← Marketecture: Get Smart. Fast.
Marketecture: Get Smart. Fast.
Episode 131: Erin Foxworthy from Snowflake on how brands and agencies are building modern data stacks
00:44:46
Loading summary
Transcript113 lines
- [00:01]
Zeta Representative
A word from our sponsors. Today's marketing leaders need more than dashboards. They need answers. That's why Zeta built Zeta Answers. It replaces guesswork with strategic control so you know where to spend and how to grow. It gives you customer foresight, using predictive AI to target intent before it happens. It delivers full funnel clarity, tying media to real outcomes online and off. It puts on demand intelligence in your team's hand.
- [00:29]
Eric Franchi
No.
- [00:30]
Zeta Representative
No dev tickets, no delays. And with live campaign insights, you can optimize while campaigns are still flying. Zeta Answers doesn't just measure outcomes, it drives them. See it in action@zetacmo.AI answers. Again, that's Z E T A C M O AI answers.
- [01:07]
Ari Paparo
All right, welcome to the Market podcast. This is Ari Paparo, I'm joined by Eric Franchi, and our guest this week is Aaron Foxworthy, who is the global head of marketing and advertising at Snowflake. I find Snowflake to be a fascinating company making sort of a jump from a commodity to database company into something much larger, an actual platform that people build on. And I had a chance to talk to Erin and can. I was really fascinated by what they're up to, so I thought she'd be a great person to bring here to kind of give us the DMP CDP data warehouse state of activation kind of point of view. Eric, what's your perspective on this, on Snowflake in this general subject?
- [01:50]
Eric Franchi
Yeah, I mean, Snowflake is a super important company. There's, like so many companies that they're talking about their Snowflake instance and they're, you know, available in the Snowflake app store or what they're. Whatever their. Their version of it is. So it's like one of these companies that loom. Looms large over marketing and advertising, but isn't necessarily like a media company. So it's super fascinating. And then Erin's an og. She's been at this, you know, really long time and has a fantastic and interesting background coming from, like, entertainment. Then she was part of the, you know, golden day of Microsoft and advertising and has agency experience. She brings, I think, like, really, really good perspective on a topic that's, like, somewhat complicated, right? Like data.
- [02:31]
Ari Paparo
Yeah, it is. It's very complicated. And also worth noting that Dennis Buckheim, who probably many people know and has, I think he's been on this podcast before, he was also just hired by Snowflake to be the, effectively Aaron's counterpart that sells to publishers and the cell. Oh, wow. Yeah. So they're really staffing up some heavy hitters yeah.
- [02:50]
Eric Franchi
Yeah. Dennis was the former head of the IAB Tech Lab.
- [02:53]
Ari Paparo
I think he was. And before that he had a senior position in Microsoft.
- [02:57]
Eric Franchi
Yeah. Awesome. Congrats to him. This is going to be a great one.
- [03:00]
Ari Paparo
Yeah, it should be a really good one. So I'd also recommend listening to this Monday's podcast with Ramsey McGrory, who's the new president of Prisma within MediaOcean. He tells us about how the innovative integration is working, a bunch of other interesting topics. So with that said, let's dive right in with Aaron Foxworthy. All right, we're on with Aaron Foxworthy, the global head of marketing and advertising for Snowflake. Aaron, thank you so much for being here.
- [03:25]
Aaron Foxworthy
Thanks for having me.
- [03:26]
Ari Paparo
So you and I sat next to each other at dinner at Cannes. And so for all the PR folks who constantly email me to get on the show, that's the secret, I guess. Just get the right seat at dinner.
- [03:38]
Aaron Foxworthy
Pick the right seat.
- [03:41]
Ari Paparo
It was really interesting. There are a bunch of Snowflake people at this dinner, and I've always been really interested in this company because we did a massive migration at Beeswax to move onto it, and it seemed like there was more than just a database, more than just a platform, but something a little more interesting for the advertising world. So we wanted to talk to you about what. What you're up to and what Snowflake's up to and why. Why it's more. It's interesting, I guess, would be the real question. So let's just start with why you decided to work at Snowflake. You had this background at Horizon Media and Advertising, and now you're. You went hard tech. How'd that happen?
