Latest / Signal & Noise: Bringing Clarity to Media, Data, Tech and AI / Referee to Co-Pilot: Ad Verification in the Age of AI with Mark Zagorski, CEO of DoubleVerify
Transcript
- 0:03This is signal and noise, bringing clarity to data, tech
- 0:07and AI. Welcome Mark Zagorski to Signal
- 0:16and Noise. It's great to have you on the
- 0:17podcast. Thanks for having me, Brett.
- 0:19Good to be here. Hi, Rio.
- 0:22Good to see you. So for those that don't know
- 0:24Mark, Mark and I actually go way back to the beginnings of my
- 0:28career in AD tech at a company called the little company called
- 0:31Exelate and sort of traverse the data marketplace, data
- 0:35management platform worlds before being purchased by
- 0:38Nielsen and launching the Nielsen Marketing Cloud.
- 0:42And then Mark did a whole bunch of terrific things.
- 0:45People probably know him as the CEO of Double Verify.
- 0:47But in your own words, Mark, I'd love to have you share with our
- 0:51audience a little bit about your trajectory in Martech ad tech
- 0:54and kind of what brought you to your current five year plus
- 0:58stint, right? Yeah, no, I, I mean, look, I've
- 1:01been swimming around in this Internet ad pool since 1997
- 1:08where I was with Modem Media, one of the first digital ad
- 1:12agencies. You some folks who are who are
- 1:15long in the tooth may remember those, those that, that crew and
- 1:20yeah and and most recently sold XLA to Nielsen then ran a
- 1:25company called Talaria small public company, which we merged
- 1:28with Rubicon to create Magnite. And yeah, and formerly Tremor
- 1:34video, right? Formerly tremor video.
- 1:36Yep we it was tremor. We sold off the the buy side and
- 1:41just focused on sell side. Turned that to Talaria and
- 1:44Talaria merged with Ruby became Magnites.
- 1:46And then I turn that over to Mr. Michael Barrett, who's done an
- 1:51amazing job with that business. And then most recently in the
- 1:55last five years IPO double verify, that's where we are
- 1:58today. So it's been a, it's been a
- 2:01really great kind of decade or so of selling companies, merging
- 2:06companies and now IPO and companies.
- 2:09That's awesome. Yeah, well, your reputation
- 2:12definitely proceeds you, Mark. We're thrilled to have you on on
- 2:15the show and welcome everyone back to Signal and Noise.
- 2:18So again, this is Ria Longacre here with my Co host Brett
- 2:20House. And as you can see, Mark
- 2:22Zagorski's on. We're going to talk into and dig
- 2:25into the state of ad verification.
- 2:27And I guess you could say Mark is really in the Co double
- 2:30verify. He's right in the pilot seat,
- 2:32right? Mark, we're thrilled to have you
- 2:34here. And I just think this is such a
- 2:36very important topic, especially like looking at ad verification
- 2:40and how this like relates to things in the age of AI, right?
- 2:44Anyone who's followed this space knows this has been a it's been
- 2:47a lot going on this year DV, like I posted some really good
- 2:52Q2 results stock bumped quite a bit.
- 2:55Looking at the numbers here is 21% revenue growth guidance went
- 2:58up and the market notice shares, shares went way up and I, I, I
- 3:02wish I'd bought more stock before that trajectory looks
- 3:06good, but in a lot of. Ways is.
- 3:10Verification isn't just a victory lap.
- 3:12I mean, I, you know, we, there's been a lot of, I think questions
- 3:15in the industry about where things are going.
- 3:17A few weeks ago we had a, a guest in our show doctor, Doctor
- 3:21Augustine Fu, who, you know, many know he's been a sharp
- 3:24critic of brand safety tactics, not specifically of, of, of
- 3:28double verify, but just in general.
- 3:30There's been argument for him as well As for many publishers that
- 3:32some blunt keyword blocking can drain revenue from quality news
- 3:36and maybe even sometimes miss real fraud.
- 3:39So like we want to put address all these concerns, talk about
- 3:43them and understand what D VS doing, where the changes are
- 3:46being made and why, why, why D VS leading the pack and how it
- 3:51relates to this recent good news.
- 3:52We're also dig into some of the Double Verify's recent moves,
- 3:55including the Cyber's, a acquisition, which I thought was
- 3:57really cool and one of the, I think a harbinger of a lot of
- 4:00the M&A activity we've seen recently.
- 4:04In addition to the auction time optimization of and as well as
- 4:08rocker box for attribution and what these bets say about time,
- 4:12trust, outcomes and the whole age of outcomes thing.
- 4:15I think industry never sits still.
- 4:17A lot, lots of change, a lot going on, especially with AI and
- 4:20really a lot to talk about. So, you know, welcome here,
- 4:24buckle up everyone, and let's dive in.
- 4:26Yeah, you really. You really kept that narrow,
- 4:28Rio. We're really gonna.
- 4:30We're gonna. There's a broad set of topics
- 4:32and. Like I think we got like 5
- 4:34minutes in. What we'll be Oh, you did say
- 4:35anything was good. You said you were.
- 4:37Available to like 5:00 PM Eastern, right?
- 4:41So, so you know, you guys have been, I've heard you say the
- 4:44word referee, right? And you know, you and I having
- 4:46spent had some stints at Nielsen, you certainly,
- 4:49certainly know what that sort of neutral arbiter role is.
