Latest / Signal & Noise: Bringing Clarity to Media, Data, Tech and AI / Rewriting the Rules of Programmatic: Adam Heimlich on Agentic Bidding, ARTF, and Building Chalice AI
Transcript
- 0:03This is signal and noise, bringing clarity to data, tech
- 0:07and AI. Everybody, welcome back to
- 0:15Signal and Noise. I'm Brett House, joined by my Co
- 0:18host Real Longacre. And today we're talking about
- 0:21one of the most interesting founders working in AD tech
- 0:25right now, someone who has seen the industry from both sides.
- 0:29Adam Heimlich spent years inside the agency world, including
- 0:32leadership roles at Verizon Media, working directly in the
- 0:35trenches of programmatic. He saw first hand how the system
- 0:39actually works, the incentives, the inefficiencies and the
- 0:41limits of the tools. Media teams for the media teams
- 0:45that were using them every day. Eventually, he decided to do
- 0:48something a little more dangerous.
- 0:49Something that I've done recently and it is terrifying at
- 0:52times is build a new company, launch a new company, go out on
- 0:57his own. So he's the Co founder of and
- 0:59CEO of Chalice AI, a company for those that don't know, trying to
- 1:04rethink how programmatic buying actually works from a gentic AI
- 1:07to new RTB standards through a containerized architecture.
- 1:11And we will talk about what that means that let's advertisers
- 1:14run. Yes, I'm sure there are many
- 1:16questions about. That totally run their own
- 1:18decisioning logic directly in the bidding process.
- 1:21And again, if that's flying over people's heads or in one ear and
- 1:24out the other, we'll talk about it in detail.
- 1:28So I, I, you know, my, my perspective on you, Adam, is
- 1:30you've got a very different way of thinking about programmatic
- 1:32infrastructure. You're certainly a visionary in
- 1:34the industry and, and you've done your fair share of shit
- 1:38posting. It added good mixed sharp.
- 1:41It's, you know, mixing good sharp insight and LinkedIn and X
- 1:45and others. Welcome to the podcast.
- 1:47It's a thrill to have you and would love for you to give a
- 1:50little introduction. I don't know if they did you any
- 1:52justice there or if that was completely biased.
- 1:55That was fair. I just want just got to plug my
- 1:58live Friday show with Gareth Glazer, Ad Tech Ad talk, since
- 2:02we're all podcasters here. That is it.
- 2:04It's a good one. Q&A.
- 2:06We're a buy side sell, a buy side optimizer and a sell side
- 2:10optimizer. Take questions, not many
- 2:12questions this week, but sometimes we get into questions
- 2:14like between him and I, we know where the bodies are buried and
- 2:18how things really work. So recommend that show.
- 2:21It's it's Fridays at noon ET on on YouTube.
- 2:25But I yeah, everything you said it was accurate.
- 2:27I'm a older founder. I was had a good 30 year media
- 2:32career, 10 pub side as a writer, journalist and 20 agency side.
- 2:41Toward the end of that, I became an ad tech consultant as sort of
- 2:43a a bridge to founding and I I founded early in the pandemic,
- 2:48so just about six years ago this March, which was also coincided
- 2:53with my 50th birthday. So it felt like between the
- 2:57pandemic and my 50th and everything I'd seen and the lack
- 3:00of freelance work available when the pandemic started it, it
- 3:04certainly felt like a a golden hour.
- 3:06Crisis. What?
- 3:09I said a midlife crisis. It was not a midlife crisis.
- 3:12I mean, I hope I live to 100, but it wasn't any kind of
- 3:16crisis. It was like, I really want to
- 3:17work. I have ideas.
- 3:19I had the business plan in the drawer and yeah, there was
- 3:23nothing else to do. Well, Adam, we're thrilled to
- 3:26have you on here. Like, like when we started this
- 3:28podcast not too long ago, there's a short list of people
- 3:31we drew up and, you know, you were on that list and glad we
- 3:34could finally make this happen. And yeah, I think one of the
- 3:36reasons why we wanted to have you on here besides just I think
- 3:39like I I love a lot of the things you say, some of which
- 3:42are, I think non, I mean, are completely like makes so much
- 3:45sense, some which are maybe a little controversial sometimes,
- 3:47which I love as well. I, I think that you've got a
- 3:50cool story, even beyond the tech like like what happens when
- 3:53someone who actually like, you know, worked and worked in the
- 3:56trenches like you, you had a big programmatic job at Horizon,
- 3:59used to be a client of mine as well as other places, and you
- 4:02spent a bunch of years like running executing campaigns for
- 4:06some major brands. So it's, I think it's cool to
- 4:09see that you went from that to actually doing your own thing.
- 4:13I think that's part of the story here.
- 4:15And then like, and when I'm looking at what you've done with
- 4:17Chalice, just whole like effort to rethink the architecture
- 4:20about taking like what was like static bidding logic, you know,
- 4:23plague Pig platforms with not a lot of transparency.
- 4:26And how you're exploring this world where leveraging AI and
- 4:29leveraging, you know, Agentic to see how we could completely
- 4:33change the bidding process, change the decisioning logic
- 4:37with this portable containerized approach that I really think is
- 4:40cool. So we're going to dig into that
- 4:42and the idea how like these initiatives you're doing, like
- 4:44the Agentic RTB framework, ARTF that's being explored with you
- 4:49along with IAB Tech Lab. I think this really brings out
- 4:53like some interesting questions like if a of a agents are
- 4:56increasingly making media decisions, what does this mean
- 4:59to for advertising and how media is bought and sold?
- 5:02I mean, I think it has some really big implications which
- 5:04are which are pretty cool with that.
- 5:06Let's kick it off some questions.
- 5:07I think Greg, you had the first one.
- 5:09Yeah. So, you know, so being having
- 5:11these leadership rolls up companies like Horizon Media,
- 5:14would that experience teach you about how this, you know, how
- 5:17programmatic actually works versus, you know, what the
- 5:21industry likes to say about it, So the positioning versus the
- 5:24reality? Yeah, I mean, because I'm old
- 5:28enough that I saw pre Internet media business and just because
- 5:32I'm kind of curious and I'd like to take a step back and have a
- 5:36level of awareness of what I'm, what I'm in from a outside, from
- 5:41an imaginary outside position. I kind of saw the incursion of
- 5:45technology into advertising as as like an as like an invasion.
- 5:49And with the, it wasn't just technology, it was VC too that
- 5:55came into advertising and really changed it.
- 5:58So I think having the knowledge of what it was like before
- 6:02allowed me to follow that as a narrative of how, you know, this
- 6:06is how the tech people think about it.
- 6:07This is how the money people think about it.
- 6:09But I'm native to this advertising and media domain and
- 6:14I could, I could trace the fault lines of, of where they're not
- 6:17communicating or where it's not working.
- 6:20And I think, I think by doing that, it put me in a good
- 6:23position just to match right, new technology with old
- 6:28problems. That's what I was always trying
- 6:30to do. And you know, I, I, I adopted
- 6:34search really early. That's how I got into
- 6:36advertising. And I realized, so just by
- 6:37adopting this thing that was lying around, you know, Google
- 6:41Adwords have been out a couple of years before big advertisers
- 6:44used it. You can make a big change right,
- 6:46just by learning the tech. And it was the same with
- 6:49Facebook ads and you know, even more so with programmatic.
- 6:53It's like I would got really into this pattern of adopting
- 6:56the technology and how the technology changes the
- 6:58advertising and advertising will shape the technology if you if
- 7:01you have a voice and how it's made.
- 7:04So I was, I saw a lot of people split off and form service
- 7:08companies around just the fact that they could use the
- 7:10technology. I stayed within the agency world
- 7:12just getting, you know, more and more executive around this
- 7:15ability to use the adopted technology to client benefit.
- 7:20But then I chose AI as my wave. Like as I said, I was turning 50
- 7:23and and I the AI wave was already gathering.
- 7:26It was clear it was going to be something I didn't.
- 7:29It might have underestimated how how big it would.
- 7:31Be the tsunami on. The longer ride than than maybe
- 7:36I thought, but I think I think that's what equipped me to say
- 7:39like, all right, so this this round of new technology hitting
- 7:43the old advertising business is like the fourth time I've seen
- 7:47it. And there are certain patterns
- 7:49around adoption and resistance that, you know, I don't have to
- 7:53get upset or knee jerky about anymore because I've seen it.
