Clean Rooms Vs. Containers: What’s The Difference In Advertising?

It's very nice to be in this very, very cold apartment. Give it up FOR AIR CONDITIONING. [laughter] GIVE IT UP GIVE IT UP FOR AIR CONDITIONING. YEAH, I'm unsure that I'm going to be leaving anytime soon. Um so, thank you all for coming. Um my name is Stephanie Lazer. I will be moderating this panel. Um I work at AWS, uh Amazon Web Services. So, um I'm a strategic like sub-industry leader. I work on the strategy for advertising and advertising technology. I've been in ad tech for entirely too long. Um so So, yeah. Um so, thank you guys so much. Um we're This is like the This is like the tech-focused uh panel of the the like a really aggressive tech focus focus panel. So, I'm um excited to moderate it and I'm going to let our panelists go ahead and introduce themselves. So, why don't we go down the line starting with Oleg? Thank you. Um Oleg Kornfeld, I'm a group president, tech and platforms for CMI Media Group. We are healthcare agency under WPP umbrella. I am Daniella Harkins. I'm SVP of go-to-market at LiveRamp. Uh we are a data connectivity platform and uh I am in charge of all of the commercialization of our solutions and products. My name is Ravi Patel. I'm the CEO and co-founder of swim.ai. Um and we are uh you know, a dynamic supply shaping engine uh and decisioning layer. So, it's there to help you stop wasting your money. So, thought I'd throw it out there. That's awesome. I'm Joel, also a fan of not wasting money. Uh [laughter] But, I'm CTO at OpenX and I run product and engineering. All right. Yeah. And we love air conditioning. That's what we've learned [laughter] so far. All right. Perfect. Okay, great. We agree on one thing. So So, one of the one of the focuses uh of this panel is we're going to talk a little bit about containers and we're going to talk a little bit about clean rooms. We're going to talk about all the different ways in which um we uh can like interoperability exists and can um uh can help our industry achieve its uh business outcomes and goals. So, um why don't we start with like what what are the problems that we're like trying to solve right now? You know, I think we like kind of leaned in uh into like what is the technology, but I think if we step back and we look at instead of being product focused, we start to be problem focused. Um let's talk a little bit about that. I can start. Go ahead. Um for us it's always accurate audience reach. Uh reach and frequency of our messaging. We're a media agency. We get hired by marketers to spend their media dollars, a lot of them. Um the way we measure it and our success how well our investments work. Um if we do not fully appreciate and understand the reach of the audiences we need to talk to and specifically in health care it becomes very very nuanced. Um we're wasting money. It's interesting. Um so I said we're a data LiveRamp's a data collaboration platform. And so when I look at it, I think it aligns with what you're saying. I think about it as data availability and accurate data availability. And that can service media, it can service measurement, it can service optimization. I should have said targeting not uh media. Uh but for me it's really about no consumer sorry, no brand is going to know everything that they can about a consumer. And so how can we in a privacy controlled way uh provide more basically provide the technology that allows much more availability of data across the ecosystem to be used across the continuum from media to different marketing use cases and beyond. Uh just to add a few uh pieces to that as well. So, the way I think of it like containers, clean room, right? Uh clean rooms um more of a collaboration and measurement problem. Uh containers is more of an execution and decisioning problem, right? Uh two different pieces of that. So, an easy way that I was thinking about this earlier was one is safe knowing, like clean clean rooms, right? Like putting the raw data in and understanding it and then uh and then the other is safe doing at the moment of impression. How do you kind of uh logic on it? That's really good, Robbie. I like that. Um I think Stephanie, I like where you started the conversation. If we zoom way out, you know, fundamentally, publishers are selling inventory and buyers are trying to reach an audience. And that's where a bunch of technical solutions, clean rooms, and containers, and a number of other things have come into play to facilitate that bridging. So, to me, that's, you know, the heart of what we're trying to accomplish here with a bunch of different technologies. So, Joe, why don't why don't you take this one? Can you Can you walk me through a little bit of like where these different type of technology shine? So, help to kind of define them, give the differences of the of the of the tech. I'm happy to start and then the experts can I'm letting you start and then everybody else join in. Add to it. Argue. I don't know. Come on, guys. [laughter] Make it spicy. I'm against ACs. That wasn't in THE CUE CARDS. WHOA! YEAH. NO, I I THINK IT'S A GREAT QUESTION, RIGHT? It's like clean rooms existed because I think you look at privacy at a number of different places. W- As a user or as a consumer, I don't want my data, you know, spread willy-nilly out there, but I understand that people need to leverage it to to know how to reach me with an ad that's effective. And so, a clean room uh facilitates that connection in a way that somebody on one side of it can get a model that they want based on a bunch of different user attributes in an aggregated fashion. That protects me as a user, right? I don't show up individually, I