Snowflake Summit 2025 | Artin Avanes, Snowflake & Gerry Bavaro, Merkle

Clips
News
More from Snowflake Summit 2025

Artin Avanes

Head of Core Data Platform

Snowflake

Gerry Bavaro

Chief Strategy Officer, Merkury

Merkle

Merkury connects the dots in ad data through AI-driven insights

As artificial intelligence reshapes the advertising industry, data security and insight delivery have become mission-critical.At Merkle, a Dentsu company, Snowflake Inc.’s AI-powered data platform is helping marketers analyze customer motivations and behaviors without compromising on governance or speed.Snowflake’s Artin Avanes and Merkle’s Gerry Bavaro discuss Snowflake’s new AI tools.“Knowing customers, knowing prospects, that’s what marketing and advertising is all about, and [then] … connecting it with the experience,” said Gerry Bavaro (pictured, left), chief strategy officer, Merkury, at Merkle. “Snowflake has really helped us to have one consistent infrastructure

play_circle_outline Merkury's data challenges and solutions before and after Snowflake integration

play_circle_outline Balancing innovation with compliance and maintaining a composable data platform

play_circle_outline Importance of security, governance, and trusted data in AI applications

play_circle_outline Snowflake's differentiation with open table formats and Iceberg integration

play_circle_outline Future focus on composable platform for client use cases and driving real business outcomes.

Info
Transcript

Artin Avanes, Snowflake & Gerry Bavaro, Merkle

Artin Avanes

Head of Core Data Platform Snowflake

Gerry Bavaro

Chief Strategy Officer, Merkury Merkle

In this Snowflake Summit 2025 segment, Artin Avanes, head of core data platform at Snowflake, and Gerry Bavaro, CSO of Merkury (Merkle, a dentsu company), join theCUBE’s Dave Vellante and John Furrier to unpack how data-driven enterprises are raising the bar on security, governance and AI innovation. Broadcasting live from the show floor, the conversation dives into Merkle’s journey modernizing its identity and data platform on Snowflake’s AI Data Cloud.

Bavaro outlines the real-world payoffs of Snowflake’s privacy-first architecture: 64% faster pipe... Read more

explore Keep Exploring

What are the three ways in which Merkury is enabling agencies within Dentsu? add

What is being brought into the Snowflake environment and how can innovation be balanced with compliance when using AI in that environment? add

What is important to consider when thinking about security and governance controls in relation to AI applications? add

What is Snowflake's perspective on the move towards open table formats and how does it affect their differentiation in the market? add

What problem are we looking to solve a year from now? add

bolt Powered by CUBE AI

Artin Avanes, Snowflake & Gerry Bavaro, Merkle

search

Rebecca Knight

>> Good afternoon everyone, and welcome back to theCUBE's Live coverage of the Snowflake Summit 2025 here in San Francisco. I'm your host, Rebecca Knight, alongside my co-host and analyst, Dave Vellante. We've got two great guests for this next segment. I would like to welcome back to the stage, Artin Avanes. He is the Head of Core data platform at Snowflake. Thank you so much, Artin.

Artin Avanes

>> Thanks for having me. This was always a great tradition now. Everything.

Rebecca Knight

>> And Gerry Bavaro is the Chief Strategy Officer at Merkury Dentsu. Thank you so much-

Gerry Bavaro

>> Thanks for having me....

Rebecca Knight

>> for coming on the show. So Gerry, I'm going to start with you. Why don't you tell our viewers a little bit about Merkury and the kinds of data challenges you were grappling with before you met Snowflake?