- 7:58Was there a particular founder moment of time you're working
- 8:00like let's say with clients agents, you said you thought,
- 8:02OK, this is broken. Like there's got to be better
- 8:04way to do this. Like I have an idea or was it
- 8:06just the AI thing and you were you were early to that which I
- 8:09think so your timing was good. Yeah.
- 8:11I was aware of customization and back end data use from my early
- 8:14days at Avenue A, which had the Atlas AD server in house and a
- 8:19really good analytics team and they would go into the event
- 8:22level data. That's what we called it, event
- 8:24level data. Now we call them logs and they
- 8:26would see that in the, you know, that simple attribution was like
- 8:30a rule of thumb for which impressions were effective.
- 8:33But if you went into the event level data and you could start
- 8:35to see which impressions actually had influence, right?
- 8:38They would, they would use causal influence methods.
- 8:41And, you know, I was very curious about this and, you
- 8:44know, I think over the years I've gained some expertise in
- 8:47what was going on, but I was, I was following that narrative
- 8:49like, well, this is better, but it's, it's slow and it's hard to
- 8:53sustain and it's expensive for the agency.
- 8:56So, you know, just see, you know, my experience of AI is
- 8:59like, wow, this is now really cheap and fast.
- 9:02Yeah, Yeah, it's almost like you find you, you get technical,
- 9:05you'll learn, you'll learn the systems and then you find the
- 9:08gaps and then there you find the opportunity, right.
- 9:11Yeah, as as, as sort of a founder and an and an
- 9:14entrepreneur in a sense, right. Yeah, I think what Yeah, the,
- 9:18the big thing that that tech and money people are always, always
- 9:21miss is the the whole enterprise market because it is a minority
- 9:24of the money and it is complicated in a way that annoys
- 9:28them if you're trying to think of scale.
- 9:30But that's where I always preferred to work.
- 9:32Like the problems are more interesting.
- 9:34You meet very interesting problem solvers on the brand
- 9:37side. And I found like, I could get as
- 9:40interested in, you know, the problems that CarMax was having
- 9:43with incrementality is was when a kid, I would get interested in
- 9:46a, in a band or a rap group, Like it was just, to me, it's
- 9:49just like a media problem and solution.
- 9:54Yeah. So starting an ad tech company
- 9:56today is a lot different than obviously 2010 or 2015.
- 10:00We're in a much more mature market.
- 10:02There's a lot of skepticism and it's it's crowded.
- 10:05What convinced you that Chalice was unique?
- 10:08That I had an opportunity to really make a meaningful change.
- 10:12But I did get a lot of encouragement from people who
- 10:14succeeded in the first wave right among in Chalice's CAP
- 10:17table, this Trade Desk's TD 7. And we pitched Jeff Green
- 10:21directly, Brian O'kelly and Jonah Goodheart, Matt and Art
- 10:26from according to Joe Zawatski and Eric from Aperium are in our
- 10:29cap table. So all these people seeing the
- 10:31business plan and being like, yeah, this, this, this can work
- 10:35was was a huge wave of encouragement though it, it came
- 10:37like a year later. But I think before that was
- 10:40like, you know, the first people I showed the business plan to
- 10:42and the 1st data scientists and engineers who were coming on who
- 10:46were very encouraging and saying, yeah, that like what's
- 10:50happening now in cloud and AI services is making this cheaper
- 10:54and lighter and faster than it would have been even, you know,
- 10:57two years ago or even faster than last year.
- 10:59And that dawning realization that the cloud companies had
- 11:03were investing billions, even hundreds of billions in
- 11:07infrastructure in anticipation of companies becoming AI
- 11:11factories was like, great. We're like early in like, you
- 11:14know, data bricks and snowflake. They love us because we've been
- 11:17building an AI factory for six years on their infra, running up
- 11:20bills and, you know, creating use cases.
- 11:22So that's so that's a huge win on on your back that I don't
- 11:27think the first wave had. Yeah, it's a it's yeah, like I
- 11:31said, you were early, which which is really cool to the AI
- 11:34story. And there's a lot of different
- 11:36AI companies now. And you know, it's interesting
- 11:38what you said Brett too, about like the cost and time it takes
- 11:41to have a start up. I think I read a stat the other
- 11:43day that the cost for building a company's gone from I think it's
- 11:463 three and a half million down to, you know, 5-6 hundred
- 11:49thousand, right, with some of the new AI tools.
- 11:52But but regardless of cost, it still is a big psychological
- 11:55change, right? I imagine going from an elite,
- 11:57going from operator to founder, like like what, like any, any
- 12:01takeaways or things to call out for other people, because a lot
- 12:03of people right now are considering that.
- 12:05I've, I've noticed so many people like, you know, Brett,
- 12:07you know, included who are. Yeah, taking the.
- 12:09Taking taking a launch because it exists.
- 12:11Yeah, yeah. I mean, I thought I my, my first
- 12:16realization was like, it's not a certain kind of person.
- 12:19It's just something you do. If you did, once you do it,
- 12:22you're a founder and then, you know, you figure it out.
- 12:24It's not like anyone. I mean, you meet people who are
- 12:26entrepreneurial in high school and are, you know, you know
- 12:29they're going to be found, but that's unusual.
- 12:31And there's no, there's no founder type.
- 12:33I met a lot of people who are founding at the same time as me.
- 12:36I'm still in touch with them. It's very, it's a very diverse
- 12:39group. And you know, you don't have to
- 12:41have the courage all the time. You have to have the courage at
- 12:44the moment of founding, then you just then just hang on.
- 12:47Yeah, yeah, yeah, exactly. And then it's like it's like
- 12:50it's the old fuck Plan B sort of mentality, right?
- 12:54There is no it's. Like having a kid too, it's like
- 12:55you're not going back. So it's like, you know, you just
- 12:58have to. Burning the boat so.
- 13:00Way but forward, right so well, I think one of the most
- 13:03interesting things, and I got to say this, being married, I've
- 13:05got 2 two boys soon to leave the house. 1's already left the
- 13:10house. But you built this alongside
- 13:12your partner, Ali Manning. Yeah.
- 13:15You know, and I, I tried to imagine myself in your in your
- 13:18shoes and I thought, wow, I don't think that could work on.
- 13:21Well, I could. Yeah, I probably could work on
- 13:23my wake. Up your earrings.
- 13:25But but how did you, how did you pull that off?
- 13:27How does that dynamic work? And how do you feed off of each
- 13:29other and see how to do each other?
- 13:31'S way. I'm going to say it works.
- 13:35It happened. I mean, me and Ali met at work.
- 13:37We met at Avenue a Razorfish. We're all working in back end
- 13:41data for a large banking client. So we had, you know, experience
- 13:47of working in the same industry and we coached each other for
- 13:50many years. She was at Google while I was
- 13:52agency side. So we were sort of dealing with
- 13:55each other's companies and we were both advancing in our
- 13:58careers and, and helping each other.
- 14:00So, you know, we, I knew that we were a team Chalice was founded
- 14:05in those first six months by me and Ken Rona, who's a data
- 14:08scientist and kind of hacker engineer, a real, a real wizard
- 14:11in the early stage stuff. But by that summer of 2020, we
- 14:17wanted to pitch for funding and make a business plan.
- 14:20Ali said his operations. So she's made business plans for
- 14:23billions of dollars worth of worth of budget.
- 14:25So I said, will you help us with the business plan?
- 14:28And I think she just she just got interested, you know, that
- 14:31it was, you know, the pandemic year, like no one had really had
- 14:34full time work and could be that engaged in what had been an
- 14:38office job and then a home job. So I think, I think the more she
- 14:42did it, the more she enjoyed it, though she had been trying to
- 14:44get out of advertising. And yeah, it's, it's really
- 14:49rocky. Like the all the reasons people
- 14:50say that it'd be hard for them or that you know, it's a
- 14:53dangerous thing are true. Like, you know, you want when
- 14:58your, when your partner is your coach from work, like they're
- 15:00entirely on your side. Whereas when you have a work
- 15:03argument, they can't really coach you through it because
- 15:05they might the physician might be adversarial at work and
- 15:09feedback, negative feedback from your partner hurts much worse
- 15:13the negative feedback from a partner.
- 15:14But I will. Say you're much more likely to
- 15:16take a little personally, right? It's got to be personal to.
- 15:19Say the least, Yeah, It's only professional.
- 15:21It's only professional. But I think it works for the
- 15:23company, like the founders should be essentially in a love
- 15:26relationship, like they should be complementary and Better
- 15:30Together than neither of them are.
- 15:32Are a part which is close to the definition of of love or a
- 15:35marriage, right? It's like.
- 15:37Without the codependency. Without the codependency.