show up as a as a cohort of sorts. Um I think a container is a way to, you know, as we were talking to people, it's like, "Hey, look, I build a model, but for me, you know, that model is IP. I don't want you knowing about my model." And so, that's where containers come into play. It's you can own your model. It's your executable, and we as a host of it can't access it, but your, you know, IP is protected in the same way that, you know, me as a consumer I'm protected. So, I feel like it offers protection at different levels. And I like what you said about, you know, at the highest decisioning level, it's What did you say? Doing? Safe doing. Safe doing. Yeah. Yeah, so to hit on the containers a little bit. Um the way that I think of it is it's like a unit of decisioning logic, right? Uh that kind of sits at the moment of impression. And it's it's a lot of the way it's evolved is something that we haven't done before. And it's crazy. Programmatic's been around forever. Uh the whole premise of programmatic was the auction mechanics, being able to control it. Uh but what we didn't realize is everything getting in between and how certain pipes are smaller than other pipes in availability, and why it makes so much more sense now. Um and you know, I'm not a clean room expert, but the way I thought of I think of clean rooms as we do work with LiveRamp and do uh do a lot of things there is um it's it's great in terms of uh knowing who you're going after. It's like great for uh you know, understanding who got the exposure uh and where is it going, and then taking that data, publishing it into the containers uh to be able to kind of act on it. Now, I think it's becoming more real-time in nature. Uh it's still hard. In containers, you can't take CRM data and publish your data and match it and do it in real-time. It's still after the fact. Uh but you could still get a little bit more dynamic and real-time in that. Yeah, and you can probably action faster out of a container, obviously, than you can out of a clean room. The other thing that I would say in terms of where things shine, on a clean room, I think there's an opportunity for multi-party collaboration. And so, when you start looking at a brand, a publisher, and potentially a measurement partner, or different a brand and different CMNs. There's lots of different opportunities for multi-party and we're seeing two three brands want to collaborate together to do everything from media to like traditional direct marketing. And so I think there's a really big opportunity with clean rooms in that area when you think about it from a multi-party standpoint. Um and then I think from a technology standpoint the privacy around the clean room and the protection of that data and the fact that you are not you're not sharing data that you are prevent you are providing a view into a set of approved data. Those are the things that I think are are pretty critical. But again, I think a container also act like the actionability is much faster in that way. That's exactly where I was going want to go over this as well as the the because I I'm work with health care marketers pharma marketers, they're by design very conservative. Yeah. Um and it's almost was interesting to see how health care hopped over the whole DMP thing and went straight into clean rooms because DMPs were not kind of they were just not ready for them. And I I always before I joined health care I always assumed they're just not going to do anything. It's not that they're don't want to do anything, they're just behind on on everything else. So it was interesting to see that with clean rooms we finally got them to a point to get comfortable Yeah. with with uh um being able to share the data match it with with potential partners. I I was just going to say just on the health care it's really been interesting to see health care from a regulated standpoint leapfrog other regulated industries with leveraging technology. our language. It's Yeah. Yeah, it's really interesting to see that. Yeah, that it it totally makes sense. Um you know, I just having a conversation Sorry, guys. Um I was just having conversation maybe on Sunday night where we talked a little bit about a few years ago like we came here and it was we talked about clean rooms, clean rooms, clean rooms. That was, you know, I think um containers are starting to like are more like a little bit later starting to like enter the ether. So, we've had a couple of years of of being able to like navigate around the adoption of clean rooms. Um So, what what do you think we've learned in that time period? Um you know, it's it's a different conversation that we're having today. Um more of more of a uh this solves everything and more of a a get down to the nitty-gritty of it. So, let's talk about uh the adoption the last few years. There's no silver bullet. I think we're in the industry of good enough and and if we're doing something better today than what we did yesterday, we're moving forward, right? Um what I've learned is who are my trusted partners like more kind of consistently consistently. I would I would I would start there. It's it's interesting. I think when we when we introduce clean rooms to the market, this is the broader way, clean rooms could solve so much. And what we learned was that unless you stay really focused on the use case and stay focused on what it is that you're trying to solve for, first off, the technology's never going to go anywhere. You're not going to be able to move forward with it. That's one. Two is um the importance of identity because of the fact and having the ability to connect all of these different you know, think about