- [04:16]
Aaron Foxworthy
It's an interesting story. So early in my career, I started on the media side. So agency girl, kind of starting right out of college, and digital actually right out the gate. So I won't date myself, but that was a long time ago when budgets were like 1% of the media plan. And I live in Los Angeles, so I ended up working a lot in the entertainment space, media and gaming, where it was a lot of kind of what we called early days, like spray and pray marketing. Right. We would just do huge reach buys all the time. And I worked in that field for quite a while in Los Angeles. I got recruited to Microsoft, so I went into a very similar role I have now. It's called category development, which was a little bit of deal like sales support, a little bit of product support, and then a lot of also product marketing. So you kind of Sat as this semi expert across all three and helped kind of drive the business forward. So this is when the Equanif acquisition was coming in. So this was adcn, this was wrapped. This was when you started to see Microsoft kind of think about having the pieces of really kind of becoming something amazing. Scott Howe was there. So really kind of amazing time to be at Microsoft. Obviously the rise of Bing changed a little bit of the direction that Microsoft was at the time. And so I had to make a decision of what I want to do next. And I told myself I would never go back agency side because I live that world. That is a tough world and I didn't want to do it. But I got this amazing opportunity at Horizon Media as an independent to basically start an incubator in their Los Angeles office or a big client. They just won. And the premise was to take an entertainment client. It was a brand new client, it was a brand new company. It was called STX Entertainment. No legacy data, no infrastructure. And to think about how to be more data driven for a typical like company that didn't have a lot of data. And so it was a really interesting journey because we were coming in through the paid ad space. But what we found in a lot of our testing between starting with third party data then going to second party data, that every single, no matter what platform we were doing any type of data driven targeting and media, first party was always the key. Whether it was look liked, it was a seed, not a surprise. But we didn't have a lot right. So we were striking deals, second party deals. And what we realized is that what we're going to have to do is start to think about building more first party data sets. So that's natural for a retailer, natural for a company that's first party rich. Right. Not necessarily natural for companies that don't have it. So that got us into the world of the dmp. So we were really early customers in the DMP space because we were taking cookies and segmenting them and driving them in the ecosystem. But that got shut down from the walled garden. So all of a sudden our DMP wasn't helpful anymore and we needed to migrate. And the problem was so I was responsible for that entire my team that DMP review. None of the tag management of the data we had for two and a half years was portable.
- [07:06]
Ari Paparo
Right. Because it's all web based and cookie based.
- [07:08]
Aaron Foxworthy
Right. There was nothing durable about it. So obviously the portability clause that we had in the contract didn't even matter. So we had to rethink everything. And I remember, I'll never forget it. I actually almost got fired because we had lost this company's entire segmentation first party data strategy when we wanted to move to a different technology. So luckily it wasn't. And I sat down with some of the team I had, amazing team at the time, some data engineers and data scientists. And they said, you know what, let's think about this different. What if we think about having identity stored in cloud storage, bringing it into a data warehouse, and then giving us the flexibility to never have this happen again. Right? No lock in. We can take it in, build it into a durable identity. It's our segmentation, our audiences, our bi tools. And so that's what we did, right? And so they came in and said, oh, we're in this really cool warehouse called Snowflake. It's like blowing up. It's super fast. It's separate to storage and compute, right? This is where everything's going. And I said, okay, I trusted them. I'm like, let's do it, let's find out. And so we heard all these awesome benefits of what Snowflake was doing and they were already in the ad tech space. Like some of the early customers were in this space. So we started to build that. But then what was happening is we were also testing AMC and adh, nfa, we're testing all the clean rooms. And then when I saw Snowflake release with Disney and nbcu, a clean room infrastructure, I said, oh, it has all our first party data. It's already segmenting and now it has the ability to have secure collaboration. Like this is, this, this is going somewhere. And so I, that's when I, this was three and a half years ago and I was, perfect time, perfect moment. There was a position open to kind of call on agencies in the buy side.
- [08:44]
Ari Paparo
Let me run you through my narrative of this evolution and you tell me if I'm right or wrong. And we had Matt Cats on this a while back to talk about mparticle and their evolution. So I want to hear your version, which is you start by saying, hey, I want to collect all this data. And so you have DMPs. And DMPs like we just discussed are effectively cookie stores. And even for privacy reasons, DMPS actually would say, you cannot store any customer data in our system. It's just anonymous cookies. And that was the wrong bet. And then consumers step two is consumers say, well, I need a cdp. A CDP is effectively like a dmp, except it uses actual users instead of cookies. So you have all these identifiers and the CDP offers a bunch of services like bi, but also privacy, a bunch of other stuff. And then some folks say, and not everyone, I mean CDP is a giant category still, some folks say, well, I want the source of truth to be a data warehouse I control and I'll rent out the parts of it that I don't do, like the privacy controls, etc. And that's what people called a composable CDP. Am I, is that a good summary?
- [09:50]
Aaron Foxworthy
Yeah, I think that's, that's right. I think what's happening is that advanced marketers and agencies are realizing that owning that data foundation is the key and that SaaS can come to it, you don't have to move it and then you don't get locked because the black box, the lock, like all the words we use to kind of describe happens on the Martech side as much as it happens on the out tech side. So I think that if you can own your key customer data, right, whatever that data spine, whatever that core infrastructure is, and now what happens is you can pick breast to beat applications to come to that. We call it data gravity. Some people call it composable. There's all these ways to think about it. But what I don't like about the word composable, at least from a Snowflake perspective, is that someone says, oh, I connect to Snowflake, so I'm composable. And in our world, the SaaS application actually lives in your governed environment. It comes into Snowflake and sits on top of your data so that it's fully composable. So it's, it's an application that lives within the governed bounds. We call it a connected or a native application that sits within Snowflake, so it's actually within your governed environment versus sending a copy of the data to a platform that now is doing, you know, whatever it is they do within the application segmentation downstream.
- [10:56]
Ari Paparo
So speaking beyond Snowflake this, you're bringing up kind of another really big trend, which is this idea of non moving data, or maybe there's a better technical term for that. But the idea being like in an old world, if you want to use platform X, your first step is to take all your data and push it to platform X. And that goes for everything, goes for every, every segment on the lumascape there's a new trend which Snowflake is a part of it. But we don't want this to be a Snowflake commercial. We're trying to like talk in general which is, you know, leave your data where it is and bring the app to you. So what are you seeing there?