- 15:39They shouldn't. Be Well, I think the best
- 15:41startups, the best startups have that, right?
- 15:42Yeah. Like Jobs, it was in the act.
- 15:44You had like, yeah, Gates, Gates and Allen where you have people
- 15:47who like, bring complementary skills.
- 15:49So. Yeah.
- 15:50And why not, right? And, and help shape each other's
- 15:52best ideas and protect the company from each other's worst
- 15:55ideas. And oh, that's good.
- 15:58But it can be real struggle for us to come together.
- 16:01The decks we've made together, some of them have been struggles
- 16:05and involved like couples therapy sessions just to finish
- 16:07the deck. But those are, those are our
- 16:10best decks and that that's our best work.
- 16:12But we do have a pretty, we do have a pretty solid department
- 16:15split. So we're not running into each
- 16:16other. She is operations, sales,
- 16:19marketing, finance, HRI, AM, product strategy and
- 16:25partnerships. Well, from the exterior, seems
- 16:28like you're making it work. You know, I follow both of you
- 16:30on, you know, online and I see a she'll post a picture of we
- 16:34brought the kids to the office because school was closed or
- 16:37some emergency, you know, seems to work is is pretty cool.
- 16:40So like like I'm a start up. Life is supposed to be like all
- 16:43consuming. Like, how do you balance that?
- 16:46I mean, I can imagine it's a lot, right?
- 16:49Yeah, it's a lot. I mean, a lot of people with
- 16:51kids work hard and we were never like, you know, like young
- 16:55founders who were like working ridiculous, you know, 18 hours a
- 16:57week and through the weekends. We we never worked like that.
- 17:01We never worked those kind of hours at agency.
- 17:04That's good. So if you've kept a little bit
- 17:06of a balance. Right, Yeah, we've kept some
- 17:07balance. But you know, the, you know, the
- 17:10reality with little kids is if you have good health, like it's
- 17:13a few hours, if there's an hour or two in the morning and you
- 17:16get them to school, there's hour or two in the evening and then
- 17:18you put them to bed. So it's not like, you know, it
- 17:23was manageable. We're making, you know, we had
- 17:25some money and really, really good help and a good
- 17:28neighborhood and some relatives and all that.
- 17:30So I don't want to exaggerate like how crazy it was, but it
- 17:33was many, many years of not really resting, like going from
- 17:37work, you know, work to the kids to the work to the kids and, and
- 17:41even a vacation being being tough like for many, many years.
- 17:44So it was, those are tough years and it was a tough thing to do
- 17:47being over 50, but it's a lot easier now.
- 17:50Yeah, yeah. It'll keep you.
- 17:51It'll keep you, you know, sharp going into your later twilight
- 17:54years, right. You know, the I'm more active.
- 17:56You challenge your brain, right. And I think I, what I've learned
- 17:59is you gain a little bit of wisdom with age.
- 18:01And I just, I'm a much more, I think I'm much more capable of
- 18:04doing this now than I was in my 20s.
- 18:06Like when Rio and I started something together.
- 18:08You know, I didn't know shit. I just didn't know enough from
- 18:12from both organizational management to multifunctional
- 18:16management. I just didn't have enough sort
- 18:18of wisdom and knowledge. Well, plus, remember Brett, we
- 18:20like because we ran the small like tech company, like, you
- 18:23know, early ad network. We, we did a lot of media
- 18:25buying, you know, for, for some clients.
- 18:27Remember, we worked a ton, right?
- 18:28But I think a lot of that was just like hanging out and after
- 18:31hours playing ping pong and stuff like that too.
- 18:33So like I think that now like I agree with you Brad, probably
- 18:36better at managing my time and better at focusing 100% on
- 18:39stuff, right, So. Yeah, come on.
- 18:41And. Team yeah like you know the
- 18:43reason it's big reason is easier now is we have a really good
- 18:46leadership team and some of these people have been with us
- 18:48with us for years our our CTO Joe Wilson and head of AI and
- 18:52data science Thailand Petri joined us in 2021 or 2022 and
- 18:57they you know have led and grown their own departments of those
- 19:00are of course the two of the hardest so it's you know Rob
- 19:05Mooney is a better product has also been with us for for for
- 19:08many years and it's. That's great.
- 19:12Well, that's a testament to your, to your leadership skills,
- 19:14to the fact that. Yeah.
- 19:16And the people that stick around, that just shows that
- 19:18loyalty is there for a reason. You know, we're on a mission.
- 19:21We're on a mission. Yeah, it's it's become a family
- 19:24in a way, right? It's so let now let's take now
- 19:27that we've got personal. Pivot of the tech, yeah.
- 19:29Yeah, yeah, yeah. Let's talk about technology.
- 19:31So you've been deeply involved in the agentic RTB framework,
- 19:34ARTF for the acronym, if people aren't familiar with it, with
- 19:38the IAB Tech Lab. And for listeners that may not
- 19:41be, you know, in this day-to-day what, what problem and I have my
- 19:46own summary of what I think ARTF is actually trying to solve, but
- 19:49I'd love to hear from you what, what you guys are trying to
- 19:51solve when it comes to autonomous.
- 19:53And how it's different than some of these other other approaches
- 19:56we're seeing too. Yeah.
- 19:57Yeah, I'm going to start with what people know, people.
- 19:59People know that AI is really powerful, and they know that
- 20:03behind this magical machine is an enormous amount of data
- 20:06processing, right? Everyone knows that, right?
- 20:09They could see that. They read about the costs.
- 20:13And in AD Tuck, there was really no way to bring what we know of
- 20:18as AI like that sort of magic into the event level for the
- 20:23exciting part is the the actual bidding because of the pipes,
- 20:28right? It's like, it's like you're
- 20:29trying to bring a it's like you're trying to bring a 2025
- 20:32hot tub into a 1930 house. The infrastructure can't carry
- 20:38that. But ever, you know, just
- 20:39everyone knows it's a lot of data processing and they're
- 20:42actual data processing limits of the the platforms.
- 20:46So that should that should be simple enough for everyone to
- 20:48understand, right? But not getting a level more
- 20:51complicated is, is we started, you know, on the idea that we
- 20:55should use every field in the bid request.
- 20:58One of the ways that that pipeline has been narrowed for
- 21:00platforms is they mostly look at the IDs and not all the IDs.
- 21:05They look at the their own IDs and they especially look at IDs
- 21:08that they could track conversions of and especially
- 21:11IDs that they've tracked conversions of lately.
- 21:13So that shrinks things down to a much more manageable level for
- 21:16their machine learning. So our first idea is, all right,
- 21:19we're going to work with bigger models.
- 21:20So we're first of all we're going to look at all IDs and
- 21:23second, we're going to look at other fields.
- 21:25And as LLMS get bigger and bigger, we started to get more
- 21:27interested in the language fields, which are mostly thrown
- 21:30away in the old ML pipeline. So there's domains which were
- 21:34turned into categories, so they could be numbers, but they
- 21:37actually started using the words in the domain and then the words
- 21:40in the sub domain, then the words in the full URL.
- 21:43And then we started working with Deep Sea, who scrapes the web.
- 21:46And we said, all right, let's every time there's a full URL,
- 21:49use all the language on that page and the metadata about the
- 21:51quality of the page. Now we have a much more
- 21:54predictive model and actually fix a big problem in the
- 21:56programmatic, which is when you only look at IDs, you don't care
- 22:00at all about the quality of the page.
- 22:01And if you're trying to get the lowest possible CPM, the
- 22:04incentive becomes to find that ID and the cheapest possible
- 22:07page, right? That's why we have clickbait.
- 22:09That's the reason, yeah. Without, without any of that
- 22:11context, which is so critical, yeah.
- 22:13Everyone knows clickbait's a problem.
- 22:15Most people don't know it's because DSPS only care about IDs
- 22:18and don't look at the page. So we're like, all right, we're
- 22:21going to have a model of all the pages on the open web and we're
- 22:24going to score them not in general for quality and context,
- 22:28but for how well they match what the advertisers trying to do or
- 22:31how effective they are for a campaign outcome.
- 22:34And then we're going to try to deploy this.
- 22:36And we had a lot of trouble deploying it because it's big
- 22:40and DSPS aren't, aren't built for it.
- 22:43We had been inserting our models in a DSP and we started talking
- 22:47to s s PS And right around this time in 2024, two years ago,
- 22:51Andrew Casally said to me, we can host one of your models in
- 22:55our server, right? He's the CEO of Index Exchange,
- 22:58which is one of the, he's one of those entrepreneurial guys who
- 23:00started our company in high school.