Gerry Bavaro

>> Yeah, before Snowflake helped us. So Merkury is the global identity and data platform across Dentsu. Dentsu as one of the larger global agency networks, has three parts of the business. One is Merkle as the customer experience agency handling client data management, CRM experience, driving that experience across any channels. Then there's media. So the media agencies of Dentsu that manage media planning and buying. And then creative. Creative and content development across the creative agencies. Merkury is enabling three things for those agencies. One is to know the person best, and that is a customer or a prospect. So client's first party data, plus we have a data product. Merkury has consumer data globally that enhances that data. Two is, know something about them with the data. And knowing customers, knowing prospects, that's what marketing and advertising is all about. And third is, connecting it with the experience. We don't care what technology is enabling that experience across media, direct mail, CRM, email, or the site experience. It's interoperability and plugging all of that data in to deliver the best and most personalized experience. So that's Merkury, as we like to say, the fuel for getting closest to customers, closest to those individuals who are not yet your customers and moving them to action.

Rebecca Knight

>> Okay. And as you said, you need to know your customers, you need to know your prospects. So that's a lot of data to be ingesting and analyzing all the time.

Gerry Bavaro

>> That's lot of data. Yeah, so before Snowflake, it was a lot harder. I think Snowflake really helped us with three things. One is, for data management environments and our platform, our client's data management environments, and our Merkury platform, we had multiple code bases across multiple cloud providers. Very complex. We had to deploy and maintain those environments, and Snowflake has really helped us to have one consistent infrastructure. The other thing is, our clients today with regulatory and privacy, they don't like to share files back and forth. So what Snowflake has helped us with is zero copy, zero share. Snowflake sharing, native apps. We have a native app in the marketplace that brings identity, brings the processing, brings our data into the Snowflake environment of our clients. So that portability is really important. And then I'd say the third factor that has really helped us with a tremendous amount of customer data and then thousands and thousands of attributes that we enhance with is traditional analytics methods of lots of data scientists and lots of complexity around managing the data is now moving to AI. So we use Cortex to really speak to the data, versus query the data. Ask the data questions and bring insights back a lot quicker.

Dave Vellante

>> So Artin, we had the pleasure of sitting down with you on Monday to go through, really the platform announcements and your perspectives on the platform. And we also heard from Customer Zero how they're using the platform. So we were just talking off camera, the customer said a lot. There's a lot of... I feel like I'm at re-invent with a fire hose. And so what are the salient aspects of platform that we should be focused on out of Snowflake Summit 2025?

Artin Avanes

>> Yeah, I think there are a couple that I would highlight. First is how we have really evolved our compute infrastructure. We just heard it's easy to use, but it's never enough. There's never an ending of further evolving the platform, make it more easier. And our announcements around adaptive compute or adaptive warehouse are really critical and next step, I think, in terms of the evolution of compute. And if you pair that with the AI revolution that we are all witnessing right now, it makes it a very powerful, easy to use platform. The second piece I would highlight is, everyone is raising towards AI. Everyone wants to use Cortex. Some are faster than others. But you want to do it in a secure and governed fashion. Security and governance are paramount and the prerequisite. If you don't have those in place, the right controls and they apply also to the AI applications that you're building, you will fail as an organization. So I would call out the trusted part in all the announcements we did around security and governance, this as my personal favorite.

Dave Vellante

>> You're ticking all the boxes because when we talk to customers, they say the biggest thing is skills. And then right behind that is, we have to have this governed. The whole security and compliance piece. You're essentially saying, we simplify that. The compute infrastructure piece, the adaptive warehousing provides greater granularity and simplicity in terms of sizing. And I want to ask you a follow-up on AI. You're saying, does AI inject a less predictable sizing dimension?

Artin Avanes

>> Yeah, I mean, for sure. I think if you look at the demo that was presented at the conference at our keynote, Christian and Jeff, you might ask questions that you don't think about the sizing. You are now suddenly interacting with a system in a very natural way. You can say, "Give me all the sales records of the last 20 years." This could mean billions, not even trillions of records that come back. You have to have a system that then automatically, in an adaptive way, is able to handle that workload. And I do think you are hitting it on the right spot here as it relates to you have to have an infrastructure that is able to scale it without needing the user to do any action. And that's exactly what the overlaps are here.

Dave Vellante

>> Gerry, you're living this. I mean, Jensen has said that he, the visionary, the guy who sees around corners, underestimated demand by 100X of computing demand. By 100X this past year.

Gerry Bavaro

>> Yes.

