um, a lot of folks are familiar with the mix panels and, and the amplitudes of the world. [00:09:44] You're kinda like calling both of them together, marketing and product analytics. And I find that interesting because, uh, we, at my last, uh, couple startups and, and I've seen this with some clients too. Part of the journey of getting off of [00:10:00] GA is really asking yourself like, do we need a tool to track what our users are doing in the product and what our users are doing on the site? [00:10:08] Like these are just two different devices. Could we just have one platform that does both of those things? [00:10:15] And so I've seen a lot of companies use Mixpanel and Amplitude for both web events and product events. Is that kind of how you're seeing it? You're just. Building and, and bridging product and marketing for, for one tool to do analytics. [00:10:28] Istvan: yes, yes, exactly. Um, so I believe there isn't like certain industries or certain segment of companies that for them, essentially the two things are almost the same. Um, typically imagine like travel, e-commerce, um, advertisement businesses where you want to have, where you actually have a huge volume of visitors or, or users you, they need, they come into their application and they basically exit in, in some follow. [00:10:55] Uh, but they may coming in later again, well, half a year later. [00:11:00] But what, what's your goal as a company is to maximize the number of visitors to this, to this application you are having? Um. Let's, let's focus on like travel industry, like Imagine Kayak, you know, search Engine or Google, Google Flights. Um, the goal of these, uh, these applications is to find as many users as possible with doors that come to the application and search of flights and book through, book through Kayak, Google search, and Google Flights, uh, uh, for the application. Um, but it's of these applications, kayaks, kayak, and let's say Google search. Google File Search is, is a product. That has to work very well. The funnels has to be like maximized to, to the best conversion rates. There shouldn't be a drop off in any of these steps. If there is a drop off in like one of the steps when you buy, buy a, like book a ticket, you know, then there is a problem and you are losing customers. [00:11:50] Uh, on the other hand, the marketing side is like you maximize the number of visitors, you know, that's the marketing problem. Like to, to bring it to the, to the application. So top of the funnel basically to [00:12:00] maximize the number of visitors to, to funnel. But they also should go through the application efficiently and exit. [00:12:05] Exit positively, essentially. So in my head, this is the product and marketing problem is one thing for certain industries. [00:12:14] Phil: Right. Yeah. In the B2B world where it's like heavy sales inbound motion, then you also have a CRM at play there. But you know, still. Maybe the product is, is still being used like a lot of the times it's, it's a SaaS tool and you still want to connect and, and stitch that data together, especially from an activation standpoint like the, the marketing automation team or the lifecycle team that's trying to like kick off onboarding campaigns or do something with product data. [00:12:42] Matching that with where they came in and stuff like it. It just makes a ton of sense to have event-based analytics for web and product just being one tool [00:12:51] together. But you're, you're doing this a bit differently too, on top of the warehouse. So instead of like having a database that's copying all the [00:13:00] users that you probably already have in the CRM, you already have in the marketing automation platform, and you already have in your CDP or in your data warehouse. [00:13:08] You're basically not creating a new copy of that database. You are just living on top of the warehouse instance where you said data lakes also, and you're just querying that data. [00:13:21] Does that make sense? Do I have that right? Yeah. [00:13:23] Istvan: exactly. Exactly. So essentially what we do is automatically generating SQL operators on top of existing infrastructure. Then those SQL queries are the typical, you know, if you, if you think of like segmentation, if you are familiar with Amplitude funnels, journal calculations, AB testing, evaluation. [00:13:43] So this is what we do, but SQL queries and to be Ed. What, why we started to do this is we've seen that after a certain stage in any business, essential you reach a point when you [00:14:00] need to centralize your data anyway. You need to have your, all your data anyway in a central location. And as I mentioned, we are working mostly for large data sets, product market analytics. [00:14:11] This kind of really makes sense not to copy that large data center [00:14:15] Phil: Right. [00:14:15] Istvan: to two, three places. Uh, you mentioned Castle Dio, uh, with daring. I assume they had the same feeling and uh, uh, when they, you told, uh, that was probably the same problem. They are, they are solving, why are you moving all this data that you already have in your data warehouse to different tools, synchronizing between different tools. [00:14:33] And that's actually just, you know, like this, in my, in my world, it is kind of a waste, uh, and often actually impossible. This is what, uh, we as well see [00:14:43] Phil: Yeah. [00:14:44] Istvan: that, uh, at some point you reach a scale that your, your data becomes so large. The volume of data becomes so large, you cannot move it away. [00:14:52] Phil: Mm-hmm. Yeah. [00:14:53] 4. Feature or New Category? What Warehouse Native Really Means For Marketers --- [00:14:53] Phil: When I first heard about like warehouse native or MarTech Solutions, it was really exciting. From a pricing standpoint also, as well as like, just like connecting and integrating data. Um, and there's, there's a lot of excitements in marketing automation, customer engagement space, like we talked about castle.io. Um, but I wanted to ask you about like category versus feature. Like is warehouse native just a feature? Or is it just like a fundamental rethinking of how we build MarTech? Because we're already seeing this today, like there's a lot of CDPs who caught up to the composability movement and were just like, oh shit. [00:15:33] Instead