- 23:02He's been running that exchange for most of his life and had
- 23:09recently re architected it and explained to me about this
- 23:12technology called Kubernetes for edge computing.
- 23:15Which is the technology of high frequency training where you
- 23:19replace a server to server connection where a message comes
- 23:22out of one place and lands in another.
- 23:24And they send a message back to doing it in a single server
- 23:27where one of the one of the message systems is in a
- 23:30container like a like a virtual desktop, a computer within the
- 23:33computer. And then the message doesn't
- 23:37leave the server. It's a lot more energy
- 23:39efficient, it's a lot faster, it's a lot cheaper.
- 23:42That's basically it. That's, that's basically the
- 23:45idea of what became containerized RTV and IRTF.
- 23:49It took a long time to build. We didn't test it till the end
- 23:52of 2024. And then in 2025 we started
- 23:55running it. One of the first campaigns we
- 23:57ran was for Zillow. It did so well the case study
- 24:00ended up winning the best in show at the Ad Exchange Rewards
- 24:04we took what index and chalice. How, how did it do well, what
- 24:08was what were some of the metrics that you were looking at
- 24:10That sort of said, this is it, we nailed it.
- 24:12It was Zillow drives leads, but then they also want to look at
- 24:16the quality of the leads like how so it turned that ID looking
- 24:19thing into like how many leads are actually going to be useful
- 24:22for Zillow as people who are actually home shopping or
- 24:24actually buy a house. So we looked in the quality and
- 24:27context of the pages to get much better signal over whether this
- 24:31Zillow lead was a legitimate house shopper and it did like
- 24:3540% better than cost per lead basis than than ADSP on its own.
- 24:40So Adam, so sounds like just to summarize in for, I guess some
- 24:43reviewers. So it's a technology that's
- 24:45using because you're not only looking at like the standard
- 24:47contextual signals, but you're really like, you're using AI to
- 24:49really understand really deeply based on domain, sub domain and
- 24:53the actual content in the page really what's going on and and
- 24:56what this and what this inventory looks like.
- 24:58And instead of just like, you know, in a normal DSP looking at
- 25:00the audience and trust in the ID just to say, OK, this is what it
- 25:04is. You're you're able to better
- 25:06target the advertising and but then you're, it's actually
- 25:09running, it's running on the SSP, the supply side, but this
- 25:13is something you're selling them to the brand.
- 25:14He's actually buying media. Right.
- 25:16Yeah, yes. The ARTF was just a means of
- 25:20deploying a bigger model. And what the model does is
- 25:23predict the cost and value of an AD opportunity or of all the AD
- 25:27opportunities, right? So just bigger model.
- 25:30That was step one. Then we took it to IAB.
- 25:33I was able to convince Anthony, the head of IAB Tech Lab to make
- 25:38it a standard. The standard committee was very
- 25:41well attended. All the top publishers and
- 25:43Netflix and Amazon wanted to sit in the committee and their
- 25:47interest is super cool. They're looking for ways to
- 25:50control where their data goes, right?
- 25:53Because these infrastructure requires them to either withhold
- 25:56show level data from everyone or send it to everyone.
- 25:59This container starts to give them the capability to send it
- 26:03only to people they trust to properly value it, which is the
- 26:07opposite of what happened with newspapers, right?
- 26:09Nobody properly valued newspapers and they lost all
- 26:11their money. That continues, yeah.
- 26:13And still continues, right. So they're being smarter about
- 26:17it. They're interested in in the
- 26:18containerization. But now that now it is a
- 26:20standard, it's been adopted by Open X very successfully.
- 26:23Pub Matic is building it now. They'll be probably a couple
- 26:25more SSPS this year. So it's a very successful
- 26:28standard. And the road map is very
- 26:30exciting because what started as just a means to deploy a bigger
- 26:33model now becomes a whole other framework for companies to work
- 26:38together. Amazon has launched something
- 26:41really similar called RTB Fabric.
- 26:44The idea of a single server environment where the bid
- 26:47request can ping around to different partners who can
- 26:50change the bid request, augment the bid request, add data, and
- 26:55then a custom model can answer that data.
- 26:57So that implications are very profound.
- 27:00We're going from a world where there was one set of bid
- 27:02requests and everybody got it to every advertiser might see a
- 27:07bunch of see different bid requests depending on their
- 27:09partnerships and they could respond with custom models with
- 27:14with much greater ease than ever before.
- 27:17So they they have a lot more control over the decisioning
- 27:19decisioning logic before the bids actually.
- 27:22And the data one the data and the data logic so.
- 27:25Right, which allows them to it kind of almost insert for the
- 27:29simple minded, but insert some of the strategic planning, some
- 27:31of the context that they that only they have as the brand into
- 27:35that play so that they know which is going to drive
- 27:37effective outcomes, right. That's right.
- 27:39It's going to be more like you said that, Brett, because, yeah,
- 27:41because a lot of people say is it really agentic or does it fit
- 27:43the definition? But what what you said is what
- 27:45makes it agentic? What makes it agentic is that it
- 27:48will serve the advertisers intent right?
- 27:51A lot of people when they talk about agentic forget to ask who
- 27:56is it, whose agent is it, and what does it do for them.
- 27:59Yeah, the agent should be. Represented the agents like
- 28:02representing the interest of either the buyer or seller.
- 28:04I think that brings it up an important question Adam.
- 28:06So like what you've created, I mean, it's like bid management
- 28:09might be one of the applications of it.
- 28:11But in theory, like the that you're, you're giving all this
- 28:15incredible contextual information, Brett, to your
- 28:17point, the brand can then insert some of what they want to
- 28:20accomplish into this like like when looking at buy side and
- 28:23sell side agents me could just apply to different types of
- 28:26deals, Adam over time and different different ways of
- 28:29negotiating between buy and sell side.
- 28:30Oh absolutely. And not just buy and sell side,
- 28:32but all the partners in between and, and then rearrange them,
- 28:36right. So like an agency roster of
- 28:41partners can you can think of them as now being more or less
- 28:44drop or drag into this container environment.
- 28:46You know, the, the tools haven't been built where they physically
- 28:49drop and drag a a tile and it happens in some container, but
- 28:52it might only be a year or two away.
- 28:55And each of these, it's a very different way of working than a
- 28:58platform, right? Because everyone is modular,
- 29:01doing their piece in an environment they don't control
- 29:05and they have to serve the end user intent, which might change.
- 29:09So it's it really, that's what that's what being an agent is
- 29:12about to me. You're a modular service that
- 29:15changes depending on the end user intent.
- 29:18We're seeing a lot of things called Agentic that is just
- 29:20workflow automation. And maybe that is a kind of
- 29:22agent, but for me it's much more exciting to ask.
- 29:26When all the ad tech partners are agentic, it will mean that
- 29:30the end user can arrange them to do exactly what they want.
- 29:34Do you see a world where like there's all these kind of
- 29:36fragmented decisioning layers like like within within RTB and
- 29:41it's like the strategy's happening in the brand side,
- 29:43then you get the supply path and supply path optimization and
- 29:46then the bidding request and brand safety, which is a
- 29:48different vendor and then budget pacing, which is obviously
- 29:51within the DSP. Inefficient.
- 29:54Yeah. And then attribution which is
- 29:55another vendor all together. Do you think containerization
- 29:58can actually help bring this? Yeah, well, Chalice Alone,
- 30:01that's Chalice Alone's mission is to put all the decision and
- 30:04logic in one calculation. And we're getting there.
- 30:07We're getting to where we can do it.
- 30:09People don't realize that that most DSPS will do a a value
- 30:13calculation like do we want to win this one or not bid and then
- 30:18do a bid shed in calculation separately.
- 30:20That's extremely inefficient because the calculation doesn't
- 30:23catch the relationships between predicted value and predicted
- 30:26cost. And those are correlated, right?
- 30:28For some of this, this has been statistically proven that you
- 30:31shouldn't do those separate, you should do those together.
- 30:33But it's a lot more. Computational, yeah.
- 30:35Otherwise you might be buying some really shitty CTV inventory
- 30:38that's appearing on mobile devices and not on televisions,
- 30:41right? It makes the error very high
- 30:42stakes, like if, yeah, right. Like if everything is based on
- 30:47like what we do is we do value to cost prediction of everything
- 30:50and then stack rank them by the ratio.
- 30:52So the first things we're going to buy or the value to cost are
- 30:54furthest apart. Then, if you're off a little,
- 30:57it's still profitable. When these two things are
- 31:00separate, being awful little can cause you to miss an option.