Dave Vellante

>> So I'm curious as to how you're dealing with that, how Snowflake affects that.

Gerry Bavaro

>> Yeah. Look, I mentioned this earlier, the golden thread of any agency business in marketing or media and customer experience is, you have to understand from the data what the insights are pointing you towards. We're in this world of AI where we like to say, it's becoming a bit of a garbage in garbage out world because AI is great and it needs lots of data, but the quality of the data matters. And it's not just about one type of data. With AI it's completely unstructured data. So customer data has IDs and attributes. These individuals buy these products, but what we're using Cortex for is the unstructured data. These are things about motivations, culture. Context, why did somebody buy? And we see that's what's driving up the compute. That's what's driving up the value of the compute, asking questions of the data. An example for an auto client that we have is, using Cortex, we can ask the data, "Show me all of the customers that have more than two children in the home, are high net worth. But with other data, have they engaged with content about large SUVs? And then with other data, do they own an older vehicle that's a small vehicle?" That is a perfect audience. The speed at which that list of individuals in the United States can be compared to the normal, these are high net worth individuals, maybe they'll buy, is so fast. And that's a valuable query. But then we're taking Cortex AI and we're utilizing it to write intelligent queries to visualize that. Business leaders can't get into that data and understand the, "So what?" So visualizing that we're seeing that AI is really about taking that and building a view of that data so you can get to the, "So what?"

Dave Vellante

>> So it's interesting because we've been pretty good at what happened, why it happened? It sounds like we're entering the era of, what are they going to do next? What's going to happen next, and what's the next best action for us to take? Is that something that we're moving towards? Prediction?

Artin Avanes

>> Yeah, I think it's all about recommendations, predictions. And with AI and with the large language models, you learn so much about the data that is at your disposal. You learn so much about the users, how they interact with the system. I think the future will be, not just returning the result sets that you're looking for, but really revolutionizing and helping with entire workflows. Because it's one thing to ask about a particular question, but then typically, it unfolds or there are many other additional questions that are being unfolded, and you as a user would like to have an easier way to drive through this forest of questions. So I think that will be the future experience, how it will look like. At least that's my point of view. I don't know-

Gerry Bavaro

>> Agentic, yeah. Human-in-the-loop.

Artin Avanes

>> Yeah.

Gerry Bavaro

>> What I just walked through of that example of an automobile manufacturer, that literally you could map out the agents from start to finish that would need to drive that entire process. Humans though will always be in the loop and guiding, but we're seeing it's freeing up our best and most strategic minds to really think about the creative, think about the business problems and not be spending a lot of time. We're saving a tremendous amount of efficiencies and cost utilizing AI and utilizing agentic AI in particular. But it is about getting to insights quicker and getting to action quicker.

Rebecca Knight

>> So in terms of those insights and actions, and you described the data about how many kids does a couple have and what kind of cars do they drive. But what about that softer stuff, that behavioral stuff, the contextual and culture stuff? How are you deriving insights from the data or how do you think you could in the future sense?

Gerry Bavaro

>> So right now we're utilizing what's called semantic matching. So AI gives you an ability to understand the unstructured data in a way where, I mean, you could point it at an annual report. You could point it at what individuals are moving across as websites they're visiting or videos they're watching. It's not identified data, it's not your traditional database marketing type data. But semantic matching is if you have attributes about an individual and you have attributes that don't connect to the individual, but it's the behaviors or it's the cultural attributes. When you semantically match, semantic meaning the words, the information, what it's about, the themes. That to me is the future. In marketing, that could be an ultimate prediction machine because now everything doesn't have to reside against the 20 million customers sitting in the database when you overlay the context, the culture, the engagement information, that's where marketing is going, and that's the beauty of AI. It opens up an entirely new paradigm of insights and action.

Dave Vellante

>> How'd you do all this before Snowflake? Paint a picture of your journey. What was the before and after? Your brain, your brain on Snowflake?