two or three different parties coming in which happens quite frequently when you have a CMN, a publisher, and a brand which come in we do that quite often. They could potentially all have different identity partners and all define consumers differently. So, I've having like a consistent way of being able to map that, I think it's really important because we'll probably get to a question on this, but like scaling it's also really hard. It's not It's not just easy. It's not the click of the button. You know, your organization has to change. There is um a lot of let's say like there's there's there's going back to the the point around like the legal, the privacy, how is that data going to get to be used? These things aren't easy to solve, and so you got to make sure that your technology infrastructure and your ability to match and do this accurately is critical. Yeah, just to add on to that, too. I think [clears throat] the the hype around clean rooms was the time of a lot of the cookie deprecation. It was. Uh that was the big thing. Now we don't need to say that word cookies anymore, right? Cuz we just forgot about it. Um [clears throat] But everybody still uses them. We still eat them. [laughter] No, so like I think because of that the hype went past. So the the issue is I think we talk less about clean rooms today and at least in my conversations I have. And a lot of it's because there's a few drawbacks that we had in it. And what I mean by that is first it was the activation gap. Clean rooms are awesome after the fact for insights, right? Understanding after the flight is over, what happened, right? Not in And then also like taking those insights and being able to activate them across your different channels has been hard. Uh we talked a little bit about interoperability, and you can't interoperate. Walled gardens have their own clean rooms, so you got what? An N number of clean rooms to deal with. Um I think a lot of that's still fragmented, and it's coming together. I think it's coming together. I think Yeah, I mean yes. And I think was it coming They also are working with a small set of partners to bring some of that together. And I think early on with clean rooms, the other piece was um it was expensive, right? It's like you needed big uh advertisers, uh big brands to be able to use them. So a lot of the mid-market was kind of taken out. And I think with now with the way that it's evolved, uh you're going to kind of start seeing it being used more and more because now you can make you can actually action on it a lot quicker with a lot of technologies. Do you think maybe that I'm going to say it. I think that Do you think maybe that paired with more like agentic automation is can kind of drive some of the adoption of clean rooms because if you're talking about you know, it just being super manual to be able to set these up across multiple different parties is that I My view is yes, uh but um and you know, I know we were uh I was trying to avoid that word agentic Listen, it's my time. Is that point in the panel, guys? Agentic is great for certain use cases. I think in the clean room capacity, it's actually amazing because now you can pull insights that you wouldn't have thought of that you have to do manually and you can start uh teaching your LLMs how you think, how your brands think, right? So, I think it really helps in that. From an execution and activation layer, maybe there's a use case, um but it's not going to change it's the same process, well, right? It's an automation problem at the end of the day. Okay. At least that's my view. Yeah, I I I generally agree. Um I think that like we're partnering with different agentic measurement partners and things like that to really kind of get the action to get the insights out of it because it is hard and it is a heavy lift. Like let's just say what it is. Um the other thing that that I think that we are starting to work on is like working with uh our set of partners to then build kind of the end-to-end agentic workflows. Yeah. So, that way you might have the clean room underpinning that with the rules, with the context, with um with uh I'm blanking on the word I'm looking for, but um and then being able to then just more seamlessly I keep using CMNs as an example, but just more seamlessly power them with and again it's not us building it it's our working with a lot of our partners uh so that you create a much more seamless and and honestly user-friendly and you are asking a question and it's then plugging in and going out and saying, "Hey, do I have access to this? Can I do this? What are the rules around this data?" That's where I think there's a lot of opportunity. I think I mean to me it goes back to the same use case a gigantic use case across the board. Um what are the most repeatable processes Yeah. along the way and if I automate them, how does it change the overall process and does it actually save anything? Or it becomes It doesn't help our decision. Is there any If the answer is yes and often it is, then yeah, there's a place for it. Do Is there anything that you think we learned from the adoption of clean rooms that we can apply to connect like to containers, right? We're talking about containers being the execution, right? We're talking about um you know, there being opportunities in clean rooms to at least use a gigantic measurement to pull insights out of the clean room, etc. Do you think there's anything that we've learned from this that now we have more tools because of AI to to be able to do that that we can apply to containers? This is not on the This is not a question on the card. I'm just asking. [laughter] Do you mind if