- 4:53As as you know, and I love that analogy for for DV, where do you
- 4:57draw the line between sort of neutral arbiter and sort of
- 5:00business partner with, with brand clients, with agency
- 5:04clients without compromising your independence as a, as a, as
- 5:08an. Arbiter yeah, I mean, look, we
- 5:10we have one main constituency who we serve and that is the
- 5:14advertiser, right. You know, over over 90% of our
- 5:18revenue comes from advertisers and our goal and not our goal,
- 5:23our mission is to be an independent, you know, partner
- 5:28to them to help them assess the quality and efficacy of their
- 5:32media placements, right? There are many platforms that do
- 5:36what we do, but they're also selling media, right?
- 5:40There are many buying and selling platforms that do what
- 5:42we do, but they have skin in the game, which is they have a
- 5:46vested interest in making sure that transaction happens.
- 5:49We don't. We have a vested interest in
- 5:52making sure that the data we provide the advertiser is
- 5:55accurate and helps them make the best decision possible.
- 5:58So, you know, there is no line in which, you know, we feel like
- 6:04we have to cross to be, you know, independent or good.
- 6:08Like we're already there. Like that is our mission.
- 6:10If if our advertisers don't trust us that we're doing the
- 6:13right thing, then they won't spend with us.
- 6:15And that is key. Our independence, the veracity
- 6:18of the data that we have and our focus on ensuring that they are
- 6:22getting the right information to make the right decision is is
- 6:26essential to who we are as a company and our value prop to
- 6:28anyone who works with us. Well, results speak for
- 6:31themselves, that's for sure. Looking back to the outstanding
- 6:34results last quarter, St. notice stock bumped.
- 6:37What do you attribute some of this success to?
- 6:40So, I mean, look, we, we started off the year saying this is a
- 6:43year of transition for us and it, and it's still, it's we, we
- 6:46are investing in solutions for both CTV and for social
- 6:50platforms because we believe, you know, more and more dollars
- 6:53are flowing there and there's less and less transparency.
- 6:56So you know, we know that this year was going to be 1 in which
- 7:00there was going to be some variability, right?
- 7:02We started the year off saying that we benefited from some
- 7:07active behavior that we had not with new products which are
- 7:11still being developed, but with kind of selling our vision to
- 7:14new customers and upselling those solutions to to our
- 7:19current customers as well. We had some really strong
- 7:21strategic wins over the last several quarters.
- 7:24Microsoft can view Charter. All of these companies, you
- 7:28know, came to start working with TV and we were able to upsell
- 7:31them some, you know, new solutions.
- 7:33So we benefited from our focus on, you know, selling the vision
- 7:38of where we're heading as a business, bringing new customers
- 7:42into the fold and taking advantage of, you know, some of
- 7:46the growth opportunities in upselling products that have
- 7:50arguably been around for a while, but have not had
- 7:53penetration with some of these new customers that we were just
- 7:55on board. And so we had a good first half
- 7:58of the year. I think we will continue to show
- 8:01strong growth, you know, based on the the the evolution of our
- 8:06solutions into other venues over the next several quarters as
- 8:10well. I've heard you talk a lot about
- 8:12like moving from just branch safety protection and
- 8:16verification into more performance outcomes, really
- 8:20looking at things more holistically, like what, what
- 8:24solutions are you bringing or problems are you think you're
- 8:26solving for CM OS? And you mentioned the brands
- 8:28themselves that that they weren't, that maybe even weren't
- 8:30solvable a few years ago. Yeah.
- 8:33So you know, the the evolution from us being what is
- 8:36classically known as kind of a protection business to one that
- 8:40also kind of has performance as part of his mission is not as
- 8:44big of a stretch, I think as some might believe.
- 8:47We started off by saying we're going to help you take garbage
- 8:49out of the system, right? Things Ads that aren't viewable.
- 8:53Fraud. Ads that show up in content or
- 8:56context that's not aligned with who you are as a brand.
- 8:59It's an efficiency plan. Yeah.
- 9:01When you take garbage out of the system, what's left is going to
- 9:03perform better, right? And I think that's the nature of
- 9:07who we've always been as a business.
- 9:09Now Fast forward to today, you know, we're helping you find
- 9:12media quality, but with the addition of Cybads, we're
- 9:15helping you find that media quality of those impression
- 9:18quality, that impression cheaper, right?
- 9:21And now with the addition of Rocker Box, we're going to help
- 9:24you prove that that impression or understand whether that
- 9:28impression actually drove a sale, right?
- 9:30So if you think about it, it's the natural evolution of quality
- 9:34data, finding quality environments, finding that
- 9:37equality environment cheaper and then proving that that quality
- 9:40environment or that impression actually drove a sale.
- 9:43So it's an expansion of our core, but we've always focused
- 9:46on efficiency, pulling garbage out of the system and helping
- 9:50what's left work better. The yes, I bids that was an
- 9:54interesting acquisition. I figure we could jump ahead to
- 9:56some of the questions we had about that Rio was and that's
- 10:00those are custom bidding models, right based on on signals like
- 10:05viewability and attention and IVT and and other.
- 10:09Yeah, I, I feel, I almost felt like that that acquisition
- 10:11kicked off like the acquisition like we've seen a lot of them
- 10:15the last like let's say year or so, especially in the AI space.