of copying all that data, let's actually connect to warehouses. And you saw all the CDPs have like warehouse connectors. And so they're kind of like hybrid models now and you're seeing a lot of customer engagement platforms. Uh, one of the sponsors of, of the podcast is Mo Engage, and Mage has recently rolled out. [00:15:50] I host native on their customer engagement platform. And so if you are on Snowflake or you use a data warehouse, you'll have to copy all that data, figure out a way to like send it to Mage. [00:16:00] They can sit on top of your data warehouse there. So what are your thoughts there? Like is this like a brand new category, we're redefining how we're building a MarTech? [00:16:08] Or is it kind of a feature that a lot of tools in the next couple years are just gonna have warehouse connectors? [00:16:14] Istvan: Yes, it's a great question. Thank you itself. How I see this, um, from the point of view of the user, the end user, it is not really a new category. They are unable to do the same type of analytics as they would do in the traditional third party analytics tools. But from point of like. The people or the team that builds these, like marketing, um, infrastructure, essentially data infrastructure for them is a completely new approach. [00:16:43] It's a totally, it's a completely new category as well. I have to admit that it's, it's pretty hard to, uh, companies to switch to a warehouse approach. It takes a lot of time. Getting stuck in one is, is much easier, obviously. Uh, especially, you mentioned CDPs and now [00:17:00] they are out of the box. Doing the warehouse, uh, ingestion. [00:17:04] Uh, so if you have that in your organization immediately collecting a warehouse, uh, customer engagement software or like a, or like an analytics software like ours, it is essentially a couple of minutes only. Uh, so it's very easy to get started to get off of these, uh, third party tools. It's a, it's a different challenge. [00:17:22] It's, uh, yeah, it's, uh, we see struggling in there, but uh, yeah, we see also more and more coming out. Of these older, let's say, let not, I won't call them older. It's like the traditional approach. Yeah. So just to summarize your question, if you think of the, the team that deals this system for them as a new category, if you think of the end user, it is enabling them to do the same thing. [00:17:48] So for them is, is not, I wouldn't see, even say it's a feature, it's completely transparent, it should be completely transparent. [00:17:54] Phil: Yeah, so what about like the folks in the middle there? So a lot of the marketing [00:18:00] technologists aren't the end users. They're not the ones that are gonna be using the marketing analytics tools or, or sending emails in, in the customer engagement platform. And they're not the data professionals that are building that stuff in, in the data warehouse, but they're sitting between both teams and they're the ones translating requirements and the folks that see this as a brand new category, and they're really excited about it. [00:18:22] Those other folks that are really technical, but they don't necessarily always get the marketing applications and the use cases. And I feel like the marketing operations, marketing technology folks are in an interesting spot with this transition to warehouse native because we can do that translation, we can translate why the data folks are excited about this. [00:18:41] And translate that into like, why marketers should care about this. Because oftentimes the CMO is gonna have like a, a, an impact on whether we're like moving to a new tool that does warehouse native. Why should the marketers care about warehouse native? So. For the marketing ops folks that are sitting in that middle and have that opportunity to like [00:19:00] translate those things. [00:19:01] Why should we care about it? Why, like what are the biggest moments of like Aha, that you've seen folks implement Mitsu, that you're just like, this is what I need to tell marketing ops folks about the power of warehouse native. [00:19:11] Istvan: yes, yes. That's a great question. Uh, definitely there is some aha. Uh, what we, what we feel from customers, we hear from, we hear back from customers. The, actually coming back to the feature versus category, this is kind maybe a feature. What I will, just, what I will just described, so in traditional tools, you might get problems with data accuracy. [00:19:31] So meaning that you sent your data there, used for them, basically you're unable to fix it if there was an outage or something, or like you made a mistake, you tracking or capturing the data, it's there, it's there, it's done. You, you are, you're done. You have to annotate the chart. There was an issue and what this technology enables. [00:19:49] It's kind of the full culture over the data essentially. You can model it, you can transfer it to you, clear it much better. Plus, if you see a product and marketing analytics, uh, [00:20:00] chart, let's say a funnel or like a journey chart, it's gonna align with the charts that you see in your bis in power bi tableau in the like, uh, like a, like a all hands meeting. [00:20:10] What is presented by, let's say the CEO in the company. So. What we got often is like, finally this is the element. Finally, I don't see this huge drop off in the chart 'cause we were missing some data. Now it's ing a data warehouse. It is, it is. Basically we are seeing the same data as what we expect, actually. [00:20:28] So that's a, that's a great value that it brings for, you know, like the, the Marketing Falls folks. Um, so they have this trust. Essentially the trust is increased in this, in these digitals. [00:20:40] Phil: Very cool. I think the, like aligning with BI and the rest of the company, like in the all hands meetings is, is a really easy one to grasp for, for marketers. The, the one that's kind of new to me that, that. You called out there is like owning and controlling that data. Like marketers are very familiar with this issue of just like, oh my God, we got a huge spike in [00:21:00] traffic. [00:21:00] We're excited, we're pumped. We dig and figure out. It's like, oh shit. It was bots, bot traffic. [00:21:05] Like, and, and like, we need to annotate that thing every time we show it. Like every, oh, like