I answer the question that's not on the card? Yeah, go for it. I think what we've learned is that ad tech is seeing the power of having new tools. It's like a clean room enabled us to do things that we couldn't do before because we started to move some of the modeling toward where the data was and you could do it in a safe way. Now with the container, we're starting to move some of the decisioning toward where the bidstream is and as ad tech evolves, I think we're questioning some of the assumptions that have been kind of held for a long time and we're seeing a big shakeup in where different pieces of the functionality live. And I think that AI will have that same impact, but again, it's just a technology and the question is going to be does it let us move pieces of the workflow or pieces of the execution flow to where it makes more sense to do them? Mhm. I think one of the things that, you know, form that question is also um when we think of clean rooms, we always, you know, also say like we got the insights and the activation match rates drop. And a lot of that was because we've always done everything on the buy side. And uh that's the crazy part about how we think about our ecosystem, right? We always thought about DSPs and stuff. Now, when you take it more upstream into the supply side, you're just getting better match rates, you're able to iterate a lot quicker. So, um I do think they they work hand in hand uh in kind of doing it. But, the question is how do you collapse the ability to pull those insights into more real-time nature, right? Like how do you not have to wait after the fact because you could push new insights into the bid stream pretty quickly. Uh but, the question is can you get those insights and have a feedback loop that's a little bit more responsive? And again, [clears throat] it's it's the input cost. Is it worth it? And so, to answer your question, no, I haven't I haven't I don't know the answer to that yet and I don't know how long it's going to take us to um figure out if it moves the needle or not. We are constantly struggling still to this day to pay publishers more in the pro- in the programmatic execution because we know the quality of inventory we get when all the input costs get get taken out and kind of constantly balancing that act and sometimes it may sound good and then at certain scale it just does not move the needle. So, conceptually, absolutely makes sense. The question will be will it move the needle? Will that input cost provide better value proposition to the supplier in the end of the day? That's and I don't know the answer to that yet. All right. So, let's um let's transition and let's kind of get into a little bit more details and talk about like what are the real use cases? Um you know, what are your companies doing today? Um how does your approach solve this more cleanly? I think like this really dovetails well into your question, which is like there's still some uncertainty about what is good and what is not good. So, the question is tactically like what have you seen work? What have you seen not work? Omnichannel orchestration works very well. Honestly, that to to us that's our biggest use case. This is where we see immediate value. Because once we can use a clean room to define the audience with suppliers with suppliers on the end being able to activate it now across multiple channels and measure it back provides very good picture for us. Okay. So, you are getting access to more measurement data and you are understanding more of your omnichannel reach, frequency, all of that kind of data. How does the audience travel? How does the audience travel? And you're able to get insights into that in a way that was always bifurcated and is even stronger today. That's cool. And that's mostly clean rooms. Have you started to play with containers yet? Not yet. Okay. Uh, from our standpoint, we see it there's a a couple of big ones. One is, I've said this a few times, obviously a lot of commerce media networks have invested in clean rooms because of the need to collaborate across multiple different parties and really kind of build and scale their brand partnerships. So, that's been that's been really critical. Measurement, the one of the biggest things is like unlocking some of our biggest uh, CTV partners, different social media partners, things like that to really unlock measurement, which probably goes back to a little bit of what you were saying around some of the omnichannel orchestration. And so, that's been a really big big driver. Um, on the CMN side, obviously we see it everything from targeting through measurement. Mhm. Um, I think we're starting to see it be used more for planning. We're starting to be see uh clean rooms uh now being tied tighter into optimization. So, like happy days and things like that to be able to action off of it. And so, I think there's a lot happening. The other thing I would say though is like I'm talking about these use cases, there's so much more we could do with it. Yeah. And this is very nascent. Like both of the things that we're talking about up here, even though we've uh we probably have like clean room or AI, like we're tired of hearing it. All of this is nascent. Yeah. And nobody knows fully what they're doing with it yet. All right. I mean So, so I pulled out from from that uh target it I pulled out targeting, I pulled out measurement. What about uh containerization? What are you seeing in that area? So, on our side we we do a lot of that, right? Uh the whole purpose of containerization, the reason for that was um if you think about programmatic uh DSPs and SSPs uh we like to say it talk at each other, not to