- 10:19I thought that was like a harbinger, right?
- 10:20What you like? Not a lot of people predicted
- 10:23it. I'd love to hear a little bit
- 10:24more about your rationale for making that move.
- 10:26Yeah, I'll be direct. Like the initial rationale for
- 10:29that acquisition was more defensive than offensive.
- 10:31And what we saw was, you know, we were in the world of pre bid,
- 10:38you know, pre bid inventory being very binary.
- 10:40It was either viewable or not viewable.
- 10:42It was either brand safe or not safe.
- 10:44It was either fraud or not fraud and we had this binary kind of
- 10:48take on how our solutions were implemented.
- 10:51And what we, what we started theorizing is, well, viewability
- 10:56is binary. But when you start looking at
- 10:58the way IB design defines it and how other advertiser defining
- 11:02it, there is actually a scale of viewability.
- 11:04And what we've like, what the IB is, society is kind of
- 11:07arbitrary, right? Anyone in the media knows.
- 11:08It's kind of like there's this line of like, hey, the IB says
- 11:12this is viewable or not viewable.
- 11:14And I think some of our customers started feeling the
- 11:15same way, not around things like fraud or brand safety, but
- 11:19things like suitability and viewability where there's a
- 11:23spectrum of, hey, you know, what if I could get this ad
- 11:27significantly cheaper and it actually has ROI against it,
- 11:33maybe I'm willing to not like to not look at viewability as
- 11:37binary, right? Maybe an ad that didn't meet the
- 11:40binary binary definition of viewability actually still
- 11:46displayed long enough for someone to click on it so that
- 11:50it actually did do something. And if I can get that super
- 11:53cheap, then why wouldn't I buy it?
- 11:54So we started thinking like, you know, is this the way the world
- 11:58is going to go? Is all buying just going to be
- 12:00more on a spectrum, more on versus binary?
- 12:03And that was the initial thesis behind Cybiz, which is like,
- 12:06let's protect what we have. Yeah, but and it also moves you
- 12:09a little bit into the activation space, you know, and you have
- 12:12this auction time optimization, which I think it reminds me a
- 12:15little bit real of our conversation with Luke Pascalis,
- 12:17who was, as you can imagine, Mark was a fascinating
- 12:20conversation. And he talked about the
- 12:21criticality of sort of signal, whether you're on the publisher
- 12:25side or the brand side, like harnessing and capturing the
- 12:29signal. Forget about just first party
- 12:30data, we're talking about third party data signal.
- 12:32And it's is is critical for success, for relevance, for
- 12:36content optimization, ad optimization, auction, you know,
- 12:40real time optimization. Is that really the goal is to
- 12:42increase the amount of signal that you're capturing I think.
- 12:45It's, it's, it's to not to increase the amount of signal,
- 12:48but to increase the usability of our signals.
- 12:51Yeah, more applications. And you know, you know, I
- 12:54mentioned viewability as one, but for the vast majority of
- 12:57customers of Cybits, they're using their own signals, right?
- 13:01So whether it's someone like Diageo who has their own, you
- 13:05know, their own KPI's, the idea behind side is it's a custom KB
- 13:09IKPI engine, right? And it builds, it builds an algo
- 13:13based on whatever that advertiser wants to to focus on.
- 13:16It could be a DV metric like attention, but it could be
- 13:20entirely their own metric. It could be sales, it could be,
- 13:23you know, clicks, it could be anything.
- 13:25It could be viewability. But the idea there is.
- 13:29So that algorithm allows you to suck in data from the client,
- 13:32whether it's whether it's their own first party data or part of
- 13:35their data ecosystem, which could be third party data
- 13:37partners through a cleanroom type of thing.
- 13:39So you can bring all that in there and then customize to the
- 13:42variables that are driving the most impact.
- 13:44Has that been combined as well with TV signals?
- 13:46I mean, that sounds interesting. Yeah.
- 13:49And and if you kind of play this out, think about rocker box now.
- 13:53Now we have the box which is pulling in attribution signals
- 13:57and all kinds of other transactional data that can also
- 14:01be optimized again. So you're right, you know,
- 14:03there's, there's accumulation of more signal that we're getting
- 14:06through rocker box and then there's the application of that
- 14:09signal in more, more places through silence.
- 14:13And you know, it's cool. So I think, you know, it makes a
- 14:16really powerful package for DV to talk to customers around.
- 14:20And it sounds like it's a, it's a real diversification play.
- 14:23It's like don't like limit your risk in terms of where you
- 14:26played and how you're being defined in market today, like
- 14:28competing against IAS and broaden the scope of what your
- 14:32business actually does to support brand efficacy and
- 14:35effectiveness, right? Yeah.
- 14:37You know, look, I think Rio, you mentioned earlier, it's kind of
- 14:40like what problems are you trying to solve?
- 14:42And I think, you know, CM OS and advertisers are always trying to
- 14:45solve the quality problem and they're trying to solve the
- 14:48performance problem, right? You know, does my, is my media
- 14:51in the right place so that I can protect my brand and does it
- 14:54actually work? And if they can do that through
- 14:57a single platform and a single customer relationship, they'd
- 15:01much rather to do that than work with 15 other different places.