Gerry Bavaro

>> Oh man, it's a great question. A lot of files back and forth and less power in the ability in analytics to get down into that data that is the, we call it the four C's. Context, culture, commerce and customer as well. It was hard. So we would have a model of trying to get the data and bring it in, and key it or connect it to some single unifying ID. Now what we're doing is, with native apps and Snowflake, we're bringing our IDs to the Snowflake environment. Bringing that processing and bringing our data into that. And then bringing the unstructured data in as well, so Snowflake Data Clean Rooms really help us with that. Because privacy and regulatory, we strip out all the PII, we're holding that. But inside the Clean Room, legal teams, privacy teams love it because it's like an ultimate sandbox of all this structured and unstructured data. That's what Snowflake has brought to the table, is very portable, very fast and very secure.

Dave Vellante

>> So you're bringing the context into-

Gerry Bavaro

>> We're bringing our consumer data into the Snowflake environment. We're bringing our IDs on USPOP, our Master Merkury ID into the Snowflake environment, and we're bringing the unstructured data in as well. And then using AI in that environment to do the work. Before it was a bunch of feeds and we said, "Can we please have that data? We need it on our side." And our partners and clients would say, "We don't want it to move."

Dave Vellante

>> And so that brings the fifth C, Cash.

Rebecca Knight

>> It's coming. So how do you balance innovation with compliance? As you said, one of the reasons you need this is that it is secure and you feel confident in it, but how do you balance that need to explore and innovate and experiment with AI, but while also making sure there are proper guardrails?

Gerry Bavaro

>> You want to hit that?

Artin Avanes

>> You want to start or...

Gerry Bavaro

>> No, you go. Go ahead.

Artin Avanes

>> Yeah. So I think it's never an either or. As I was saying before, the first question that our customers are all asking is, "I want to have access to the latest and greatest AI products you have." We heard this conference, Snowflake Intelligence, and everyone wants to get immediately started. The second question though is always, wait a moment. How do I should think about the security and governance controls? And that's where you might have heard it. It's being said over and over again, "A strong AI story or modern AI applications won't exist with a strong data layer and strong foundation with the right security and governance controls, right where the data is on the data itself." So it's not an either or. We have to innovate on both aspects in parallel. And that's why I would say there was an overload of announcements because there was a lot of work on the security and governance side that we have done. We just heard about data clean rooms. It's an example of how we have evolved our data sharing story to ensure privacy-preserving collaboration. And that's another example many years ago I would say, or a couple of years ago, where we innovated with data sharing, became the data cloud, but also innovated in the privacy and privacy-preserving collaboration. And you will see a very similar pattern with AI as well.

Dave Vellante

>> You have a pretty highly differentiated platform. I think anybody would agree with that. As the world goes toward open table formats. And Gerry, I don't know if that's something that you guys are exploring, so that may or may not be relevant to you, but how do you think about that differentiation? Is that a threat to the differentiation or is it an advantage?

Artin Avanes

>> There were a lot of debates about this many years ago, even internally. I think at the end of the day, it is, in my opinion, and I mean it when I say it, a strong differentiation. Because we heard words like portable, being interoperable. All these are attributes that our customers are asking for. So we want and we embrace the open of table formats. And obviously we believe, and we think Iceberg has won that battle now, and we have been one of the very first ones who embraced Iceberg. So we are putting a lot of investments and building an Iceberg-first experience. So if you want to have your data outside of Snowflake, you can do it and you will not compromise on either performance or on security governance. So it will, and there's no doubt about, it will be a differentiator for us moving forward. At least I'm convinced about it.

Gerry Bavaro

>> Yeah, I agree with that. We talk about interoperable all the time. The one consistent is going to be change that the client has in the technologies they use, the types of data that they need to utilize as well. And that's why we love Snowflake. It's just so flexible for us. So using Cortex, we have our own proprietary AI models, but then we're utilizing others. I presented a slide a few months ago where I explained the Snowflake app marketplace like an iPhone. And everyone said, "That's really cool. Wait a minute." There's all these apps, it's a composable ecosystem for what you do with data. And that open and interoperable model is very important because clients don't like lock-in. They really don't like that. They're going to come to the table, they might be working with Adobe. The next year they might be using Salesforce. They have a mixed world of technologies and data from various partners in the marketing and advertising world especially. And then you got Google, you have Meta, you got Amazon, you add all those partners and you have to have something interoperable and flexible.