- 31:02You're really supposed to win, yeah.
- 31:06Who is this approach to resonated the most with so far?
- 31:08I mean, it sounds for a brand perspective, it sounds pretty
- 31:10appealing. I mean, I also think for people
- 31:12on supply, right? I mean, it's because it's like
- 31:15it's demonstrating like the spec, well, let's say the people
- 31:17with high quality inventory, right?
- 31:20This must be great for them. Like what are you hearing from
- 31:22the marketplace as you talk to different people?
- 31:24A lot of excitement from agencies as a way that they can
- 31:27add value again, because it really plays to their strengths
- 31:29of having a full marketplace view and be able to offer
- 31:34different partners and different configurations and handle that.
- 31:38I think the agencies that are looking for a way to charge
- 31:41transparency for a service advertisers can't do themselves.
- 31:44Oh, transparency's a service. I love it.
- 31:47Right. I, I, we have a, we have a buddy
- 31:49that's like contributing to signal and noise.
- 31:52Mike who, who, who coined claw as a service, right.
- 31:56So everything seems to be as a service, but transparency as a
- 32:00service. That's an interesting play
- 32:01because that is a big we talk about all the time.
- 32:03We had Doctor Fu on the podcast. We've had a bunch of people
- 32:07about ad fraud and and how it continues to be this persistent
- 32:11problem in the industry and there's it's just a complete
- 32:13black box often times to the advertisers.
- 32:15Yeah. So I mean, live ramp Sidecar was
- 32:18the 1st commercial container and it adds a live ramp ID to a bid
- 32:21request in one millisecond. And Dr. Fu can have a container
- 32:25that adds his estimation of the percent fraud in one
- 32:29millisecond. And Christophe from Analytics
- 32:32can add his in the same millisecond, right?
- 32:34Because those don't depend on each other.
- 32:36So you could see it, it's endless.
- 32:38You could have hundreds of them going simultaneously and some of
- 32:41them have to be sequentially. Like Chalice isn't going to bid
- 32:43until we get all these. But that's what I mean.
- 32:45Like then we could calculate all at once, Foo's, Foo's
- 32:49estimation, Kristoff's estimation, the estimated cost.
- 32:53You know what Gareth said about this placement history?
- 32:55Like all in one calculation, that's the most computationally
- 32:58efficient, the most energy efficient, and it very much
- 33:01aligns with what we're seeing with AI.
- 33:03The reason these tools are so magical is they're doing tons of
- 33:06computation in the back end all together, right?
- 33:10The language. Scale and quicker.
- 33:12Yeah, it's amazing. Do you find that you're in
- 33:14clients of it? Is this a tough sell?
- 33:16Do you find yourself doing a lot of explaining so that they can
- 33:19sort of wrap their head around or or are there people in the
- 33:22industry on the DSP side or otherwise that are sort of
- 33:24pushing against this sort of new approach which is different than
- 33:28what we've seen before? Yeah, there's a lot of, I mean,
- 33:30that's, this is what I was talking about adoption.
- 33:32Like there's there's cultural barriers to adoption of new
- 33:35technology every time just because it's new and it disturbs
- 33:39workflows and it's magnified with AI because it's scary.
- 33:42Although people were scared of search, social and programmatic
- 33:45too. Not maybe not as scared, but I'd
- 33:49say we got the early adopters and Chalice was fortunate to get
- 33:52the attention of early adopters and get money flowing and
- 33:55quickly become profitable. And now like we, we know that
- 33:59the middle and late adoption is, is a heavy service lift, heavy
- 34:03evangelism lift, a heavy marketing lift.
- 34:06You know, we hope it's not so expensive that we don't need to
- 34:08throw giant parties that can, but we will if we have to.
- 34:13I saw you. I saw you a can last year.
- 34:15I think you're. We go, we bring a big team, but
- 34:17we haven't rented a big space yet.
- 34:19But it's like I say, they come from agency and and this is
- 34:23something we have to people need service.
- 34:25Like we have an excellent customer service team and we
- 34:28expect we're going to run 110K campaign and hopefully show them
- 34:32something I've never seen before.
- 34:34It's going to let. Winning hearts and minds, right?
- 34:37No, you had, you had a post that I read quickly yesterday.
- 34:40I think I don't know how recent it was, but you said, and I
- 34:42thought this is a little more theoretical, but I thought this
- 34:44was a really good point. I think it ties into what the
- 34:46point you're making about people adopting this in kind of moving
- 34:49through this sort of resistance layers of what they know or the
- 34:52workflows they currently work in.
- 34:54And you said, I just want to point out the idea that we all
- 34:56have that will have AI agents that can do things for us that
- 35:00that is not consistent with the idea that they'll do exactly
- 35:03what we do without them today, right?
- 35:06It's natural. The first thing people did with
- 35:08movie cameras was film plays before they knew what a movie
- 35:13was. So do you see us?
- 35:15I mean, to me, that's a, do you see that?
- 35:16We're in this moment right now where people are almost
- 35:19containerizing to use the word, their own view of what AI can
- 35:23do. And it's really just, it's sort
- 35:26of in a box. It's what we do currently today.
- 35:28It's just. It's just.
- 35:30Making sure it's. Compounding this.
- 35:31I don't get too sad about it, Mcluhan said.
- 35:33We drive into the future looking in the rearview mirror.
- 35:37Yeah. And that's true.
- 35:39Most people can't visualize things.
- 35:41Like I've been deep in this six years, so I could visualize it.
- 35:45I could explain it as well as I can, but the best I could do is
- 35:48show you, like, let me run a campaign of any size and
- 35:51hopefully I could show you something you've never seen
- 35:52before. I used to get more worked up
- 35:54about it, but it's like if someone's never had home
- 35:57cooking, they've only eaten McDonald's, Wendy's and Burger
- 36:00King. Like you could tell them about
- 36:01it, but you know until they taste it, they're not going to
- 36:04believe you. Yeah, well, but it's
- 36:08interesting. I mean, at the point you made
- 36:09bread about like when like when A is first coming out, like
- 36:12we're just going to try to use it to automate things we already
- 36:14do that maybe we shouldn't even be doing right or maybe they
- 36:16will look very different one once A is actually applied.
- 36:19I think Jeff Green set up market tech heard a couple people
- 36:21commenting on the crowd who's saying how like AI, like
- 36:25programmatic, fewer things lend themselves better.
- 36:28I'm just paraphrasing it to to AI to AI than programmatic.
- 36:32And well, then I guess I don't know, like programmatic is a way
- 36:35of automating things with right with primarily through real time
- 36:38bidding. So when we actually go a gentic
- 36:40trading, the do things look the same?
- 36:42I don't know. I don't know if you've had a
- 36:43point of view on that, but I think they might look different.
- 36:45I think there has to be an infrastructure upgrade.
- 36:47Yeah, scaled SAS is, is a, is a, a plumbing system.
- 36:52The scaled SAS of Nat tech is it is an outdated plumbing system
- 36:56that needs an upgrade. And it's not crazy.
- 36:58I think, I think Trade Desk has done big parts of this upgrade
- 37:00and and Index like kind of took a year off from promoting new
- 37:05products while they were internally doing these up.
- 37:08Re. Architecting and and it'll
- 37:11happen and that the cultural stuff is, is moving pretty
- 37:14quick. But there's something about the
- 37:16mentality of scaled SAS that's going to be tougher for the
- 37:20platforms. Like the idea that we're a hub
- 37:22of data and data comes in and we make the AI and sell the product
- 37:26back to you. That doesn't look as great an
- 37:29idea to me as it as it did 10 years ago.
- 37:31It's because it's, it's, it's not the only way and it's, it's
- 37:35probably not the best for a big advertiser because there's
- 37:38multiple hubs, if there was one maybe.
- 37:42But like I have to send all my data to Meta and to Google and
- 37:46to Amazon and. Multiple point solutions,
- 37:49multiple data warehouses as opposed to managing it all
- 37:52through one system. Whatever you are, data bricks
- 37:54house or snowflakes house and then plug in the application
- 37:58layers, the IL layers. Sometimes I lead a pitch with
- 38:01saying, hey, we estimated your data like a, you know, trillion
- 38:05events and two terabytes a day, and it's everything.
- 38:08It's massively effective for training AI.
- 38:10So whose AI are you training now?
- 38:16Trained to frontier. Trained to frontier model
- 38:17probably right? Let that sink in.
- 38:20I was like, oh, I don't know, I'll tell you.