- 15:05We also occupy a pretty unique spot in the fact that we're not
- 15:10just open web, right? We're not just CTV or just
- 15:14social. We see data and transactions
- 15:16across open web, across mobile web, across mobile app, across
- 15:21social, across CTV, across audio like and we do so and that we
- 15:26know reason why we can be there is because we're agnostic and
- 15:29we're not competing against any of those guys.
- 15:31We don't compete against Meta, we don't compete against
- 15:33Netflix, we don't compete against Trade Desk.
- 15:35And that puts us in a really unique and agnostic position.
- 15:39Is, are these acquisitions changing who you're selling to
- 15:42to a certain extent like agency versus brand CMO?
- 15:46I mean, is it? Has it changed that dialogue at
- 15:48all? I think it's brought more folks
- 15:52into the discussion, right? So.
- 15:55Function more functional folks from.
- 15:57Different areas, you know, when you start talking about
- 15:59optimization now you're now you're pulling traders and, and
- 16:01media buyers in. You know, brand safety in many
- 16:05cases was a discussion at the sometimes at the CMO level,
- 16:10sometimes at the media level. And in some cases they actually
- 16:13had specific brand safety leads, right.
- 16:16You know, the brands had their own brand safety folks.
- 16:19This now has now brought in performance folks, analytics
- 16:23folks. Which changes the nature of the
- 16:26conversation too, right? You have to get a lot more
- 16:28technical a lot faster with those data science gurus that.
- 16:31Are a lot more technical, but also a lot more bottom line
- 16:34driven to like, show me what you're going to save me.
- 16:37Show me what results you can help prove.
- 16:40And I think that's, again, I think that's good.
- 16:43We want to have those discussions with advertisers
- 16:46because, and we've said this before, it's kind of like you
- 16:50grudgingly pay for insurance, I'm sure.
- 16:53Neither of you, I think. I think I've heard you switch to
- 16:55that before. In public today, neither of you
- 16:57love paying for insurance. No one does.
- 16:59But you love like paying for performance like this is great.
- 17:02I want to work with you. I want that to happen.
- 17:04So. You love with with the insurance
- 17:06actually kicks in to support you in a time of need, Yeah.
- 17:09It's good when it works, but why am I?
- 17:12Paying this amount, where's the value of this?
- 17:13Yeah. So, so for the for our audience,
- 17:15because we should we define cybeds for for the audience, for
- 17:18rocker box. And this is sort of close to
- 17:20heart for probably the last five to seven years of my life at New
- 17:23Star TransUnion heavy duty MTA, we did obviously some of that in
- 17:28MMM and some of that at Nielsen with Visual IQ.
- 17:31We are we are commercializing that as part of the Nielsen
- 17:35marketing cloud as well as their home built CPGMTA solution.
- 17:38But just for the audience and tell me if I'm wrong Mark, you
- 17:42know rocker box is sort of a consolidated solution.
- 17:44It combines the capabilities of MTA classical multi touch
- 17:48distribution and marketing mix modeling and incrementality
- 17:52testing, right, to give you sort of a single source of truth.
- 17:56Now that's kind of a positioning statement and I apologize for
- 17:59that. But can you break that down for
- 18:01the audience in terms of the benefit 'cause I know the full
- 18:04capabilities of these types and they can get very heavy and very
- 18:08deeply algorithmic and very predictive and complex and hard
- 18:12to implement. Can you kind of talk us like how
- 18:15the how Rocker box sells into brands and agencies and how do
- 18:18you see it in the ecosystem compared with both the open
- 18:21source models like the Meridians and the and the Robins from Meta
- 18:26as well as like the big players like TransUnion and IRI?
- 18:29And, you know, there's a bunch of players in that, in that
- 18:31space, econometrics and. It's interesting too, Brett, you
- 18:34bring this up and I'm curious to hear what you have to say about
- 18:35it more because I felt like MMTA attribution in general, lots of
- 18:39questions about it. These are coming up our people
- 18:42gaming attribution and you know, with signal loss, like is that
- 18:45is there a feature for MTA? And I saw a lot of brands start
- 18:47to really shift focus. Two more.
- 18:50MMM, right. And maybe we should you know,
- 18:51which I think it kind of fallen off for a while, right.
- 18:53So I thought that was kind of refreshing and you know, like I
- 18:56think end of last year, beginning of this year,
- 18:57incrementality incrementality questions about it for retail
- 19:01media. How do we do this?
- 19:02I even we did an episode on it on another podcast a few months
- 19:05ago, like just talking about like, what is it right?
- 19:08So you know, definitely an important topic.
- 19:10Love to hear you have to say. Yeah, I look, there is no lack
- 19:15of complexity in our space and it at once anyone feels like
- 19:19they get their arms around it, we just throw another thing in
- 19:21there. And I mean, we're not even
- 19:24talking about like, you know, large language model fueled chat
- 19:29bots and you know, when they get ad supported how we're going to
- 19:32figure out how we attribute things to them.
- 19:34Yeah, when, not if, right when they get AD supported.
- 19:37They will all be AD supported. You can mark this.
- 19:39There is no doubt in my mind they'll be AD supported just
- 19:41like all CTVS. Well, Perplexity just lost their
- 19:44head of ads, right? I think Taz Patel left, which
- 19:46is, yeah. I think it says more about the
- 19:48chaos there than than their goals.