- [11:30]
Aaron Foxworthy
I mean that's in the three and a half years I've been here. It is, it's the biggest trend that we're seeing. I mean it's a constant conversation of I need to rethink my entire Martech and at tech stack and I have a data set that's sitting in a plot in a data platform that's super powerful performant, it can house all this data at scale. The last thing I want to do. And there's a couple ways to think about it. When you get into the data engineering or gets it's efficiency, I don't want to maintain a hundred APIs and every time I it's a copy of data and it's expensive. Egress is expensive. Right. Data going in and out can cost money. So how do I do it from an efficiency perspective? The other piece too is what you just said is when you get into compliance in the security office, right? Especially in regulated industries, this is where they're like I am not allowing you to take that customer data anymore and ship it off to right. A SaaS application, an identity provider, they're going to cut it up and then they're going to send it out to the endpoint like that. We're not moving it anymore. And so there's two things that Snowflake helps with or cloud in general kind of helps with. Is the first one is that idea is that that SaaS is an application, it's built on a cloud infrastructure. It can now live on top of your dataset again in your boundaries of security. The other part is what you hear a lot is Snowflake just calls it a data share. So in cloud world with collaboration that just means you're basically exposing a view of a table to another party. So it's access to actual view where they can query it, right. But actually not take possession of it in any way. So they can basically query the data. It's live, it's cross cloud, it's real time. So as the data changes what they're querying, they see those changes automatically, but it builds into the table, all the governance. So if I said you can't see this column, you can't see this row, you can't see this, I'm going to mask this specific object. We have all these interesting controls in the table so that when they're querying it, they can't see what you said they're not allowed to see. So it's a really efficient way of collaboration without actually making a physical copy. So that's where you hear a lot of this like non movement of data. It's that idea of applications coming. But it's also like when you do collaborate, there's a way to do it without actually fully making a copy and losing kind of that governance of control.
- [13:30]
Ari Paparo
If I'm looking to activate my data like through a DSP like the Trade Desk or Google, et cetera, is it possible to do that without moving the data? What's the current state of play for non moving data? If you were to have like a matrix of here's my use case, here's how far along I am of being able to do it without moving my data. Can you speak to that specifically about activation?
- [13:53]
Aaron Foxworthy
Yeah, I'll give you a couple I can talk about publicly. So this is actually something that the Trade desk worked on us quite a while ago. So we have a first party pipe that comes out. We call it a pipe, it's not technically a pipe to trade desk seat through a datashare. So if you want to actually expose a table to trade desk, they can actually query that data from Snowflake and land that in the seat without taking a full copy in position. So we've actually created a lot of those type of integrations not only with DSPs but also with a lot of SaaS like MarTech applications like CDPs. So we have quite a few AdTech and MarTech partners who will take a data share as a form of activation from Snowflake. If you want extra, I know you're going to ask me about this if you do want extra privacy controls where there's not even a view over exposure. Now we're talking a data clean room.
- [14:37]
Ari Paparo
Yeah. Is a data clean room a thing or is it a feature like it sounds like you just said earlier that you can make a given row available or a column available to a partner. How's that different from having a pet layer in front of a row or column?
- [14:53]
Aaron Foxworthy
So really for us a data cleaner was an advanced data share. So what it has is specific features like aggregation and projection constraints. So basically you can't query under a certain threshold volume. So what we do is on top of that data share, we basically inserted code that gives exceptional like very specific types of ways that prevent certain queries from running to give you protections. But that's what's interesting about data cleaner rooms. We call it a product, right? People say I have a data cleaner room for a Snowflake, a data cleaner room is literally right. Just extra features as part of that data share. Now we bought a company, Samuha, as you know, that allows for a UI layer to make it easier to use, but underneath the hood really it's Snowflake data sharing with extra pet enhancements. Right. That's the difference between a data share really and a data clean room.
- [15:38]
Ari Paparo
And back to that trade desk example. So the client's not moving their data, the trade desk gets a view into the data and probably they're copying it or matching it with their IDs so that they can run it efficiently. They're not doing like a real time lookup that would probably not make sense. So there has to be some sort of translation going on.
- [15:58]
Aaron Foxworthy
Yeah, there's a translation. It's obviously it's a HEM and a made. Right. So there's an identifier within the view that's allowed that to make the match.
- [16:04]
Ari Paparo
And do the customers have to know how that works or. I'm just very interested in what parts they have to compose and which parts other people compose. If I'm a first party data owner and I have all a bunch of data in Snowflake or another data warehouse, do I also have to rent some sort of identity spine or can my activation vendors do that work for me?
- [16:24]
Aaron Foxworthy
So it depends, right? If you have a first party data set where you have those identifiers, obviously depending on what the end platform takes, some of them say I'll take a hem, I'll take a maid, I'll take an IP address. So you're deciding that other times, obviously if you don't have those identifiers and you do enrich, that's exactly, that's where we get into kind of this concept of right identity resolution. But here's what's interesting. So go back to the Data Gravity Applications and you kind of heard this announcement with us with Axiom. So whether it's Axiom or Merkle or Liveramp or Transunion or Name a lot of our identity providers, they've built applications to come on top of that first party, resolve the data in your environment so you don't have to make copies and then activate. So if you do need it, the nice part is that they're, they're bringing that resolution now at least to your first party data set versus you sending it out right to their platforms. They're cutting it up, they're resolving and then they're activating from kind of their infrastructure.