- 38:23But yeah, but but but we're looking at like scaled like
- 38:26scaled SAS and the impact here. I mean, like, I don't, I mean,
- 38:29there are having people saying recently on, there's a lot of
- 38:32conversation on X. We'll get into the social thing
- 38:35in a minute. I know this is something we have
- 38:36to talk about. But but which I still call
- 38:40Twitter most of the time. But but, but, but you know, most
- 38:44of that tech is on there and people and people, people
- 38:46weren't like the other day saying, well, will there even be
- 38:49a, will there even be DSPS in 5 or 10 years?
- 38:51I think there's going to be disruption.
- 38:52But but you're looking at martech and I think martech's
- 38:54more probably going to be disrupted more.
- 38:56The martech stocks are really tank to tank to bit.
- 38:59I don't see like like it's an enterprise SAS going well
- 39:01anytime soon. But I think their ability to
- 39:03sell licenses and the whole effect, the model they go to
- 39:06market with and bill for may, you know, probably be disrupted
- 39:08to a certain extent. I think ad tech's less
- 39:11vulnerable if you ask me, because these are big models.
- 39:13They're fast. To your point, it's a lot of
- 39:15moving parts, but I do see the way people work, people work on
- 39:18these platforms changing quite a bit.
- 39:21Yeah. People aren't great at 2nd order
- 39:22effects. If the way things play out along
- 39:26the same track, it's going to be so much more competitive.
- 39:29There's going to be so many more companies and the ease of
- 39:34bringing a product to market and making it big, the cost of that
- 39:37is going to go down so much that our whole economy will look very
- 39:40different. So the question to me is, is
- 39:43will that be allowed to happen or a few big companies going to
- 39:46be able to shut it down? I hope it's allowed to happen.
- 39:50I mean, it would be, it would be a shame if it's not right.
- 39:52I mean, because you know the point we made earlier about it's
- 39:54so much cheaper to create a create technology.
- 39:57Yeah, we've had guests in compression.
- 39:59Yeah, we've had guests in this pod who are saying how like they
- 40:01like 111 dude can like do all of the work with clients, gather
- 40:07requirements, prototype design, vibe code, you know, say 6 or 7
- 40:12roles like people and they can do things in a couple weeks that
- 40:14would have taken months before. So it is kind of amazing how
- 40:16quickly things are going. And and the marketing like, you
- 40:20know, Facebook, Facebook has changed the economy and that
- 40:23anyone who builds any sort of gadget can market it on Meta and
- 40:26find their first, you know, 5010 thousand 100,000 customers for
- 40:31like a few bucks each. That's insane.
- 40:33But it's only Meta and only gadgets and direct to consumer
- 40:38energy drinks and stuff. But that could it could be the
- 40:40whole economy and a lot more than Meta able to do that.
- 40:43And we'll see a lot more companies in every category
- 40:47should. Yeah.
- 40:53So should. So should we pivot to Adam?
- 40:55To Adam the ship poster. Yes, do it.
- 40:57Yeah, we had a couple questions in there, Adam, so I'd love to
- 40:59hear you. Yeah, I'd love to hear your
- 41:00thoughts on it. I mean, I love, I love some of
- 41:02the wars you get into on, you know, with, with some of there's
- 41:05a lot of trolling going on and I'm nothing by you.
- 41:08But in general on on social media ad tech is is is the funny
- 41:13ad tech, you know, on Twitter slash X is a bunch of bunch of
- 41:15people. Yeah.
- 41:17Curious to get your thoughts on it.
- 41:19I know that you have some rules. You told me the other day we're
- 41:22hanging out at ramp up about like who you engage, who you
- 41:25engage with, who you just say I'm blocking you too bad.
- 41:28Like, maybe we'll start there. The rules of engagement for Adam
- 41:33Heimlich on social media. Look, I just, I think when I
- 41:37when I got into the media business in the 90s, it was not
- 41:40that professional. It was not an elite profession.
- 41:43And even advertising back then, it was starting to become
- 41:47professionalized and elite and the culture of the American
- 41:51elite that you're, you never act out and you never meltdown.
- 41:57It never really worked to me. It doesn't seem right to me.
- 42:00I think I think it's repressed and it causes problems.
- 42:03It's just disingenuous. Disingenuous and like, yeah,
- 42:06like a lot of America hates it who aren't involved.
- 42:09I hate it too. And yeah, it's just it, it's, I
- 42:13think it, it, it's sort of an aristocracy creates one thing.
- 42:16And I think, I think it creates, I think one of the reasons you
- 42:19see like really rich, successful people act out so much is they
- 42:22feel like I've, I've held in bad behavior so long and now I've
- 42:25made it and I'm going to. Who, Yeah.
- 42:27Who's going to restrain me? Like, who's like, like you
- 42:30can't. Nobody can fire me.
- 42:31Yeah, you know, and yeah, so it's so so that's, that's where
- 42:38I come from. I was raised in New Jersey and
- 42:40New York City family. And I don't think, you know, I
- 42:43don't want to hurt anyone's feelings.
- 42:45But if I'm like sort of rough with someone that I think their
- 42:47post is bullshit and they say so, I don't think they should
- 42:51think it's the end of the world, even if they think I'm a Dick,
- 42:53which is fine. But I don't think they should
- 42:55cry about it. Or if they do cry about it, they
- 42:57should get over that. So it's it's healthy debate,
- 43:01that's what it is. Yeah, like, but yeah, so that's
- 43:04that's said I'm, you know, I'm not always proud of my social
- 43:07media behavior and arguing. Sometimes.
- 43:08Sometimes I get I'm edgy, you know, sometimes I'm too and
- 43:12angry or edgy or things come out come out wrong.
- 43:15But most but my best, I'm trying to be funny.
- 43:18I'm trying to lighten the situation and add some humor and
- 43:21levity to these environments wherever people are promoting
- 43:25themselves or or say, you know, doing their best at saying
- 43:28accurate things but missing. Yeah, yeah, yeah.
- 43:31They're either shameless shills, or they're or they're or they're
- 43:34just inaccurate. Some people take they think
- 43:38themselves way too seriously. Some of these.
- 43:40People, yeah, I mean. It's.
- 43:42The deflating comments can be really fun.
- 43:44You know, you know, there's, there's other people engaged in
- 43:47the same way. But you know, if I make people
- 43:49laugh, sometimes I meet people at conference and they say I
- 43:52posted something that made them laugh.
- 43:53And that's, that's good. That's what I'm.
- 43:54That's what I feel like. Yeah, well, sometimes, sometimes
- 44:00people like us that are from the New York, New Jersey area, we
- 44:03can come across us, you know, we we show our cards and we play it
- 44:06straight. And yeah.
- 44:08And there's no obfuscation here. It's just, you know, this is,
- 44:11this is what it is. So New Yorkers called a spade a
- 44:13spade. I love that about they're
- 44:14brutally honest, which which is good to be being New Yorker
- 44:16myself. And but one of your rules, if I
- 44:19remember though, was like, like if if you tell me, don't sell me
- 44:21and someone continues, they're done, right?
- 44:24Yeah, yeah. Oh yeah.
- 44:25I I accept most LinkedIn connections, but if they pitched
- 44:29me I said I say not interested. Another pitch after that and
- 44:33mostly it's bots now anyway. I just blocked them, which
- 44:35resulted in me having like thousands of people blocked on
- 44:38that then. But there are also a few people
- 44:41who their commentary has annoyed me so much.
- 44:43I've blocked them and I've heard that's really hurt, hurt some
- 44:46feelings. I'll try not to do that.
- 44:48If you could unfollow them, you know, in ways you could get them
- 44:51out of your feet, in ways that don't they don't see and that
- 44:53hurt your feelings. But there's one but.
- 44:56If someone's gonna, if someone's gonna go in political
- 44:58commentary, like you gotta accept, that's part of it,
- 45:00right? If you're gonna challenge people
- 45:01and argue with them, they can block you, right?
- 45:03I mean, I don't go ahead. I cut you off, Adam.
- 45:05Oh yeah, political, yeah, but it was actually there's there's an
- 45:08there's an industry guy who every is posted is like the
- 45:11startup CEO should and the startup CEO must and the
- 45:14communications of the startup CEO.
- 45:16And I was like, I feel like this guy is like yelling at me like
- 45:20yeah, everything wrong. I feel like I'm getting
- 45:21constantly negative feedback so I'm.
- 45:24Sort of a dogmatic approach, right?
- 45:26Yeah. I mean, yeah, you don't have to
- 45:28say in my opinion before you say everything, but you could also
- 45:31be not so prescriptive, right about the.