everyone gets excited. What was that spike? Oh. Like that, the annotation says it was a bot. And we see this in email land also. [00:21:16] There's a lot of like, um, enterprise email clients who will just like mark every email as open and like, you see a huge spike in engagement and it was all fake. And, and so that's really powerful there because like a lot of the times, um, I've gotten those tickets on the marketing ops team from the marketers saying like, okay, it was bot traffic. [00:21:36] How do we like fix that? How do we just remove it from the dashboard so it doesn't affect all of our average metrics? Because our average opener right now is completely inflated because of that one little thing. So you're saying because we get to control that data because it's on the warehouse, we can make those tweaks and the annotation isn't static. [00:21:52] It actually lives in the data and it doesn't affect like future reports. [00:21:57] Istvan: Exactly, exactly. One more thing I would like to highlight [00:22:00] here is that the end-to-end traceability of why we see a number on the chart is, is possible with this type of technology. Uh, for example, via us, we just present the s QL query that we generate. You can copy paste it, actually, you can copy paste your BI tool if you want to have it in your, like, in the, your static dashboards. [00:22:18] Uh, the same query. You can create the same chart in the be Able tool that just the same as what we built or what we show. And this end-to-end traceability is the one that is making it not a black box anymore. It's, it's an open horizon, essentially. You can look into it how the things work. You can see the joins. [00:22:34] If we use joins. Uh, you can see all the other things that we do in the SQL and data analyst, like marketing ops. People can review it line by line. And often our customer calls, we do it. We, we do review line by line, the follow up query why why things happen. What is the, how do we calculate conversion rate? [00:22:52] How do we calculate time to convert metrics? Um, all sort of things like that. We, we can like review and, and talk about it. And they [00:23:00] might have like a different opinion about like how to calculate retention for example. Um, but again, that's a good discussion to have because now we can finally actually talk about this. [00:23:08] Yeah. [00:23:09] Phil: Very cool. Yeah. It could lead to finally folks having, uh, the same data dictionary and the same metrics dictionaries. Um. Yeah, so like one, one of the things that I wanted to ask you about is, um, like [00:23:23] 5. How Decoupling Storage and Compute Lowers Analytics Costs --- [00:23:23] Phil: one of the prerequisites for doing warehouse native or even thinking about going warehouse native is obviously having a data warehouse in place or, or a data lake for a lot of orgs, marketing teams especially. [00:23:34] This is still pretty new, like, um, but, but a lot of folks would argue that, you know, data warehouses. For some companies that are leading in this space, a lot of companies are, are on the tech side. For a lot of folks, data warehouses are actually old. Like they're, they're not new tech, but they're more excited about the data lake side. [00:23:52] The data mesh side, data, fabric, semantic layer, real-time data, the platforms. Um, can you help us like bring clarity to the data [00:24:00] management space? I know you're really deep in there. What, what, what advice would you give to marketers there that are just like waking up to the Doty warehouse, but we're talking about like what's next already? [00:24:10] Like is the data warehouse kind of dated already? What's the difference between a data lake? What are your thoughts there? Stephan? [00:24:16] Istvan: Oh yeah, that's a good question as well. Like a data warehouse, how I grew up, how I like as a professional. Data warehouse was already a new thing. [00:24:25] Phil: Mm-hmm. [00:24:26] Istvan: Uh, if you imagine like the old ones, actually, I don't even know the names and the brands who were like, you know, before in the two, in the years of like 2000 or something like, like that, those years there was, there was data warehouses, like Oracle maybe had [00:24:38] Phil: Mm-hmm. [00:24:40] Istvan: I don't know wanna, that was like, oh my God. That was like the, that was like the first data lake trial I would say. Uh, from Google and I think it was Google. Yeah. Anyway. No, it's actually now, I would say, eh, how I see it, everything is a data lake kind of. And um, but we, we [00:25:00] still call them sometimes data warehouse because that's something that you can tell to others and they'll understand that a lot of data there. [00:25:06] Um, just a slide note here, like compared to a database, if you tell somebody the database, what's the database? It's like an operational thing. You, you keep your like, I dunno, subscriptions there or like, uh. I don't know, some operational stuff about the application that you're building in a data lake. Data warehouse. [00:25:22] You collect data and you forget about it. You don't talk, modify that often. Maybe if you fix the issues yes, then you modify. But it's most like isec only for, in my head, essentially everything is a data lake. And what's the difference between the, the data warehouse and data lake? Um, data lake storage of it is completely decoupled from computation. And what does this enable is that you can collect your data virtually for free into a storage layer. When you want to compute, let's say a funnel or like a journey chart or something, then you speed up a cluster that relay the data directly from the, the, like the, the [00:26:00] storage layer, compute, the, the computation that actually measure like conversion rate, shut down and show you the result essential, or basically show you the result and the shutdown. And this is hugely cost effective. Because the storage is where you have the most, the biggest problem, like, you know, like you, you collect all these billions free events monthly. You store them essentially mutually for free. I would say compared to like, you know, how much a classical data warehouse would cost, like imagine classical data warehouse would run all the time just to have the ingestion possible. [00:26:33] You know, like when you were ingesting data to