- [17:15]
Ari Paparo
Yeah, that was going to be my Next kind of area to ask you about which is what, what is the agency's role here? What, what data are they owning and how is that changing over time?
- [17:26]
Aaron Foxworthy
So in the world, as I obviously work with a lot of agencies, there's a lot of different takes here on the, the big holding companies, a lot of them own their own identity assets so they're able to have their own first party data, enrich it with their clients, right, in privacy, safe ways and kind of build audience and segmentations. And that's kind of feeding into these large platforms they've built, right? So interact from an IPG example or Blue, which is Horizon's platform as another example. So they'll have identity kind of baked into, so they have graphs, right. They're baked into kind of powering the structure, right, of, of their, their platforms a lot of times that are built on Snowflake, so they have their own data assets. There are some smaller agencies though that are say, I'm never going to actually own your first party. What I'm going to do is I'm going to work out of your ad seats and I'm going to take the data, right? And I'm going to, out of, out of the exhaust that comes off those ad platforms and I'm going to build audience segmentation and think about planning in aggregate that I'm actually never actually testing your first party data. So I see kind of different routes, right? Some agencies saying, I don't want it, don't ever give it to me. I can, I can make decisions without you handing me the first party data. And then I have others, right, that have the infrastructure to actually create their own first party data assets. So different kind of ways of thinking about it, right.
- [18:41]
Ari Paparo
They're right in the middle of this question about whether you move data or not. And they're, and they're also a data owner in many cases. So it gets pretty confusing. All right, so we have to talk about AI. We should get some theme music whenever we're ready to bring up AI. So I keep hearing from folks saying things like, well, everyone's got AI, but we have the data, so that's what matters. How are you seeing your customers think about the data in a data warehouse as it relates to AI and what they're bringing to the table?
- [19:12]
Aaron Foxworthy
Yeah, this is interesting. It's, it's happening really fast. But I would say people are really just toe tapping at the beginning, which is, you know, not a surprise. I think where a lot of agencies started in AI was Google has a new Gen AI feature and platform. I'm going to use it, right. I can a lot of it happen on the creative side in more of the data side. It takes a lot more effort, frankly. And I think that. So what you're seeing is, is that the agencies that have kind of worked on building that data foundation we talked about. So if you're bringing all that data in, right, you've created, right, a schema that allows you organize the data and then you've created a semantic lay. So just think of a semantic layer as like a business layer on top of the data that allows you to think about what that data is. Now you can start doing AI on top of that work. And so there's a few agencies that were. That had that organization of the data, right? They, there's this one we talked publicly about, a small one called Power Digital. So they actually had their E Comm agency out of San Diego. They had, you know, hundreds of clients, mostly D2C. And they are bringing in that data into Snowflake, right, and organizing it across all those buys. And they were able to quickly start to build a on top, where they built a chat agent that was able to come in and say a marketer could come in and actually ask questions of how's my meta campaign doing today? What does this benchmark look like? What are the ads that are performing today? So instead of going in a tableau report, right. Their marketers can now actually ask really intelligent questions. And the nice part about how I think Snowflake thinks about AI is that we bring the models inside your governed infrastructure. So what's nice is that you're not making a copy again of data out to an LLM. The LLM has to come to you and can only train on the data that you have inside of Snowflake. So again, we have that governance feature and then you get a picket so you can actually see like what's the right LLM to use, what's going to give me the most accuracy, what's the cost implication of that specific LLM. So we built like a full service orchestration to help you decide on how you use AI and AI layers into your data sets.
- [21:09]
Eric Franchi
You have a CEO that is a former head of ad product for Google, Sridhar Ramaswamy. I was curious to see since he's been in the seat, it's probably been about three years now, right?
- [21:20]
Aaron Foxworthy
Oh, he's been there for three years, but I think CEO for like a little over a year.
- [21:25]
Eric Franchi
Got it. Have you seen any evolution in how you work with marketers and the media ecosystem, just with his influence.
- [21:33]
Aaron Foxworthy
You know, I think that the interesting thing about coming to Snowflake is that, you know, media and advertising in a cloud organization is one piece of a huge narrative. Right. Like what we do within retail and financial services and all these other verticals is deep and wide. And so when you think about marketing and media and Attic, it's important, but it's definitely not the only part of his day job.
- [21:55]
Eric Franchi
Now he's saying marketing isn't the most important part.
- [21:58]
Aaron Foxworthy
Right.
- [21:59]
Eric Franchi
People are at the center of the world.
- [22:00]
Aaron Foxworthy
And it's so interesting being a cloud infrastructure because it's definitely not all we think, but to that point, yes. Right. Having him, obviously the company and understanding kind of. I think the challenge is that it's an ecosystem. Right. This is a buy side, sell side ecosystem. And not everyone understands that. Right. So what we do on one side affects the other and vice versa. And I think that his understanding has helped us think about product integrations, what we purchase, how we think about organizational structure to solve those challenges. So that's been very helpful, definitely at senior leadership that they understand the complexity of the ecosystem.