- 19:51So yeah. But, but you know, winding back
- 19:53to to rocker box, I think, you know, you're exactly right that
- 19:57their focus and our, you know, our focus is trying to ensure
- 20:01that we can pull enough signals and run the right types of
- 20:07modelling that helps tie, you know, the, the, the exposure to
- 20:14the outcome and do so in a way that is proprietary, unique to
- 20:19them. Clear and actually focuses on
- 20:23what the advertiser is interested in, right and I think
- 20:27you know that is unique to rocker box where I think the
- 20:34connective tissue starts to make sense.
- 20:36And this is something which is pretty is the data point which I
- 20:40actually just heard recently and it makes total sense is that we
- 20:43found that almost 30% of of ads that were being considered as
- 20:51being driving attribution were actually not viewable, right.
- 20:57So so exposure does not exposure and credit, IE cookie bombing
- 21:02from. That Oh yeah.
- 21:04And they do the same thing in mobile, like mobile mobile
- 21:06device bombing, basically, yeah. You know, driving attribution.
- 21:09So you know, to me that shows like, hey, there's a desperate
- 21:13need to start with quality data, IE stuff that DV can assess is
- 21:18viewable or not fraud and pull that into the model as well so
- 21:22that we can make what Rockerbox is doing even better than
- 21:26anybody else out there because we're doing that kind of that,
- 21:29that filtering first. So Rockerbox sells directly to
- 21:32customers. They they've not, they don't
- 21:34sell the agencies, they sell directly to mostly performance
- 21:38driven advertisers. So that's interesting.
- 21:41So and performance driven advertisers like it doesn't
- 21:43matter the verticals necessarily.
- 21:45Oh there's some verticals are more performance driven than
- 21:46others. Is it?
- 21:47More, more lower funnel. Lower funnel, so roots mailers
- 21:52like Lowe's hotels. DTC.
- 21:55Alfreders, Warby Parker, you know, Weight Watchers, people
- 21:59that are really driving online transactions.
- 22:02That's kind of their specialty is that they, because they've
- 22:05been able to really come up with a, a, the model that shows true
- 22:09ROI on spend. And then obviously apply that to
- 22:12the, the immediate mix model like where we can attribute
- 22:16revenue or attribute sales to this.
- 22:19Now this is where you should be shifting your dollar.
- 22:21Yeah, exactly. Strategic budget allocation
- 22:23changes and shifts the have you. And I'd be interested here
- 22:26because one of the challenges that we certainly faced at New
- 22:28Star was there was always a data hygiene and data unification
- 22:33issue, especially with the big brands.
- 22:35Like we always talked about how GM who is a major client of ours
- 22:39had 60 different data warehouses.
- 22:41So like the first six months of our contract with an MTAMMM sort
- 22:45of hybrid was combining that data, getting that cleansed and
- 22:49enriched, etcetera, before you could actually begin starting to
- 22:52kind of power the models and getting more predictive and
- 22:55prescriptive in your do you. And so implementation times
- 22:58could go for 10 months, a year before they even see any results
- 23:02outputs. I mean, is Rockerbach struggle
- 23:04with that or is it a different approach?
- 23:07I mean, look, from from my perspective, it is a new
- 23:12universe, right, Because I'm coming from implementations that
- 23:15are much faster, much better, you know, we, you know, we
- 23:18control the the engagements in many ways and how we're like
- 23:22pulling data in from third parties.
- 23:24It's based on just ad impression data, right?
- 23:27Yeah. So, you know, so that a, yes,
- 23:30the, the, the ramp time, it is still considerably longer for a
- 23:34rocker box customer than it would be for like a straight DV
- 23:37verification customer. And the complexities of pulling
- 23:40that data in are not small. I mean, you know, but I, I do
- 23:45feel like this is a crew that's been around for a long time.
- 23:50They're digital first folks. These are not like legacy MMM
- 23:55guys from you know. From the from the old Nielsen
- 23:59days, they got rid of that. Axiom days yeah, like the you
- 24:03know so so they're pretty adept and they know and and most of
- 24:07the customers that we're they're working with are digital 1st and
- 24:10and digital transaction first companies right yeah.
- 24:13So the data that they're pulling in is not as heavyweight as as
- 24:19maybe what. Yeah, something like a Bank of
- 24:21America. Or GM is like, you know, GM is
- 24:24looking at autos, you know, like they're, they're, they don't
- 24:27have online transactions per SE. You know, Urban Outfitters and
- 24:31Warby Parker and a lot of these other folks, I mean, like a vast
- 24:34majority of their transactions are online.
- 24:35So the lift isn't as heavy. And I think it has to do with
- 24:39the kind of customers we've targeted as well.
- 24:41Yeah, Plus GM, they have the three tiers.
- 24:43It's, it's a global company, right?
- 24:45They have dozens of agencies. I mean, Brett, I've worked in
- 24:47those big MMM projects. They could take six months to a
- 24:50year just gathering the data, aggregating it, building the
- 24:52models. I agree.
- 24:54Really, really, really not easy and maybe AI could speed some of
- 24:57that up, right, in terms of the data collection, normalization,
- 25:01etcetera. On the subject today, I'd love
- 25:04to hear, I don't know how much you can share, Mark, but love to
- 25:06find out about how is like any interesting things in a road map
- 25:10related to this. There's been a big lot of buzz
- 25:13about a Gentek AI recently. Love to hear like you know,
- 25:16whatever you're able to share about that.