a, to a data warehouse, originally you had to have the cluster spinning all the time. So there was a revolution when you had like outer scaling enabled, you know, so you could scale it down, and it was only ingestion and it was not computation for high computation. [00:26:49] You needed to scale it up the cluster, but now you basically scaled up to zero, uh, as the com be the computation, and you just store the data. And essentially you can, you can leave, like [00:27:00] for example, we have customers that they, they've billions of roles, but they may be around a single chart, uh, like single competition once per week. And for them, um, this is really makes sense because like they, they don't need to pay essentially nothing for the storage and the computation, which is expensive. That is like once a week and they don't care about it. [00:27:19] ​ [00:29:11] 6. How Composable CDPs Work with Lean Data Teams --- [00:29:11] Phil: one of the questions I have in my head is like, we're obviously big supporters of the composable movements, uh, on, on the show, like Census was a sponsor for, for a long time. Um, we did this like deep dive on package CDP versus composable, CDP. The composability route isn't the best solution for everyone. [00:29:28] Like traditional platforms are totally fine for some situations. Plenty of companies have invested heavily in CDPs and and in traditional bi, and they're doing okay. Do you disagree that like, like do you think these. Investments in traditional BI and CDPs were misguided, and the fact that they're doing okay is a reflection of like their decisions were the right ones. [00:29:51] Is there a path to integration there? Like what are your thoughts? [00:29:56] Istvan: I believe that the current technologies, the [00:30:00] current CDPs and, and. Like, so even product, uh, having their own storage layer in built it totally makes sense for, for the majority of the companies. Essentially, I believe in what I see, it is probably less send less, uh, whether at, let's say the, like the companies that should switch to a warehouse data approach are more and more of these. [00:30:22] You see more and more of these, I believe, um, one because there's much more data now than let's say two years ago. And this data lake and all these approaches, like, I think it makes it, um, possible to, to, to run a company from the get go on top of the, like, data lake, the, the, but I think it's like this, these traditional approaches are very good when you, like in the early stage and you can like keep, keep it running. You don't need to hire a person. That's true for this, this warehouse native and this warehouse approach is that you [00:31:00] need to have least one person in the team that knows what he's doing. [00:31:03] Phil: Yeah. Yeah. One, the one person [00:31:05] data [00:31:05] Istvan: person. And in fact, it is interesting, like, like I would say most of our customers is like, have a very, very small day to day. Yeah. [00:31:15] Phil: Very interesting. [00:31:16] Istvan: if we have some that doesn't have any, the team, like zero people, uh, essentially the data, they rather like big query. Uh, we help them, basically we give them advice, what to do. [00:31:26] Uh, it's not a consultancy what we do, but like we just send them like a slack message on like a slack connect that this is what you should do. Uh, and basically they're enabled just by that. So it's like a backend engineer takes care of the, of the, of the data warehousing, but definitely you'll need an expert. [00:31:43] And, um, I would say it's still hard to come by a good, good expert that can understand. Platform problem. The data analytics, problem marketing use cases, product use cases, all the other use cases and deal with single, uh, thing, which can support all these, all these use cases. [00:32:00] But there are these folks out in the wild. [00:32:05] Phil: Yeah. [00:32:05] Yeah. The, the technical data, um, unicorns, I [00:32:09] Istvan: Yes, [00:32:09] Phil: of folks call them. [00:32:10] What do you think are like the ca the, the characteristics of companies where maybe folks are working at some of these companies right now listening. What are the characteristics that you would point out that you would say you guys are good candidates to explore Warehouse native? [00:32:26] 'cause you said a lot of companies early on and for a lot of traditional companies. Traditional tools are fine. They don't need to go warehouse native. [00:32:33] But you, know, you you obviously grow in, there's hunger in this space for, for this kind of data management, what are the characteristics that that jump out to you? [00:32:41] Istvan: This is a great question because this is our go to market [00:32:44] Phil: Yeah, yeah. Product market fit [00:32:46] Istvan: Exactly. Uh, I have to be honest, like it is hard to find, uh, like the common, like a segment of companies where this like really makes sense. Yeah. I, we, I denied that a few, like I. [00:33:00] Like typically e-commerce, travel and all these companies in, in those segments, like, you know, like, or media, entertainment, gaming, just because the sheer volume of data is big for, for, in my head it makes sense to use this, go for this approach. [00:33:13] But on the other hand, they really have quite some, uh, inbound leads that they are not coming from this, this segments. And what I've found there is, um, typically have all one or two or small data teams, maybe one or two data engineers and or a small data team. But they are kind of enthusiastic about this. So it's a, I would say if you are listening to the podcast and if you are, if you knew about this solution, you might be, you know, like you are a better of, of this. You are already a good candidate to try it out. While what we have very, like the big struggle is to evangelize this, educate people about this. [00:33:53] It's an uphill battle. To tell, like if I, if I go out and talk to companies and like, you should switch to [00:34:00] this social because it just makes more sense for you there, it's, it's very, it's very hard to tell them why. It's a, it's, maybe they understand why, but, but it's like the cost of change is, is huge, obviously. [00:34:11] Phil: Yeah. [00:34:12] Istvan: So, [00:34:13] Phil: Yeah, cost of change for anyone in the audience who like