- [22:32]
Ari Paparo
Yeah, I actually overlap with Sridhar quite a bit at Google. They used to like have me and the other doubleclick people, like, try to give him advice about, like, stuff. And he was very polite about not taking it. So I appreciate that. So. All right, well, Aaron, thank you so much for this. We're going to take a break and we're going to come back with our refresh news of the week. This podcast is brought to you by audiohook, the leading independent audio dsp. Audio Hook has direct publisher integrations into all major podcast and streaming radio platforms for providing 40% more inventory than what could be accessed in omnichannel DSPs. What's more, audiobook has full transcripts on more than 90% of all podcast inventory, enabling advanced contextual targeting and brand suitability. Audiohook is so confident that in addition to CPM buys, they offer the industry's only pay for performance option, where brands can scale audio and podcasting with peace of mind, knowing they are only paying for outcomes. Visit audiohook.com to learn more. That's audiohook.com.
- [23:39]
Eric Franchi
All right, we are back everybody, with the refresh on the docket for this week. What do we have? Some big news out of TTV, a brand new RMN. I think this is RMN number 426 or 427. Some AI news, actually, big breaking news raised out of Substack, maybe a couple other things. We have some time. Let's talk about ttd. But before that, actually we have a correction. Ari, you want to take it?
- [24:04]
Ari Paparo
Yeah, just a quick. We get things wrong here a lot and we try to be very, very aggressive about. Well, I get things wrong. You do your research. So last week, minor flub. We talked about Kathryn Perloff Scoop about addict staffing up to help their SSP business. And we attributed it to Adweek where she used to work. But she has moved on. She's now at the information, so she's doing great work there. So thanks, Catherine. Sorry for the misattribution.
- [24:29]
Eric Franchi
Yes, thank you. And Ari is sorry. Okay, let's talk about TTD. So earlier this week I was on LinkedIn and I saw a post from Jeff Green announcing that the trade desk has been added to the S&P 500 and the stock shot up on the news. Why does the stock shoot up on the news of being added to The S&P 500? One large reason, one major reason. Because once you're added to the S&P 500, which is the most popular index across 401ks and all these passive vehicles, funds need to buy it. So it is kind of front running the addition, which happens, I think later this week with people kind of arbing a trade. So that's why the stock went up. It is now in the low 30s or mid 30s, I'm sorry, mid 80s versus being in the mid 70s last week. So that's a great thing for TTD holders. But I think it's like pretty cool that there's a pure play ad tech stock in the S&P 500. Like sign of legitimacy for the industry. I mean, definitely sign of, I think, legitimacy for the company. And it's like an awesome thing. I should also note that Snowflake is and has been an S&P 500 company for some time. But this is awesome. Congrats to ttd. What do you think, Art?
- [25:51]
Ari Paparo
Yeah, my only concern is why they weren't added to the SP 500 plus, which is a totally different product offered by trade desk.
- [25:59]
Eric Franchi
So it was written, it was foretold, it was foretold.
- [26:02]
Ari Paparo
S&P 500 plus. You know, one day Jeff Green can hope to reach the heights of Applovin. You know, I think that he's got a goal.
- [26:09]
Eric Franchi
Yeah, for sure. In the meantime, congrats to Jeff. This is also, I think, you know, there's like an interesting maturing going on right now at ttd. There's A thing that he posted this week was that a couple of longtime execs are exiting. And he said it was like very intentional. We're saying about it two trade desk colleagues who've worked there for a decade now, the Chief Commercial Officer or the Chief Client Officer and the VP of Customer Success. So it's just kind of interesting if you take a step back. Right. Like inclusion to the s and P500. Longtime execs leaving feels like some sort of changes in the air at TTD and largely positive. So congrats to them.
- [26:53]
Ari Paparo
Yeah, I think they're consistently the most interesting company in ad tech as well. They're. You can second guess a lot of their moves, but they tend to have a good result. So it's hard to second guess ultimately. But you say they should do more M and A, they should expand beyond open web advertising. They should stop with the TV operating system. They should. Blah, blah, blah, blah, blah. And, you know, only time will tell.
- [27:15]
Aaron Foxworthy
Yeah.
- [27:16]
Eric Franchi
Okay, let's talk RMNs who had HP on their bingo card for launching an RMN. Aaron, what'd you peg the odds at prior to this week?
- [27:26]
Aaron Foxworthy
Yeah, it wasn't me. Definitely low odds on that one.
- [27:29]
Eric Franchi
What about you, Ari?
- [27:30]
Ari Paparo
I mean, HP tends to be pretty aggressive with their ads, their data use, their terms of service and all this. There's two HPs, right? This is the one with the printers, right?
- [27:42]
Eric Franchi
Yeah.
- [27:43]
Ari Paparo
Okay. Just want to make sure.
- [27:44]
Eric Franchi
Also the number two PC maker.
- [27:46]
Ari Paparo
Yeah. Right. So I'm not that surprised, honestly. I mean, they've been. When you get a new printer, you have to sign in your life away just to get ink. I don't know how effective the actual ads will be. I'm going to take us off topic for one second. Have you watched this season's Black Mirror?