- 25:18Yeah. Look, AI is changing operations.
- 25:22It's changing our product Rd. mapping and it's changing the
- 25:25environment where our customers are sending ads like, you know,
- 25:29starting from the tail end there.
- 25:30We know, you know, there's so much AI slop in the open web.
- 25:35Yeah, we talk about that all the time.
- 25:37It's. It's insane.
- 25:38You know, we just published a, a, a paper on kind of AI
- 25:44generated recipe sites of which like vast majority of recipe
- 25:49sites now are all AI generated by yeah, creators and moms who,
- 25:54you know, are also, you know, we find as they were telemarketing
- 25:59pictures and from a Russian website, you know, like, you
- 26:02know, so the, the content environment has become really,
- 26:06really kind of. Awash with a lot of crap.
- 26:10A lot of crap and we have look we've had customer engagements
- 26:13who've come to us and said and and like this won't last, but
- 26:16they've said we don't want to be around any AI generated content.
- 26:20Yeah, good luck. With that.
- 26:21Yeah, good luck with. That we're like you're in
- 26:23finance and by the way, 90% of finance articles are AI
- 26:27generated. So that that being said, but
- 26:29like there is something like we are detecting, for example, deep
- 26:32fakes in political content, right?
- 26:34That is something that advertisers don't want to be
- 26:36around. So it's created a.
- 26:38Yeah, Donald Trump was supposedly dead, like a couple
- 26:41days ago, right? That was a deep fake.
- 26:43That was a big. I don't know AI's for all the
- 26:47good things AI is doing, it's doing a lot of.
- 26:50Yeah, well, and it's interesting, Mark, we did an
- 26:52analysis. It was, it was like Krepsick and
- 26:54I and he, he presented this in a, in a webinar that we just did
- 26:57on sort of the content universe and how that landscape has
- 26:59changed. When you talk about UGC and how
- 27:02in today's world 50% of all content that's produced is sort
- 27:06of short form UGC based content, which has much lower CPMS,
- 27:10different ad models, right. We, it's, it's the creator
- 27:13economy, 30% is sort of your Hollywood economy, like produced
- 27:18original content that we watched on Netflix and the like.
- 27:21And then the 20% that hasn't changed in like 20 years is
- 27:24basically live events and in news like live sports and live
- 27:27events. And so that to my, you know,
- 27:29the, that UGC bucket, which it was a surprise to me 50% and
- 27:34this was 10% eight years ago. 7, you know, nine years ago is
- 27:40going to be just a complete sea change in how we advertise
- 27:43against that content, right? Because there's such a, it's so
- 27:46voluminous, there's so much of it.
- 27:48So how do you see, how do you look at that?
- 27:49And this question, I'm just sort of top of my head.
- 27:52How do you think about, you know, managing sort of
- 27:56verification efficiency spend against this just robust content
- 28:03source that's only going to grow in in volumes?
- 28:05So we've, I mean, we've already encountered this.
- 28:07So you mentioned short form video and short form videos
- 28:10exploded. So we work with TikTok, we work
- 28:12with Reels, we work with Shorts, we verify for suitability and
- 28:15safety and, you know, content context and all that stuff.
- 28:20And you know, the shorter answer is we used AI to to to analyze
- 28:25AI. Yeah, yeah, yeah.
- 28:27So we do predictive modeling, for example, on short form
- 28:30videos. So rather than look at every
- 28:32frame of a video to determine what happens, we look at 1 frame
- 28:37per second and we determine that what's going to, you know, we
- 28:40predict what will happen in the next 30 frames that happened in
- 28:44that second of HD video, right. So we're using AI to do
- 28:48predictive modeling. We're using AI to do labeling
- 28:50now. And we've always used large
- 28:53language models to analyze context, but a lot of those have
- 28:56been initially fed by what we call labelers people, right?
- 28:59Yeah, feed the models. Now we're using AI to actually
- 29:03do the labeling as well. So which has increased our
- 29:07scalability is increased our speed to contextualize content.
- 29:11So I mean, we're just using the same tools that people are using
- 29:14to create content, to analyze content as.
- 29:17Yeah. And it's all about like applying
- 29:19metadata in a sense to that content to label it that, you
- 29:21know, taxonomize it, Orient it, right.
- 29:24It's interesting because, you know, and you've been talking
- 29:26about the big move towards CTV and social media and social
- 29:30video and digital video. We did some analysis that showed
- 29:33him of the like by 2028 we're going to have about a $500
- 29:36billion advertising ecosystem, 220 billion of that is video
- 29:41driven right and and 50 billion is CTV about you know, these are
- 29:46approximates. I mean, you know, you can look
- 29:48at E marketer. By 2028, that's what they're
- 29:50saying. Right, yeah, by 2028 and 120
- 29:52billion is going to be social video and digital video, which
- 29:56is the, you know, your, your guys, your guy in college now.
- 30:00Yeah, he just started. Just started.
- 30:02But mine did too, because we we used to talk about our boys and
- 30:05their the skiing adventures they went on.
- 30:07Yes. Where'd he go?
- 30:08Just just off the. Top Why you NYU Film School?
- 30:11Nice. Well, they need.