- 45:35Yeah, and there's there's a generational thing too, like
- 45:38being being our age, like the whippersnappers like great, more
- 45:41power to them and you're up and coming in the industry.
- 45:44But sometimes, sometimes they're.
- 45:48Back to the cabin we made earlier about wisdom.
- 45:50Yeah, knowledge and wisdom are two different things, right?
- 45:54Yeah, fundamentally. And that it takes experience to
- 45:57know. Yeah.
- 45:58The difference I. Think I don't.
- 45:59I don't think that's bad. Like when I walked into a
- 46:01newspaper and I was if I said something, I thought I knew that
- 46:04the older people and I was wrong.
- 46:05The older people would not hesitate to mock and be down in
- 46:09the harsh as possible. And put you put you in your
- 46:11place. Well, yeah, well, but but that's
- 46:13the generationally. I remember my first couple of
- 46:14jobs being yelled at, being told I was garbage, like every.
- 46:18So some of it was really tough, but it did it did.
- 46:20It did make me like, like you got to be good at your job.
- 46:23Like I remember first got into consulting, right?
- 46:25Like Cartners would tell you this is crap like, like go redo
- 46:28this. I can't believe you showed this
- 46:30to me. And you're like, wow.
- 46:31Like I thought I was pretty smart.
- 46:33And you realize why I didn't know as much as I thought I
- 46:34knew, right. And then this person's been
- 46:36doing this forever. So I actually thinking I was
- 46:38good and I'm not, I don't I wouldn't treat anyone like that.
- 46:41But but it did make me better at my job.
- 46:43So I do understand looking back, why they did that.
- 46:46Yeah, you shouldn't shouldn't be harsh and hurt people's
- 46:48feelings, but in in writing, I mean, it's another difference,
- 46:53right? Because I was a writer first.
- 46:55I see social media as a writing exercise.
- 46:57And some people are thinking like, this is me, like this is
- 47:00actually me and this is me speaking.
- 47:02I don't think that. Yeah, I think this is your
- 47:04writing persona and I'm gonna respond in my media persona.
- 47:08Yeah, and that, and that's, that's back in comparative
- 47:11literature days. I mean, I studied rhetoric, you
- 47:13know, and that is that's what that that's rhetoric, right?
- 47:16Like I'm using a rhetorical approach that this is how I
- 47:18represent myself professionally on X.
- 47:21Yeah, it's my voice. Right.
- 47:22It's it's not who you are at home with your kids.
- 47:27Yeah, plus just Oh yeah. Plus that.
- 47:30It's knowing this is your work too.
- 47:31I mean, it's not you, right? To your point, Adam, so I
- 47:33remember when I first started writing professionally, bringing
- 47:36bringing drafts of things to the editor just crossed out, crossed
- 47:39out correction, correction. First couple times like, well, I
- 47:42thought I was a good writer. This kind of sucks.
- 47:43But you realize this is making me a better writer.
- 47:45It's not personal. So I don't think that's a
- 47:47personally now, right. And I wouldn't.
- 47:49So I don't think people should be so.
- 47:50I got some cloud feedback the other day that really made me
- 47:52want to punch cloud in the face. That's awesome.
- 47:58Yeah, Yeah, the so, so, so should we go to some of our some
- 48:01of our questions? Yeah, Quick.
- 48:03Yeah. So, so which part of the ad tech
- 48:05stack do you think is most? I mean, we've been talking about
- 48:08AI, we've been talking about containerization in your in your
- 48:10solution is most at risk being replaced by DSPSSSPS measurement
- 48:16vendors. We talked about incrementality,
- 48:18whatever we want to call that, media agencies.
- 48:20Where do you see the the biggest sort of weaknesses?
- 48:24I think, I think s s PS are ahead right now and what and
- 48:28what index and open X and some others are doing, We'll give
- 48:31them a new breath of life. But ever since header bidding
- 48:33came out, I think their position has been a little bit
- 48:36precarious. Like what really is the
- 48:38function? They're not doing yield
- 48:40management. They don't have original bid
- 48:42requests. Like what are they really taking
- 48:44money from? And you know, shifting to say
- 48:48like we're infrastructure for all kinds of of agentic flows at
- 48:52the bid request is a good one. But you know, how many of those
- 48:55are are we really going to need? I think, I actually think the
- 48:58DSPS at the moment are more antiquated and maybe more
- 49:02threatened. They have a more or less
- 49:05permanent foothold in being the, the buy side source of monetary
- 49:09truth, the budget truth and pacing, which is a really
- 49:12important function that's hard to do at the edge, you know,
- 49:15because you need like the, the steady state of the total
- 49:18budget. So I think, I think buy side
- 49:21tech is a little more stable than than sell side tech.
- 49:25I think most measurement and data companies better be ready
- 49:28for a price war. Everything in there is going to
- 49:30be heavily commoditized. They'll all they'll all survive
- 49:34and have a function. But it's already the case that
- 49:37people don't really want to pay. I don't believe in premium data
- 49:41or premium measurement. That's just.
- 49:43Good. Yeah.
- 49:44And a lot of those sort of advanced analytics and when I
- 49:46came from the new star side of the fence for a little bit.
- 49:49So the market share solution analytic partners and a lot of
- 49:52these are kind of heavy implementation load weights,
- 49:56they're heavy long managed services cycles, right?
- 50:00And I just don't think that the model's getting blown up.
- 50:02Yeah, that's yeah. It's taking, it's too expensive,
- 50:05Yeah. It's consulting with a little
- 50:07bit of tech to kind of, but it's Yeah.
- 50:09And. It's taking it take it six
- 50:10months or eight months in order to like you have people come in
- 50:13and audit all your data and collect it.
- 50:14Yeah, hey, I can do that so much.
- 50:16Quickly see what happens. They don't get worried, Yeah.
- 50:19Yeah, but but it's like s s PS I mean, I would have, you know, I
- 50:23don't know. I mean, a year ago, I remember
- 50:24there's a debate about are they all, what is it?
- 50:27Are they all resellers right. And then but but I don't know,
- 50:31DSP is like with some of the changes are maybe looking a
- 50:34little more vulnerable, but I think that is interesting.
- 50:36But looking at if you had to place a bet like will RTB still
- 50:39be the dominant buying model in 10 years?
- 50:41I mean, I think it's like 6065% right of all programmatic as RTB
- 50:46right now. I mean like like like will like
- 50:48will that change in the future as we go to a gentic?
- 50:50Will there be, I mean, obviously there's different types of like
- 50:53like deal structures, but like do you think that changes
- 50:55dramatically? Yeah, I think, I think it'll
- 50:57split along enterprise and S MB. I think enterprise wants a
- 51:01really big auction, really needs it and to and to be able to
- 51:06fiddle around in there and on their models and that it doesn't
- 51:09matter so much in SMB and they could, they could do more behind
- 51:12the scenes. So I think, I think they're, you
- 51:15know, it's kind of like finance, like there will be a market
- 51:19that's for big players, for for high rollers.
- 51:21Like it's like there isn't finance and they'll be more
- 51:24commercial markets for everyone else.
- 51:28So what's the worst take you see repeated on LinkedIn talking
- 51:31talking about things that perturb?
- 51:34You repeat it. What's the worst take you've
- 51:35seen? Period.
- 51:36Yeah, worst thing you've seen either repeated or period about
- 51:39AI and advertising. When they compare platforms and
- 51:42they're like this one has the data and this one doesn't have
- 51:44the data. That is so stupid and I'm so
- 51:47tired of it. Like the this idea that some
- 51:50there's some data ownership that's determinative of success
- 51:53in a platform. This got disproven by Verizon
- 51:57and AT&T like, and you know, so many companies that tried to go
- 52:03into ad tech and and copy Google.
- 52:05It's, I mean, Google has. Where they can build a mode
- 52:09around some sort of data advantage that doesn't really
- 52:11exist because the data is relatively ubiquitous and it
- 52:13lives in big cloud. Infrastructure, yeah, you hear
- 52:16it over and over in the trade discreticism.
- 52:17Well, they should buy someone that has data.
- 52:20What is that going to do, going to piss off everybody in the in
- 52:24the other side of the market? Yeah, they've got the same data
- 52:27as everybody, they're saying. They're saying that's the big
- 52:29problem with the Spotify in the Amazon musics of the world,
- 52:32because you're not getting anything unique everybody's got
- 52:35exactly. The same.
- 52:36Catalog of music right, that's their data and it's like you got
- 52:40that repeated over and over again so each one of these
- 52:42services it's just a race to the bottom from.