- [28:01]
Eric Franchi
No, I've been told that I must.
- [28:03]
Ari Paparo
I watched one last night. It freaked me out. There's this episode with Rashida Jones where a corporation takes over parts of her brain and starts having her speak ads to everyone she talks to. And, you know, everyone in ad tech should watch that one. It can't be worse than hp.
- [28:18]
Eric Franchi
All right, it's on the list. Why is HP doing this? So, as I said, they're the second largest PC maker, second to Lenovo. So maybe Lenovo is next. Big audience. So global audience of 830 million people across 100 million devices in the US alone, 160 million monthly people across 19 million devices. So they've got the scale. And how this is going to be manifested, according to the news, is Ads are going to appear on HP laptops and in HP apps. Somewhat hearkening back to the days of like you want to say it Ari.
- [28:55]
Ari Paparo
Are you going to say are you talking Gator and things like that? I'm talking when you, when you and gator.
- [29:00]
Eric Franchi
I'm talking gator like something installed on your computer and you're running ads but in this case you know obviously it's the manufacturer themselves and they've right the consent from the user but it's a little bit of back to the future.
- [29:09]
Ari Paparo
Yeah, I mean Microsoft has pushed in this direction multiple times over the last 20 years where they've said well we're going to put an ad here and there on the operating system or in the start menu and stuff like that. I don't think it's ever been a meaningful part of their business so much as the homepage of MSN on people's browser was a huge business. So I'm a little. There isn't a lot of evidence that you can make a pre installed apps advertising, you know, scale and be a really good business.
- [29:36]
Eric Franchi
Aaron, what do you think? Coming from the Microsoft heyday, I was.
- [29:39]
Aaron Foxworthy
Gonna say the same thing that Ari said. We. I remember them trying to actually have inventory within the Windows ecosystem. Obviously they had to be really careful from consumers feedback but it never got traction or scale. So it's gonna be really interesting to see if this does.
- [29:52]
Ari Paparo
I'd be very interested in knowing if you buy an HP laptop if they claim ownership to all the data like are they sniffing every single browser URL, whether or not you've opted in or maybe you've opted in but don't know it. That's where I think a red line could be crossed that I wouldn't appreciate personally and it's sort of analogous to your tv. When you buy the tv, pretty much the TV manufacturers know what you're watching. It just feels very different on a PC that the browser or another piece of software within the PC might also have sort of data leakage to the owner of the of the device.
- [30:29]
Eric Franchi
So Lauren Johnson from adweek Shout out wrote this article. Let me talk about data for a second. HP's ads can be targeted by device data like when someone uses their PC, what apps they use, their email address, purchase history and location according to the pitch deck. For example HP can identify people who have downloaded tax preparation software or someone who spends more than 10 hours a week using conferencing apps.
- [30:53]
Ari Paparo
So eh, it's not that bad. That's not as bad as it could be. It's not your, it's not URL level and it's not the data in the apps.
- [31:01]
Eric Franchi
This is correct. And it's also desktop only, which is, you know, quite limited and not necessarily the fastest growing market per se. That said, apparently this comes from a pitch deck. They're planning on some fast product, but unclear if that's just like more video inventory that runs on the laptop.
- [31:23]
Ari Paparo
Yeah. Can you uninstall it? Can you do a wipe of your device? Are they subsidizing the device and then preventing you from getting kind of the root level access that people expect on a PC? Those are pretty interesting questions.
- [31:35]
Aaron Foxworthy
Also I'm just thinking for like gaming clients that could be really interesting or just a lot of clients who are dependent on desktop.
- [31:42]
Ari Paparo
Yeah. In gaming. I had a company on here called Overwolf which is effectively like widgets on your desktop, like leaderboards and stuff like that that help with PC gaming. And it's actually a really big business funded by a 16Z. So that's an interesting play as well.
- [31:59]
Eric Franchi
It's interesting. Final point on this, it does feel like this is the logical extension of what we're seeing with the CV OEMs all building their ads business and taking advantage of the fact that they've got kind of locked in real estate. This is a PC version of this. So again, does Lenovo do it? Does Apple do this?
- [32:22]
Ari Paparo
No, I don't think so. I don't think so. But there's also an interesting contrast with Chromebooks. So Google is a pretty big manufacturer of or licensor of operating systems because of Chromebooks. But their strategy has been to go low cost students international. And I think it's a data collection play. But they don't monetize it. They don't monetize no ads on the Chromebooks. But maybe an independent Chrome thinks differently about that.
- [32:48]
Eric Franchi
Okay, let's move on to some AI stuff. So a couple of interesting hits here. So According to the IAB who released a report this week, nearly 90% of advertisers will use Genai to build video ads. If you read the marketexture newsletter this Monday, this would not be a surprise to you, right?
- [33:08]
Ari Paparo
90% surprising. I think that, you know, the horses left the barn, as it were. The video models have gotten so good already that it'd be crazy not to at least try them and see what you could do.
- [33:20]
Eric Franchi
Yeah, to be clear, the 90%, they didn't give a timeframe for that. Where they did give a timeframe for basically buyers said that nearly one third of their creative currently is built from scratch or enhanced using Genai and they expect that to jump to nearly 40% again, building from scratch or using Genai for video next year. So it's up to 40% next year.