- 30:12Yeah, Miles, my son's in Syracuse film school.
- 30:16Oh. Nice.
- 30:16Yeah. So we'll we'll have to compare
- 30:18notes offline, but. Our unemployed children's now
- 30:22they don't have jobs because film is dead and everything day.
- 30:25I'm like moved AI moved to digital video.
- 30:28But yeah, so, so, so it sounds like you guys are you're taking
- 30:30that challenge of digital video and the and the just the sheer
- 30:34growth of that, right. My point being with our boys,
- 30:37they're on that. I bet they're on that six hours
- 30:39a day watching, watching short form video on TikTok.
- 30:43It's incredible the behavioral changes you see in the youth
- 30:47around. Us, the youth.
- 30:48The youth. The youth.
- 30:52Yeah, it is. And and again it.
- 30:54Whoa. I lost my light here.
- 30:57Hold on. I have to.
- 30:57I have to stand up, Brett. Yeah, the motion sensor has to
- 31:01pick you up again. Yeah, come.
- 31:03On Give me something here. Here we go.
- 31:05Here we go. Yeah, no, that's the Beastie
- 31:09Boys pitcher in the back. It's not Backstreet Boys, which
- 31:11Brett thought it was earlier. Just just so we're clear.
- 31:13He was still out of focus. Guys, I know the Beastie.
- 31:17Yeah, no, it's a seed change for sure.
- 31:19You know, engagement is, is all over the place now with regard
- 31:23to like, you know, how consumers are being engaged and and, you
- 31:29know, advertisers. It's a, it's a, it's an
- 31:33interesting universe for an advertiser to be in.
- 31:35We work with a ton of CPG advertisers, right?
- 31:38And, you know, they've always been the sights, sound and
- 31:41motion people because we need to create a connection to a brand.
- 31:44CPG is all about branding. And how do you do that?
- 31:47You know, 30 seconds was hard enough.
- 31:49How do you do that now in five in which someone's going like
- 31:53this? I think, you know, the rise of
- 31:55influencers, the rise of, you know, trying to, you know,
- 31:59breakthrough in these short form video universes is, is is really
- 32:03interesting. But also the fact that you are
- 32:05now at least when I used to run a commercial in daytime, I knew
- 32:10that if I was going to be on during soap operas, I knew what
- 32:12the content was. I was felt fairly confident that
- 32:16I knew what was going to happen. And you know, if something bad
- 32:19happened, then you know, someone, the network could hear
- 32:21about it. Today, it's like I have to be in
- 32:25short form. I have to be in these social
- 32:26media environments. And what I don't want to have
- 32:29happen is have my ad show up in between two terrible posts about
- 32:33something that's really uncomfortable for me to be
- 32:36around, right? But there's no other place for
- 32:39me to be. And that's right, I have to be
- 32:42there to sell product because that's where my where my
- 32:44audience is. Well, Mark, looking at like
- 32:46responsible media, like UGC versus news organizations, like,
- 32:52how would you describe your role helping brands navigate that?
- 32:55Because I know it's, it's challenging for them, right?
- 32:56They, they need to be in certain places.
- 32:59But you know, you look at some of the content that gets
- 33:01uploaded social with some of this pretty horrific, right?
- 33:04I mean, I go to a lot of ad tech is on X.
- 33:06I'm there all the time. And you know, you have to you
- 33:08have to curate your feed in order to just not go crazy
- 33:10sometimes, right? Like how would you describe your
- 33:12role in helping brands navigate this?
- 33:14Pick where to go, yet at the same time support credible and
- 33:19responsible journalism and avoid disinfo.
- 33:21How would you describe that? Yeah.
- 33:23I mean, look, we the, the one thing that we are not in the
- 33:26business of is kind of directing ad spend anywhere.
- 33:30Like, you know, we are there to be the advertiser's partner and
- 33:35to help them navigate the places where they want to be.
- 33:39You know, you mentioned news and I think it's a place where, you
- 33:42know, we still just philosophically that as a
- 33:45company and personally feel like it's important for news to
- 33:48thrive. It's important for there to be
- 33:51quality news across the spectrum, you know, and for that
- 33:56content to be easily accessible and, and for it to be
- 34:00monetizable by advertisers. Advertisers want to be there.
- 34:04So, you know, we launched the news accelerator last year to
- 34:07help publishers. We were with big publishers.
- 34:10I mean, we work with folks like News Corp and Fox and Reuters
- 34:17and New York Times and The Post. You know, we work with a lot of
- 34:22big publishers. And you know, the goal of that
- 34:25news accelerator was to help them better understand how they
- 34:30can patch their content, to help communicate and educate
- 34:33advertisers on the quality and importance of news.
- 34:37And to help advance our tools, the technology that measures use
- 34:42to find their way to good content and make sure not good
- 34:46being appropriate for them. And to stop using blunt tools
- 34:52like keyword blocking to ensure. Domain blocking, yeah.
- 34:56To ensure that they're being more sophisticated and, you
- 35:00know, that's there's a lot of noise around how terrible
- 35:05keyword blocking is and how it destroys news, I mean.
- 35:09Yeah, yeah, Lupus call is called in a Roach motel, right.
- 35:12It's like they go it. It's not that basically his
- 35:14whole point was, you know these keywords and these domains that
- 35:16you've blocked these lists go in and they don't come out right
- 35:19And they're not dynamic. They're not they're.