- 52:44A price, Yeah. Data is commoditized.
- 52:46Modeling and service are not, and a lot of people have that
- 52:49backwards. Should read the Google trial.
- 52:51They had the most data by a mile and they had this whole stack
- 52:55and they still had to break the law.
- 52:59Modeling and service is not imminently replaceable.
- 53:03And that's, I think that's key. That's AI plus the human kind of
- 53:06orchestration, right? And conducting.
- 53:08That's what you'll be able to charge for, yeah, I think.
- 53:11Yeah. Well, there's definitely more
- 53:13and more demand for specialized, you know, advisory and services
- 53:16around, you know, I mean, not only AI, but I think tech in
- 53:20general. I think that's the interesting
- 53:21thing, although AI is disrupting providers of this to a certain
- 53:24extent and their business models is more demand than ever.
- 53:27What gets overlooked a lot is that the adopters will make a
- 53:30lot of money. And Jerry, the investor, Jerry
- 53:32Newman has made this point about shipping containerization.
- 53:35Shipping containerization is like AI in that it added a
- 53:38massive amount of efficiency throughout the industry and it
- 53:41didn't change who the big players were.
- 53:43Just like in AI, Amazon was big before.
- 53:45Amazon's going to be big in AI, same as Google and Microsoft.
- 53:49But IKEA, who adopted shipping containerization to become a
- 53:52global shipper, right, changed, changed everything for them.
- 53:56And Walmart who's has just said we're going to have the cheapest
- 53:58prices on everything no matter what, right, becomes much bigger
- 54:02and more profitable. So he's his point was like for
- 54:05an investor, it's actually really hard to pick winners
- 54:08because AI and service hard as hell to predict.
- 54:10And how big do they get anyway? Too late to invest in the models
- 54:14and the adopters is a crap shoot.
- 54:16Like unless you have inside information you don't know who's
- 54:19going to be the IKEA or Walmart of AI.
- 54:22That sounds like the sort. Yeah, that that that's like the
- 54:24source of your idea here, right? Were you thinking of the shower?
- 54:27You're like containerization. Think about what they're doing
- 54:29and how do we actually apply that in the text space of it.
- 54:32It makes sense. It's very.
- 54:33Fortuitous this article is articles called AI will not Make
- 54:36You Rich. Brian o'kelly posted it and I
- 54:38was like God this is like the smartest thing I ever read.
- 54:41I read it a bunch of times. Thank you Brian for posting it.
- 54:44Then I talked about it on my show and I was like, I was like,
- 54:47I was like, I've been fascinated with this article.
- 54:49And then this man, I looked him up and he was he was Corp David
- 54:52Omnicom and then early investor in the Trade Desk and became an
- 54:56investor. So after the show that I
- 54:57mentioned it, I got an e-mail from Jeff Green with him, CC D
- 55:00and was like Adam beat Jerry, great guy Jeff.
- 55:05So I got to meet Jerry after studying his article and I got
- 55:09to talk to him for an hour. But an incredible thought.
- 55:12And yeah, a funny coincidence that shipping containerization
- 55:14includes that word containerization.
- 55:17Well it's crazy to think that they ship people used to ship
- 55:20stuff and just like I guess manually just stack it in ships.
- 55:23Right. In holes out of the.
- 55:24Box and put them in another box just to take it off the ship
- 55:29out-of-the-box, then in another box and on the truck.
- 55:32And this people come along and said why don't you just ship it
- 55:35on the truck box or the train box?
- 55:37And it reduced shipping costs at least 70%.
- 55:40Yeah, well, it's funny, a friend of mine used to at his parents
- 55:43was I guess his family business at the port of Munich.
- 55:45No, no, port of Hamburg was like they would because the ships
- 55:48would come in. There's always if there'd be
- 55:49damaged, there's stuff with some pallets back then, this is pre
- 55:51containers, right? Stuff would move around and get
- 55:54damaged if there was a storm. So a lot of these pallets would
- 55:57be write offs and they would just buy them in bulk and then
- 55:59sell them through these retail and we never know what they
- 56:01would get. So, but that whole business went
- 56:03away when they switched to the. Container, yeah, I mean, that's
- 56:05the cultural adoption. But if your job is to take the
- 56:08things off the boat and put them on the truck, you're not going
- 56:12to love this new technology. It might disrupt you.
- 56:16It took so, yeah, it took a few years.
- 56:18But once you know, it's inevitable that it's going to be
- 56:21completely adopted, What's his point?
- 56:23Because it's so much cheaper and the cultural resistance can't
- 56:26hold out over a critical mass, That's what's going to happen.
- 56:32So I'm looking over the next few years, like any big predictions
- 56:35or are there any misconceptions you think people have that'll
- 56:38that'll get cleared up? Like where do you see things
- 56:40headed? I think that a small business
- 56:42enterprise split is starting to happen and then a lot of people
- 56:45have missed it and they talk about advertising like it's,
- 56:48it's unified across those two. And it has been largely because
- 56:51of Google like, but that's a really weird thing that the
- 56:54Coca-Cola Company uses the same tool as your pizza place.
- 56:59That's not the norm. Well, for social and for Google,
- 57:03but how about for like for TV? I think that's interesting.
- 57:05I mean, like, I, I actually think that the opposite might
- 57:07happen, right? Like some of the like the like
- 57:09the generative video video tools might actually help small
- 57:12businesses, Yeah. Because it reduces production
- 57:15cost of the actual commercials themselves.
- 57:17And then when you have programmatic inserts at the
- 57:19local level, right, it does sort of democratize advertising.
- 57:23Yeah, I would say it'll be easier.
- 57:25To get into the big, it'll be easier to get a ticket into the
- 57:27big company markets than it was before, but that'll be
- 57:31substantially different than the the small company markets, and
- 57:35for the better, because the degree of complexity they need
- 57:39has not been met by Google and Meta.
- 57:43Got it. So yeah, I think that I think
- 57:45that covers everything, right? We've we've we're a little bit
- 57:47over an hour. Thank you, Adam.
- 57:50This is great, super fun, very well read.
- 57:54I can see all the books in the background.
- 57:56And for those that didn't know the he's got he was a green
- 58:00screen, right? That he that he customized, but
- 58:03it was awesome having you on the show.
- 58:04We'll probably see you at possible.
- 58:06I'm sure you'll be there. Everybody's going to be there.
- 58:08It's the tan of the US now in Miami, April 27th Rio and I will
- 58:12be there for signal of noise. Yeah, it will be quite as it
- 58:14will be quite as hot as can. It'll be great.
- 58:17Yeah. And for all of those that made
- 58:20it this far, the episode, this was a pretty quick episode.
- 58:22Sometimes we go for an hour and 15 hour and 20 minutes.
- 58:25Yeah, be sure to rec recommend Signal and Noise to all your
- 58:28friends and colleagues. You can find us at Signal and
- 58:31Noise. Smash that subscribe button.
- 58:33Follow us. Yeah.
- 58:34You too ad talk too. Ad talk as well.
- 58:39Yeah, and every, every extra Adam ad tech ad talks every
- 58:41Friday, right? It's a lot.
- 58:42It's a live event, right? Fridays at noon Eastern.
- 58:45And if people want to find it, how do they do that?
- 58:47Do they go, do they follow you in LinkedIn?
- 58:48Do they? Works out at YouTube, Yeah.
- 58:50Or LinkedIn. My name's Adam Heimlich.
- 58:53OK, cool. And if anyone.
- 58:54Wants to get a hold of chalice and like contact you about to
- 58:56learn more about this containerized approach or
- 58:58anything? You could write me at Adam at
- 59:00Chalice AI or LinkedIn, but don't pitch me.
- 59:04Yeah, exactly. I connected with Adam, but I did
- 59:07not send him a pitch. I knew.
- 59:08I knew his his reputation preceded him, but I warned him.
- 59:13My homework, Adam, is that, you know, one of the big takeaways
- 59:15here. But just generally, your vision,
- 59:17your founder growth here and and your ability to run a business
- 59:20with your wife, I have huge respect for.
- 59:22Yeah. That's terrific too.
- 59:23It's very. Yeah, yeah.
- 59:25Is this is AI will not make you rich.
- 59:28I think that's an article we should all go out and read.
- 59:30Yeah, it sounds like it's super interesting.
- 59:33So I'm I'm definitely going to take a take a read of that over
- 59:35the weekend. Thank you.
- 59:36We'll we'll talk to you soon. All right.
- 59:39Thank you. Adam.
- 59:41Thank you.