- [33:42]
Ari Paparo
Yeah. And the point I made in my newsletter wasn't my best newsletter, to be frank, but you know, every Monday it's a lot of work. And I'm interested in your point of view on this, Eric, which was Genai has gotten so good so fast that it makes you think that maybe there's not an opportunity for vendors in this space because you could use the raw power on your own, more or less. And it feels more like the analogy I made was Microsoft Excel where anyone can use Excel for anything and you don't need another, another layer on top of it to make, you know, Excel do the thing you wanted to do. What's your, what's your hot take on my hot take?
- [34:18]
Aaron Foxworthy
Yeah, I agree with that. It's going to be the power of the model. I think it's going to be which models perform the best and give you the best creative. But I also, what I think a lot about is just someone who lived in this world is the traffic and that creative, the trafficking, the ad ops. Like how do you get the ISCI code to the right platform, to the right. Like that's the stuff I'm also interested in like who's wrapping around that piece of the gen AI because it's great to create it, but you got to. Right. Traffic it out to endpoints and get it into production. And I'm kind of, I'm kind of curious more about that than I actually am of the gen AI itself.
- [34:48]
Ari Paparo
ISCI code blast from the past. Right. We're gonna, we're in Ad ID land. And the answer is codes got replaced by ad IDs like 30 years ago.
- [35:02]
Eric Franchi
Yeah. From my perspective, similar to Aaron, I think that if the value prop for a company or a startup is produce creative using gen AI, I think it's going to be very tough given the off the shelf tools. And obviously VO3 is the best example out there. But we're seeing companies and again what Aaron said, what they allow the AI produced creative to be done, how you provision it in a multivariate way, how you deploy it omnichannel, how you then optimize it, how you, you use it for, you know, driving purchase behavior. There's tons of opportunity there. And those are, those are companies that, you know, we're looking at. And some of Them are growing quite fast. So I don't think it's a category that's like doa. And if, you know, you read your psr, you might have gotten that impression. Right. It's like creative is uninvestable. I don't think creative is uninvestable, but it's how it is incorporated into a larger offering.
- [36:06]
Ari Paparo
Yeah, I buy that. I think that creative plus ad network makes sense. Creative plus trafficking maybe makes sense. There are definitely places for it.
- [36:16]
Eric Franchi
Does creative become part of the dsp? That's the real question.
- [36:19]
Ari Paparo
Historically, creative has never been part of a dsp. I don't think there's a single example where a DSP offered creative services and it worked. So I'm going to go with no.
- [36:29]
Eric Franchi
Aaron, I'll take.
- [36:30]
Aaron Foxworthy
I'm going to throw Rebecca a question. Have you seen anyone have concerns around? If I'm using this model within this environment, a YouTube environment, I'm using this model within Meta's environment. I'm using this model in the DSPs that are not all the same models. You want brand consistency, right? It's such an important part of a brand's ip. I'm curious how people are controlling that. Right. So if you don't have one central model that's actually the voice of your creative, how are you controlling for, you know, brand sustainability across all these platforms?
- [36:57]
Ari Paparo
Sounds like a startup idea right there. Call it brand, brand alike or something like that. AI checks your AI creative to make sure it's brand safe.
- [37:06]
Eric Franchi
Yeah, it's, it's, it's interesting. Right? So my answer to Ari is like, will creative become part of the dsp? Is like, well, if you look at what Google is doing with PMAX and what Meta is doing with Advantage plus, they're already connecting creative to media and optimizing it and creating like incredible outcomes. Right? So there's a Check in the 4 column there. But then on the other side, because these are siloed, you risk that brand consistency, you risk having all of the brand guidelines upheld across platforms. Which is why the client base for these products are performance marketers or SMBs. And that's why a lot of people are somewhat bearish outside of this. Hey, is there a way to do this across platforms on these. Set it, forget it. Performance products really being meaningful for brands.
- [37:59]
Ari Paparo
The feature I would build if I was running a DSP would be a sort of a checkbox, very simple option that automatically created an AI generated creative or multiple and did an instant a B test and took 3 to 5% of the traffic and ran it with a creative that you didn't come up with and saw if it worked and gave you visibility into whether it worked or not. Because I don't think you're going to be able to convince people you need to actually just do it and see if it works.
- [38:27]
Eric Franchi
Agreed. That's a great idea. Okay, let's talk about a company that's not an ad tech company that we talk about every week. OpenAI. This will be just a quick one because I think this is like super interesting. So they're working on an integrated checkout experience according to an article this week. So you will ask about a product, it will surface a recommendation, it will allow you to buy that best recommendation while never leaving OpenAI. That feels big, that feels like funnel collapse, that feels like the ability to build a big. You know, to your point, as you said before Ari, many times, affiliate business on these LLMs. Like I think this is something.
- [39:07]
Ari Paparo
It sounds like something I think, you know, OpenAI just. I think we covered four or five months ago on this pod that they put out a new product called Operator which was like a headless browser that would let an agentic agent shop for you. And we looked at that and went, eh, that doesn't sound right. So now, not that they're giving that up, we don't have any information about that, but this gives an example of a different approach which is instead of going out to the web, you bring the web into you and then there'll be other folks who have shopping experiences that are agentic but use MCP to be able to shop your catalog. I think this is a very exciting emerging area. Actually my co founder RAM from Beeswax is at a startup that is doing this sort of thing. MCP ad catalogs from E commerce feeds. Let me know if you want a connection to him.