manages MarTech solutions and has like a decent integration in place and they're all talking to each other, you know, a salesperson hitting them up out outbound and saying like, Hey, do you want a new analytics platform? It's like, [00:34:28] get outta here buddy. We got 75 things on our, on our [00:34:32] list, [00:34:32] 7. How Seat-Based Pricing Works in Warehouse Native Analytics --- [00:34:32] Phil: but one of the. The, I guess, benefits that resonated with me the most as someone who owned the marketing technology budget for different companies is the cost savings angle to this. If you are a company that has a data warehouse already, and you've got one or two data engineers that are building and managing that warehouse, and they own the data strategy, like, why not? [00:34:57] Why wouldn't you explore this? Because it comes [00:35:00] with a lot of cost savings, like this whole idea of. Copying data to other systems and like customer engagement tools, CDPs, they all charge by number of events. They all charge by number of users in the database. Talk to us about like the pricing for Mitsu and like how that is different from the amplitudes in the mix panels that are all charging by like raw number of events, number of users in there. [00:35:25] 'cause I feel like that's one of the differentiating factors, right? [00:35:27] Istvan: Yes, yes. Oh my God. This is a huge topic. Uh, I like, I love it actually. It's, uh, it's interesting. Like, let's, let's talk first about the Interation approach. You mostly pay by the volume of data you send over. Uh, it's sometimes based the number of unique users, like, or check users, empty use or like in case of other solutions you pay. [00:35:50] The number of events, how you, what you send to these tools, um, multitrack events. For us, it is seed based. 'cause we don't have any of [00:36:00] your data. We don't store any of your data. We are basically a, a, a bi tool essentially. If you, you can think about it like it's a BI tool, just it is, it's a self-service. Um, so we are, we cannot even charge for the value of your data you store in your own cloud, in your own data. [00:36:18] Like it would be, I mean, some of the competitors, they do say that they charge with a number of users who storing their own data warehouse. I find it a bit. I don't know, countering Tweety or like, I don't know, like I, I get the business point of view. Why does it make sense? Uh, but on the other hand, I think this is all for our, our ourself. [00:36:40] The seed based pricing is also a bit of an advantage because it just, for some companies it's much more, makes much more sense. So I have like this 15 people in product and marketing, uh, but I have, they want to use the application, but I have like multi 10 billion events. If I would send that out to a third party social, it would cost me [00:37:00] monthly, maybe half a million. [00:37:02] It's like, there is, there is numbers like this in this space. Like if you have 12, if you have all your data out, it is, you can have paid, essentially the graphy investment is not justified. [00:37:12] Phil: Yeah. [00:37:14] Istvan: And basically, I, I, it's, it's hard for a business for us to, you know, like, uh, to say, okay, but you get from ourself like for a thousand a month maybe. While the competitor would charge you few hundred thousand. It's so hard to like your to be reading this spot. [00:37:32] Phil: It, it feels like too good to be true in a [00:37:35] Istvan: Exactly. Exactly. It, exactly. It is, uh, I, [00:37:37] Phil: the catch? [00:37:39] Istvan: exactly. That was a catch. Uh, what we measure from our comp, like our cost customers, we enable 80 to 90% of cost savings on product market analytics for the segments that you have targeting with large data sets. So of course for some companies they wouldn't be cost setting for low volume data sets. [00:37:58] We would be actually more [00:38:00] expensive. Um, but I would say, so this is the, the selling point. This is why companies come to us. Um, obviously large enterprises would go for like 200 seats. It wouldn't be cheap. But how I see there is there, I guess it's like a chart. What I should draw at one point, like how much data you need to have. To justify the cost of switch, you know? And there is like, if you reach like 80% cost savings, then you should just definitely switch. [00:38:30] Phil: Very cool. [00:38:31] I like [00:38:31] the. I, I like the self service angle to that, because o obviously, like you're, you're encouraging people to invite other folks from the team. That in increases the number of seats. Um, I, yes. I feel like this is by far like the strongest argument. Like if, if I was like managing that MarTech budget and internally, you know, like we had a ton of things on our roadmap, blah, blah, blah. [00:38:57] But I came to the table and I said like, [00:39:00] Hey, we have billions of events in our CDP and in our product analytics tool right now. This is how much we're paying per month. We could cut our cost by 90% by switching to a warehouse native [00:39:12] solution like that would turn. Like, like eyes would just be like, wait, what? [00:39:18] Istvan: yes, yes. [00:39:19] Phil: this is too good to be true. Like, and, and so like, I, I could champion this and explain it and we could easily get that on the roadmap. Like if it came with 90% cost savings, 'cause then we could use that money for a bunch of other [00:39:31] stuff. Like our ads team wouldn't be [00:39:33] really pumped if we like shaved all of those costs. So yeah. I feel like we're, we're riffing on, on product market stuff for, for your audience here too. It is one of the most exciting parts of this for sure, because like one of the, like the bane of the existence of marketing operations people is integrations and APIs and connecting data between all of these different systems. [00:39:55] And now we have a central repository. And it's funny because like [00:40:00] you're, [00:40:00] 8. What a Data Warehouse Does That Your CRM Never Will --- [00:40:00] Phil: you're working off of this assumption that, you know, customers have a data warehouse and also that customers see the data warehouse as the source of truth. And that marketers also see it as the source of truth, and that's where like there's some disconnect in the