Artin Avanes

>> Give customers choice.

Dave Vellante

>> I was going to say. That's the operative word.

Artin Avanes

>> And then let the best engine win. As simple as that.

Dave Vellante

>> Well, this means you have to have the best engine as things move open, as things move up to the governance catalog. And so that's what a lot of the platform enhancements are about this week.

Artin Avanes

>> Yeah, absolutely. You mentioned, we have a very differentiating platform. And I think I see that always as the next challenge, and meeting where our customers are. If that is what our customers want, we give them the choice. And as I said, we want to ensure the same Snowflake experience. It cannot be that the experience on Iceberg will be different than on regular Snowflake data that sits inside Snowflake. Because our philosophy is, if we embrace these new open formats, whether it's the data formats or the catalogs, we want to ensure the same exact experience.

Rebecca Knight

>> Give them what they want.

Dave Vellante

>> So Gerry, you guys are using, I think, based on the notes I have, Snowflake's AI features, Cortex. You mentioned that. I think maybe document understanding as well. So you talked about essentially what Snowflake has enabled and the problems that it solved for you, how it's made your life easier. When we are here a year from now, what problem do you want Artin to solve that's not solved today?

Gerry Bavaro

>> What problem to solve a year from now? Okay. I think number one is how we can build a composable platform for clients utilizing Snowflake apps. Not just our apps, but all of the other partner apps. Because what's happening is the MarTech stack of the future is not a future two or three years away. It's right now. When we get RFPs it's, "Should we have a CDP or should we make this about the Snowflake environment with the connectors needed?" So I know there's a lot of talk about composable CDPs, but we like the composable data platform. The composable data platform is one where the business cases, the use cases are what drives what you have as a Snowflake-based core with native apps that you bring to the table, and they can change. They can definitely change, but that flexibility is needed. So a year from now, I don't know if it's an ask of you for a year from now, the things that you're doing. I think you're doing the things needed. I think it's more of a mindset shift. I'd want to sit here a year from now, and I have clients saying, "Look, we've tried out CDPs, we have our ESP, we have our media agency that's getting our first party data and doing really cool stuff. But the main thing that happened in our organization is we finally made sense of our data. We finally enhanced it with Identify, with Experian and other attributes and Merkle attributes. But we've gone to this world of now making the unstructured data, pointing AI at it where we're closer to customers than ever before. And the business value that's coming out of that is huge." I think we're still in that age of everyone's testing things out, everyone's talking AI, but has it really driven business results? Has it driven the sale of the automobiles off the lot that a client wants to see? I think a lot of C-level clients are having a hard time tying all of these things that are happening and the news cycle of, "Oh my God, and I got to do this, right? What do I do?" To real business outcomes. So I'd want to sit here a year from now and tie all the techie stuff to very clear business outcomes that A CEO and a CMO says, "This thing drove a tremendous amount of value and transformation for the order."

Rebecca Knight

>> You sold a lot more SUVs.

Gerry Bavaro

>> I'm running with that example.

Dave Vellante

>> Great answer.

Artin Avanes

>> I love that. I took a mental note on this. Awesome.

Dave Vellante

>> Very good.

Rebecca Knight

>> Artin and Gerry, thank you both so much for a great conversation.

Dave Vellante

>> Thank you.

Artin Avanes

>> Again, thank you for having us, okay.

Dave Vellante

>> Excellent. Thank you. Appreciate it.

Artin Avanes

>> Thank you.

Gerry Bavaro

>> Appreciate it.

Rebecca Knight

>> I'm Rebecca Knight. For Dave Vellante, stay tuned for more of theCUBE's live coverage of the Snowflake Summit. You're watching theCUBE, the leader in enterprise news and analysis.