- [39:56]
Eric Franchi
What's the name of the company?
- [39:57]
Ari Paparo
Man, you know how bad I am with names. Don't put me on the spot with names. I'm an investor. I have no idea what the name is.
- [40:04]
Eric Franchi
Sorry Aaron, I cut you off. Go ahead.
- [40:06]
Aaron Foxworthy
No, it's fine. I'm just invade into payments, right? Then you start thinking about is there a subscription, like a prime, Is there credit card in the. I mean there's all kinds of amazing things that that could start to do.
- [40:15]
Eric Franchi
Yeah, well Shopify is the, is the connection point. So it's integrated with Shopify. And Shopify's obviously got the payment data.
- [40:22]
Ari Paparo
So yeah, goes big octogen, like octagon with a GN at the end.
- [40:27]
Eric Franchi
Very Cool. Nice. One more thing. I saw this this week. We should drop this in the newsletter. It's a really long article, but it's apropos the Cloudflare News where they were basically like firewalling LLMs from seeing publisher content. This substack called Digital Digging, which I had never come across before, had a giant piece about like the six different ways LLMs actually bypass publisher paywalls and dropped a pretty eye opening statistic. Basically OpenAI's ChatGPT perplexity and XAI's Groked successfully access protected content approximately 50% of the time. So like, if you think you're firewalling this stuff off or your paywall protects you, the answer is actually no. So super interesting.
- [41:15]
Ari Paparo
Yeah, I was at the Acquired Live event at Radio City this week in New York. It was fun. The Acquired podcast, for those of you who don't listen to it. And they had, among other guests, Meredith Levin, who's the CEO of New York Times. She gave a pretty impassioned defense about why they're suing these folks over content. And I really, I like the way she put it, which was they're paying unsold sums for compute, they're paying unsold sums for engineers and distribution and all of these other expenses, but they don't want to pay for content. And that's not right. And I thought that was quite sharp observation of hers.
- [41:48]
Eric Franchi
Absolutely. Hey, on the subject of paying for content, this was kind of breaking right before we recorded this. Substack announced that they're raising 100 million from TCG, Andreessen, bunch of other investors. And on Substack, creators can be paid for their content. So I thought this was pretty neat. They're going big.
- [42:07]
Ari Paparo
Yeah. Substack is trying to create a walled garden, one unemployed journalist at a time. And it's going pretty well. All right, well, let's talk about Clinch for one second. So clinch, I had them on the podcast a couple years ago. They're a challenger, I'd call them in the buy side ad server market. Have really nice looking UI that allows you to do dynamic ads and all this other cool stuff. And they are claiming that they're offering a flat SaaS model for buy side ad serving, which is interesting. Like, you know, is it real, is it not? I have a lot of experience trying to convert people to flat pricing and not a lot of success, especially at the agency level. So I think this is one one to watch. Is is this real or is this just a. A upstart competitor trying to shake things up.
- [42:58]
Eric Franchi
We are an investor in Clinch and agreed the product is awesome and they've got a ton of momentum and I'm not, I think, you know, speaking out of turn here or giving anything confidential. It's in the article. We'll drop the link to the Adweek article. You know, apparently this is helping them win new business. Every new business with with brands. They're estimating that SaaS ad serving is saving like 30 to 50% for big brand customers. So it's real savings. And yeah, the real question is, and Ari, obviously given your ad serving background, even Aaron back in the day at Atlas, moving from usage to SaaS has real big implications like you can put more dollars to work for media. So I think it's awesome.
- [43:41]
Ari Paparo
Yeah, Devil's advocate would say, like you could just continue offering variable but offer 30% off. And so offering flat is basically a sales technique to try to offer a discount but but to lock it in. And there are different models. I just wonder if there's anything really unique here or if it's effectively a pricing war.
- [44:04]
Eric Franchi
We should have Oz on the pod.
- [44:06]
Ari Paparo
Yeah, Oz is a fun guy. We should have him on.
- [44:08]
Eric Franchi
Want to call it here?
- [44:09]
Ari Paparo
Yeah, sure. So this was a great conversation. Aaron Foxworthy, the global head of marketing and advertising for Snowflake, thank you so much for joining us.
- [44:17]
Aaron Foxworthy
Thanks for having me.
- [44:18]
Ari Paparo
And Eric, always great to chat.
- [44:21]
Eric Franchi
Always. Thank you, Aaron. And we'll see you next week, everybody. Bye bye. Thank you for subscribing to marketecture. New interviews are added every week at Marketing and your favorite podcasting app.
- [44:41]
Ari Paparo
Thank you for listening to the marketecture podcast. New episodes come out every Friday and an insightful vendor interview is published each Monday. You can subscribe to our library of hundreds of executive interviews at tv. You can also sign up for free for our weekly newsletter with my original strategic insights on the week's news at News Market. And if you're feeling social, we operate a vibrant Slack community that you can apply to join at adtechgod. Com.