space. [00:40:15] Um, my co-host Darrell, like he, he's put out a couple of, uh, polls. He's got a huge audience on LinkedIn, and the majority of folks still think that the CRM is the source of truth for [00:40:24] marketing. And so we're still facing that like slow transition to waking up that the data warehouse should be the source of truth. [00:40:32] Because like you said, a lot of teams. Like, maybe they have one or two people managing that data warehouse and they don't care about marketing. And so marketing is off in their own little land. They're doing their thing in the CRM and they're just like, ah, data warehouse is something for engineers. Like, I don't, I don't put my marketing data in there. [00:40:48] Like I don't really care about it. So I. [00:40:49] Like that's, that's the other uphill battle you have for warehouse native. Is that like the change management process for folks waking up to a better way to manage your data. But it's funny from an engineering [00:41:00] standpoint, because like marketing data stack architecture has been around forever and it's just taken so long from marketers to wake up to that [00:41:07] Istvan: Yeah. Yeah, yeah, definitely. I mean, CRM is an interesting one. I, I I see this, that some companies have this as like a RO truth. Yes. Uh, although I, I consider it this like a CRM tools, like a glorified excel sheet. You're very easy to make a mistake and you never find it out. And, uh, in the data warehouse, it's like a slightly harder to make a mistake. [00:41:33] And, uh, like you have to have like a, a job that updates the table and runs every day. And you somebody review that job. And if you find a mistake, that is then if you make a mistake that it's a mistake for a thousand, uh, records or like a million or like a billion records even, because it's much easier at the spot as you think. Um, yeah, that's definitely, but it's, um, evangelizing this as a, as a solution. It's, it's [00:42:00] hard. It's [00:42:01] Phil: Yeah, it's hard, but it's exciting and, and like once folks see the, the cost savings and the unification potential of, of, of data, there's, there's a cool story there to, to tell. Um, uh, [00:42:12] 9. How AI-Assisted SQL Generation Works Without Breaking Trust --- [00:42:12] Phil: one of the other like exciting parts of the product is the self-service side of things. And, and we've chatted with different folks in this space too, like democratizing bi and allowing, you know, like non-technical users to finally not have to know it's equal and be able to like, query stuff and build dashboards and basically have conversation with their data. [00:42:30] There's a lot of excitements in, in, in this space. But I wanted to ask you about like, the risk of replacement for analysts. We do, we asked this to, to a couple different folks on the show recently. Um, a lot of people think that it's like some of the jobs most at risk for AI disruption in the next year or two are a mix of like reporting in BI roles. [00:42:51] Like think of analysts that. Are basically just on the front end of the dashboarding and the data viz side of this. And you [00:43:00] know, their jobs are already getting super easier. Like NLP is changing [00:43:04] stuff, you know, like non SQL experts are able to have a conversation with their data. It's a category of like recommendations and pattern detection features that are, you know, reshaping the category. [00:43:15] There's like predictive engines. I'm less convinced here, but there's also tools like innovating and the data prep and the modeling automation and cleaning integration. What are your thoughts here? Like are [00:43:25] are analysts that are just building reports and go into marketing and saying like, Hey, what dashboard do you need this month? [00:43:32] Like, are those folks up for replacement? What should they be doing, uh, differently? [00:43:37] Istvan: Okay, this is a good topic. I like it as well. I'm a big believer in ai. Honestly. I like AI tools. I use them daily basis on, on a daily basis. I use Core Server. If I do to do some, some code or whatever sounds, skip. Uh, I use JG PT for marketing stuff, like a blog post a lot, you know, so it's a, I I really, I really believe in, it's like it's, it improved whole lives. [00:43:57] Uh, with that said, I, [00:44:00] I'm also, uh, disbeliever in like, it replace, uh, current, like in the next couple of years. Like, uh, any role significantly, let's put it significantly. I mean, it might replace some people. Uh, you might get like. Less people hired in software or less people hired in data. But I wouldn't say that you will have a, a future where there is no data engineers, no data analyst, there is no software engineers because of AI in the data analytics. [00:44:31] I think it's even more critical to have, you know, trustability then let's say software engineering, I would say. Because imagine you are like, imagine you are in a meeting with the, like in the store, so board meeting or something like that. And there's a question popping up. How many, how many sales I know items we had like last month. [00:44:51] Can you type it in the chat box and like that chat box and like some number come up and it was like, okay, do we trust this? But the dude ate something? [00:45:00] Uh, actually I have a really good, uh, tour test. Let's say, let's go the during test. Ask the AI the same question twice. If they, if it gives you the same answer, that's most likely okay. [00:45:12] Uh, for me, I still, even with the g pt, very new model, new like, uh, models, spikes very often get for the same question, two different answers in the, if I like, I, for example, I ask like a seql type of question, like, how would you solve this in this data warehouse? This is my data, this is my table. How would you calculate something? And it often still creates syntax errors. So I, it, it is getting there, but it's not there yet. Um, so I would use it as an enhancement to your productivity, uh, in, in every aspect, in every domain. Uh, data analytics, software engineering, marketing, right? I wouldn't trust a, uh, chat bot to write like, uh, a blog post entirely. I would use it to like enhance [00:46:00] my, my, uh, it's a mediocre English. This I news very often. So like, you know, I write it down. I have like a couple of ideas. I just put