each other. Um on the supply side there's a lot more supply that's actually sitting there. Uh a lot more bid requests, a lot more of vails than actually get to the DSPs cuz of UPS throttling. And so, basically the pipes into the DSPs are smaller because they're not listening to the entire supply stream, whether you know, you're a really big DSP or a really small DSP. And because of that not having that feedback loop, you don't know what should be sent into the DSP to go and auction on. Um and so, the way that we solved it was we needed to be as close as possible into uh the infrastructure of the SSPs cuz we're not an SSP, but we need to take that decisioning logic. It's like what works and what doesn't work, place it inside the bid stream and be able to action in real time. So, whether it's an audience match plus a uh you know, what kind of bid request match, right? Uh and be able to shape that in a dynamic way. Um so, we're seeing a lot of great use cases for that. On the pharma side we work with uh you know, a large independent agency as well that does a lot of pharma advertisers. And pharma as you know, you can't individually target a lot of it's zip plus four backed into either IPs or user IDs in the consumer set. Um but yeah, so like a lot of those pieces, it's like how do you also go across device? Uh and it's possible now, right? Like if someone saw it on a CTV in a household, how do you find them also on other way uh other devices? And you could do that within the bid stream itself now. If you're solving the throttling problem, we're going to need to talk afterwards because in healthcare, making changes is hard. biggest issues when So we're looking for the needle in haystack and by default have the traffic has already been throttled out. We're fishing in this different pond not not in the pond we need to be fishing in. Yeah. So that's the we don't change the QPS. What we change is what goes inside the pipes, right? And you want what's inside the pipes. there, yeah. Yeah. Makes sense. Do you want to take this? We're really a facilitator. You know, we're there to run your container with our real-time bid stream API. We're there to give you transaction-level data with our auction insights API. And we've got a graph which is kind of a hybrid. So I yeah, I think our goal is to make our make our partners successful and give our customers the tools they need to to find what they're looking for. Simple way I like to say it when you said that was would you rather be fishing in the ocean or fishing in a lake? What's the better one? I guess it depends on the size of the fish you're trying to catch. Yeah. Boom. [laughter] Or what kind of fish you like? That's right. What kind of cookies you use as bait? Can we keep going? [laughter] We could really take this one. I know. We could go we could go many places. go a lot of different places and I could I could definitely tie them to different ad tech theories as [laughter] well. Um okay, cool. [cough and clears throat] So let's um talk about something that we sometimes forget about. Which is the consumer. [laughter] What What are Never never never. Um uh What approach is better for the consumer? So when you think about these two different approaches, is there one that's better if you had to choose how data is processed and stored? Like what are the things that for a consumer um like what it how does how do we tactically execute on these things? And how does it intersect with consumer behavior and and and consumer needs? I mean consumers are always going to say less ads, right? Um and that's not I don't think that's going to happen with what as a a from a consumer standpoint again um it it varies so much in that some consumers want more relevant pieces served to them. Other consumers just want a better experience. Um and they they don't care on that. So like if we could take a lot of this and have less placements and make it a better experience, I think it's a win-win for everyone. You know, one of the things I always said about like retargeting for example is like it's like consumers are like I don't want ads following me around the internet. I'm like then stop clicking on them. Because those are the ads that you click on the most. But yeah, it's uh I mean in parallel with that like if we can streamline this and again publishers get to see higher net CPM of what we're willing to pay in the first place, they don't have to have as many ads on the page and that creates a value for the consumer because the content experience becomes so much better. Yeah. I mean at like as a person who works with publishers um if they can create an ad product experience that is uh significantly better and but also monetizes better um because it's got relevant data to it, it performs better than then yeah, we can actually see a world where there is less ads, but they're more effective ads. Yeah, I think the consumers want the best experience possible and they want it without using any data, right? Or like you know what I mean? Like we all want our data protected, but we also want a highly relevant experience. I think that if you do I think there's a place for both um in this world and in in kind of driving relevancy for consumers. I think that the types of data often times are a little bit different to degree, right? That are that are included in in both of them and so you probably need different privacy and and and uh rules and regulations around how they're managed. But I think if they're done right, there's a place for both of them because at the end of the day, the goal of all of this is to one drive better consumer relevancy and ultimately drive a better return for the