like bullet points. It writes an amazing like, um, paragraph. Uh, but it actually, I'm the, I'm still the, still the ideal owner. What, what should go there? [00:46:19] The samples for data, I would say you should know what you are doing and you should use AI just to speed yourself up, essentially enhance your, or improve your performance. Um. [00:46:31] Phil: Yeah, definitely. Uh, I agree there. I, I'm, I'm like both super surprised and impressed sometimes. Uh, I, I have both a subscription with, with Clode and, uh, and OpenAI and I often like will ask the same question to both, and sometimes they're very similar, sometimes completely different. One is super helpful, the other one's useless and they kinda like, uh, trade places there, but. [00:46:56] Yeah, sometimes it's super impressive. Sometimes it's just like really, [00:47:00] like you weren't helpful to me at all. Like I was just troubleshooting a templating language issue with, uh, with a client the other day. And, um, they were building liquid code, uh, templating language to personalize an email. And the token kept breaking. [00:47:14] We couldn't figure out why. And we threw it in in GPT in both code and we're just like, here's our syntax. Here's like our, our fields. Why is this breaking? And it took like 10 iterations of back and forth. Still doesn't work. Try this still doesn't work. Try this. They didn't get it. They, they, they couldn't figure it out. [00:47:30] Sometimes they do, but yeah, it, it, it is like moving super, super fast. But yeah. I was curious to get your take on, on the analyst stuff specifically. 'cause I know you're in the space and. There has been a lot of cool excitement there. Like I, I was just researching a, a tool that, uh, ThoughtSpot was using, I forget the name of it, but it's essentially like you have a chat bot experience, like it's an LLM, they built it, it's proprietary and it sits on top of your data warehouse [00:47:57] and. You literally have a conversation with your [00:48:00] data and you can also say like, Hey, like this query that I just said, like, turn that into a, a SQL syntax. And, uh, so like that, that's been, been changing super fast and, uh, yeah, I don't know. It's, it's exciting. And, and, and also, yeah. [00:48:15] Istvan: I think everybody is coming up with this text, uh, solution. Quite strong opinion about that. Sequel language itself is a bit too granular, like fine grain to be supported by like an LM Um, I'll give you another example. Like for, for coding, let's say live, we measure live coding. If the LLM writes every, every single line of codes, you are really, really easily getting into like a trouble that you, you are not able to get out. [00:48:49] Uh, but there are coming tools that, like, there are some tools that are not using the, the line of code as a block of like unit of, of, of the tail and con, but some kind of bigger code [00:49:00] blocks. So like, this is definitely must be here in this, in this application, let's say, um, authentication piece. You don't need to re invent. [00:49:09] There's, as an lm, every time you write an application, the authentication piece, it's already done. It's like a solved problem. Or like a Skype integration or like, I don't know, integration to your whatever type of like, uh, database or something. It is solved already. You have a lot of snippets on GitHub that you can just copy, paste. [00:49:29] Problem with these alliance, they're not copy pasting solved solutions. They're kind of, I would say, reimagining it with their host nations. It's always different, and I would say the same goes for data analytics, that there are certain blocks of iCal code that you don't need reinvent. I'll give you an example. [00:49:47] Like for product marketing analytics, calculating a funnel, it should look like roughly the same in every company. Not, it is not black magic. If you have like, either you do a join or there is like another way to do with window [00:50:00] functions. Uh, each has some down like pros and cons. Uh, but the funnel cock is SQL is like, there's two ways to do it essentially, and you shouldn't reinvent the wheel as as, as another. [00:50:12] So you should just take this block, you know, there is some place service you plug in. And, and, um, just execute that block. Uh, so what we are trying to build in the future is we try to, we are embracing people, AI to the application as well, but we're not going to reinvent, like right in the SQL code. We believe that these blocks of sqs are, are like they're sold. [00:50:36] The AI should just pick these blocks up and teach them to better and there therefore there be much less, uh, space for error. That's my, that's my, uh. Very often like think of ai, especially for us in the application. [00:50:51] Phil: Very cool. Uh, got a couple last questions for you. We're, we're getting close on time, but I, [00:50:55] 10. How Warehouse Native Analytics Works --- [00:50:55] Phil: I wanted to get your take on like, you know, if you were to give advice [00:51:00] to marketing operations folks, data professionals that work with marketing teams in the next like three, five years. What are like a couple of things that they should be doing to prepare for this future of ai, you know, displacement or, or innovation there, you know, like learning resources or like, maybe you'll say like, everyone should still learn SQL and, and not rely too much on, on text to sql, uh, technology there, but like what, what comes to mind? [00:51:28] Like how, how should folks future-proof their careers? [00:51:32] Istvan: Oh, that's a good one. I don't know the future. Uh, I would say to have a good understanding of the fundamentals is crucial. Still, you know, I, what I don't like as to show with like employees and like you are hiring is if you hire like a junior person, it's very often that they will turn to ai, uh, with much bigger confidence. [00:51:57] They, they should, and [00:52:00] they, they essentially. They don't choose their time to understand the fundamentals. I would, I would go the other way. I, I would, in the first years of your career or something, you should definitely, you should use AI to understand how it works and, and, and like, you know, learn it and see the limitations. [00:52:18] But if you are not familiar with like, the fundamentals, if