brand and the publisher. I'll say like influencer marketing is similar in that way. Like you know when that at what when the message was shoved in there like a paid that versus when it's naturally kind of part of whatever Yeah. they're trying to do. So I think I can I can draw a line there. All right. Now we're going to start telling the future. So this says if you had to make a prediction in 5 years, I'm not going to do five. I'm going to do two because I think things are too are moving too quickly for us to to do five, but I think two. In 2 years, um how do you see these approaches evolving? What new types of things do you think are going to emerge or would you even like to see investment in? Um is there something today we haven't even talked about or thought about things that you're um that maybe this week has bubbled up for you that you're like ooh this might be uh a a direction in which our our industry is going. I would say as buyers we will rely less on DSPs to make all those decisions for us. I think we need to continue to work much closer with the supply side solutions. That's the reason why I'm here in the first place. [snorts] And um all very specifically uh data management of it because today it's all done through DSP, which is a couple of clicks away from where it needs to be. Well, there's been this like thing the last couple years where it's been like the DSP had all the data that we were bidding on, right? And now the supply side has more and more data as well, and it's becoming increasingly more and more important into driving the outcomes. And also the like actual contextual understanding of the environment in which we run ads is also expanding. So And uh kind of the I don't know, the expectation of better match rates when you're closer to the supplier, to the actual publisher. Yeah. There you go. It's interesting also to watch brands get closer and closer to the supply side as well. Right? They are moving and educating themselves and directing the agency, and they are spending more and more time with the supply side as well, which I think is interesting. In terms of what I see, um So I don't really know, but what I will say is that I think as automation, as AdJintic, as all of these things continue, the tech underlying technology will remain critical. And yes, we will continue to have new technology that's built into this infrastructure. But none of that means anything unless we have a context layer. And I think that is going to be one of the most critical things as AdJintic continues to scale and drive efficiencies, having a context layer, I think it's going to be one of the most important things moving forward in the next couple of years. So on my side, right? Like to kind of hit on the AdJintic uh train, um you know, to me AdJintic means I don't want to use a UI. I want to go through one place, right? Like we have so many different platforms to log into. How do we kind of bring collapse it all together? Um and bring it all together in one place. The piece that's going to like it helps decisioning, too. I mean, it doesn't just let you Anyway, you go. It could. It could. It could. Um I think uh where was my Oh, now I got it. Uh uh lost it for a second. Um so, I think it's the way I think of it is there's a there's a control layer and an intelligence layer. Uh for the longest time, a lot of these platforms owned that layer. And I think a lot of it's going to start moving over the next few years into the agencies and the brands that are actually controlling the media. And it should um because then you can go on any platform anywhere that you want based on what is right for you in the moment. As long as you own all that logic that gets in. So, that a lot of that comes in through normalization. The problem in our industry isn't developing new custom algos or models or anything. The problem is the data that gets fed in is it properly normalized and clean that you can actually action on it so that you don't get the wrong answers. Cuz in an LLM, that will happen if you don't have good data and it's going to be a terrible answer. And you don't want to just lean on that. Here, here. I'm going to answer both questions. You asked about the consumer experience and then where are things going. Mhm. What we've seen is that's parts of the decision process are moving closer to the edge. So, you know, decisioning moved from the DSP, parts of it are moving now into the SSP. I could see a world where it now moves upstream to potentially the user agent, whatever that may be. And then you see data moving around, too. Talking about things that don't exist today, I would love to see a world I understand the value exchange. I'm okay with giving my data to different people to do different things, but I want to control my data. So, I would love to see a world where there is a clean room that can access my profile that I control, and then there's decisioning that happens as close to me as possible that gives really good insights to what types of ads I want because I like free content. And I don't mind seeing an ad for a guitar here and there. Andy, or you know, uh or a Huckberry ad, right? It's like those things are okay with me. You're about to get targeted with those for the rest of Can-Trip. Yeah. I like guitars, motorcycles, and no. It's putting It's putting like the money where the mouth our mouth is, right? control where it needs to be. Yeah. I mean, yeah. People are terrified it won't scale. All right. So, you guys heard it here. Relevant ads, context, getting closer to the supply. Good summary. Everyone drop your mics. We're out of time. So, all right. Thank you guys so much for coming. Thank you.