marketing and anything essentially like software development, data analytics, I don't think it's, uh, then you're going to be replaced most likely by the, it's bottom line there. Yeah. Because as a, as a human, your power is coming from like talking to other humans in a language that both of you understand. [00:52:39] And if you don't have a fundamentally like baseline for talking, then basically cannot, you cannot function in a, in a company, in an organization. [00:52:48] Phil: Yeah. [00:52:48] Istvan: Yeah. So you, you keep really replaced by an AI that in that case it's a, [00:52:53] Phil: Yeah, great advice, uh, [00:52:55] ish fan. This has been super fun conversation. I really appreciate your time. [00:52:58] 11. Happinness --- [00:52:58] Phil: One last question for you. You're, [00:53:00] uh, founder, CEO marketing analytics leader. You're also a dad and play a lot of Legos with your son. Uh, one question we ask you, everyone on the show is like, how do you remain happy and successful in your career building, running a company, but also having, uh, a young family? [00:53:15] How do you find balance between all the things you're working on while staying happy? [00:53:19] Istvan: Yes, so. Like, in my free time, I spend my all, all my time with my, my kid and my family. It is very, it's a great time now because, uh, he's like three, well, three years old. It's good to be there. So I play with the Lego with him, which is great. Um, keeping the sanity is hard. I, at work, definitely being a founder is, uh, oh my God. [00:53:41] This is a huge challenge. Definitely. Previously I was an individual contributor in many, many companies, and. You know, like you are used to, uh, a lifestyle that basically you don't have responsibility, uh, besides your like line of work that you do work like every day and suddenly it's like 180 [00:54:00] degree, like turn. Uh, I have, I have a lot of sleepless nights, uh, during this, uh, building company and, and decisions, judgment calls. Oh my God. That's a, that's a very huge one. Uh, the good thing is that I feel it is less and less, um, stressful to have a judgment call sometimes to an extent that I, I stop worrying about things that I should worry about. [00:54:25] And that's interesting. That's, uh, that's different than experience. Like I know, like that's why a car, I did something about the car. Uh, I like checked out. Okay, this is good. took while I was like. [00:54:39] Phil: not worth my energy anymore. [00:54:41] Istvan: Exact with, I had like half a hour for this project. Okay, fine. Solid. [00:54:45] Phil: Yeah. [00:54:46] Istvan: Previously it would have taken like half a year to find a good one, but the could correct price and like good parameters. [00:54:52] Now I just don't care anymore about like, significant decisions in your life, uh, which is amazing by the way, from I get, uh, a lot of, uh, [00:55:00] comments about this river from my wife, five. The, the, the main thing that makes me happy at work is like seeing customers coming in. Uh, that's, that's amazing. It's like I, as an ic uh, selling a software individual contributor previously now selling software to companies that are like way above our like league, like we are punching up, we are punching [00:55:24] up a [00:55:24] Phil: Yeah. Yeah. [00:55:25] Istvan: and we still close deals with companies that like a hundred times bigger than us, a hundred times than us, and they, I think, clicked when they look at our LinkedIn page. [00:55:34] They're like, okay, let's, let's close this linear page. They're like, I don't believe there are only three people or six people, or, you know, let's close it. And they still didn't sign with us. Uh, which is amazing feeling. And, uh, that's something also I have to, I have to learn that. Uh, I, I, if you come from the IC background, you are like serving others in the company and your goal is to, uh, help them as much as [00:56:00] possible. Being a sales founder, CEO type of personality, the goal is to maximize gross revenue, whatever metric that you're looking for in the, in the, in the startup. Um, that, that, that mindset has to be changed. And I'm changing that as well. Like I, I am, I'm more and more bold in asking a bigger deal whether I'm doing a sales course, but it, if it happens, it's like ultimately makes me super happy and like excited about it. If it doesn't happen, it's like that. It goes down very [00:56:31] Phil: Yeah. Yeah. It's hard not to be emotionally in, in, [00:56:35] in invested on, on the sales side when there's like peaks. You're like, it, it's so good and you're [00:56:41] feeling so great and, and when there's dips and valleys, you're. It's like rethinking your whole like mission statement [00:56:48] and you're just like, oh man. Like I was really hoping that one was gonna [00:56:51] close and it didn't work [00:56:52] out, and it's like, do we need to rethink this or that? [00:56:55] But yeah, I, I [00:56:56] like your, your stand of, of just like spending your whole free time. Like we were chatting [00:57:00] before. I was like, well, what are the hobbies you have going on and stuff. You're just like, ah, I don't have hobbies. Like, I play with my kid and my family, all the free time that I [00:57:07] have and [00:57:07] I love [00:57:08] Istvan: That's it. That's it. One last thought. One last thought about this happiness part. [00:57:12] Phil: Yeah. [00:57:12] Istvan: for other founders who are listening, maybe like, definitely get the co-founder if you can, just to have a shoulder to cry on. Yeah. It's like, it's, yeah. I have my fond friends that we come together often, like in a coffee or something and just complaining about life, and it's absolute [00:57:32] Phil: Like group therapy [00:57:33] for for founders. Love it. Awesome. [00:57:36] It's been super fun ish. Fun. Uh, thank you so much for your time. We'll, uh, we'll link out to, uh, Mitsu. Folks can check it out if they're curious. Uh, hopefully, uh, this conversation got folks interested in and curious about the, the warehouse native space. [00:57:48] I'm really bullish on it. Obviously you guys are. And, uh, yeah, really appreciate your time and congrats on what you've built so far. Excited to see the growth. [00:57:55] Istvan: Thank you so much. Thank you so much. It was a pleasure to be here.