- 35:21Not curated for they're not. Yeah, they're not contextually
- 35:25predictive to say, hey, you might block the Hillary Kent
- 35:27Clinton keyword, but that's kind of irrelevant.
- 35:30Today's ecosystem. She's a grandma on Scarsdale,
- 35:33you know. How do you think about the
- 35:34nuances of? That you know, there's a couple
- 35:36plays, you know places we play there.
- 35:38The 1st is we don't recommend any advertiser use keywords as a
- 35:43solo way of blocking content. Yeah, right.
- 35:46We've got categorization tools that are, you know, we have
- 35:50hundreds of suitability categories with multiple levels
- 35:54of sensitivity that are very sophisticated that look at
- 35:57things like sentiment. That's what we promote, right?
- 35:59We should be using those tools. Keywords, you know, in
- 36:04conjunction with that. Well, keywords right now
- 36:06represent .2% of all the blocking that occurs.
- 36:10Oh really, 0.2%? Wow, .2% of all the blocks the
- 36:14blocked impressions are blocked by keywords is 2.2%.
- 36:19So that aside, even with keywords, what we've done is
- 36:24we've built new tools to allow keywords lists to be to
- 36:29advertisers to be alerted when keywords are stale and not have
- 36:33been around too long, make it more seamless and easy for them
- 36:36to up to up to update and upload keyword block lists if they want
- 36:42to do that. Can that be automated through
- 36:44some sort of AI decisioning? Engine, I mean, we have
- 36:47automated alerts, we have automated, you know, we're
- 36:50automating expiration periods, We're automating all of those
- 36:53things for advertisers. I mean, you know the lat.
- 36:57This really comes down to advertiser preference.
- 37:00We can do everything from expire the keywords to ensure that
- 37:05those keyword lists are not stale and alerting them to that
- 37:09to enthusiastically and, you know, telling them you should
- 37:13not be using keywords on their own.
- 37:15You should be using sentiment analysis.
- 37:17You should be using our other tools.
- 37:18It really comes down to the advertiser and ensuring that
- 37:22they are the ones who are turning the knobs.
- 37:24We can't turn the. Knobs.
- 37:25Yeah, you can't control what they're.
- 37:27Doing you can't get more panned out that road.
- 37:29And look, there's a reason why we launched the news accelerator
- 37:32and it was because we don't want like poor behavior to actually
- 37:37impair news growth. It's not a good thing.
- 37:41It's not. Why it impairs the open web?
- 37:43Right. And, and look, the reality of it
- 37:45is, is that without tools like ours, advertisers just won't buy
- 37:50news at all. That that that's, that is the
- 37:53truth here. And, and the reality is they
- 37:55don't need to buy it there. There's an infinite amount of
- 37:59content. Advertisers don't need to buy
- 38:01content. Should they buy it?
- 38:02Yes, because I think it's important for them to be there.
- 38:06And if they have tools that allow them to navigate around
- 38:09things that is more appropriate for them, they should leverage
- 38:11those tools. And, and that's what we want to
- 38:13promote is the idea of you can use a scalpel or you can use a
- 38:17meat cleaver. We'd rather you use a scalpel.
- 38:19And this is the way you use it. Yeah.
- 38:21Yeah, we we've talked a lot about news publishers than this
- 38:24show Brett, right, which I think it's been good like Lou is
- 38:26great. Last night, laser.
- 38:28Yeah, we stuff laser and we have, you know, great one with
- 38:30Jessica Ho coming up from hers. But I think it's interesting how
- 38:34like, and I get it like complaints about keyword
- 38:36blocking and not and how it's impacted display.
- 38:40But I I do think that like if there's a broader problem about
- 38:43that the business of news publishing and you know, whether
- 38:47you go back to Craigslist or how that eviscerated classified ads,
- 38:52right? And how, you know, look at
- 38:53Google Meta's impact on siphoning traffic away.
- 38:56It's just been a broad just I think change of the business
- 38:59that they haven't responded very quickly to with not a lot of
- 39:02easy answers. So I think blaming keyword
- 39:04blocking and is in AD verification, brand safety is
- 39:07probably just AI don't know. I think it's missing the forest
- 39:10for the trees to certain extent. Yeah, that's a good point.
- 39:13I mean, it's yeah, it's, it's a little, yeah, it is a little
- 39:16maybe a red herring, right, that like innovative dot publishers
- 39:20do have to innovate how they engage in attract, you know,
- 39:24brand dollars, whether it's their content models, their
- 39:27advertising models, etcetera. Yeah.
- 39:29No, it's, it's look, it's, it's a, it's a tough time to be a
- 39:32publisher And, and you know, we want to help advertisers better
- 39:38direct spend there. I mean, the, the, I think the
- 39:41existential crisis for, for large publishers right now
- 39:44obviously is, is AI chat bot summaries.
- 39:47I mean, you know, we already know it's impacting publishers.
- 39:51So there needs to be a universe in which they can figure out how
- 39:54to monetize that as well. Otherwise, you know, or create
- 39:59content that's unique, that isn't just, you know, fact
- 40:04based, you know, snippets. It has to have a POV, right?
- 40:08And again, I'm not a news person.
- 40:09So there's, there are much better prognosticators here.
- 40:13But you know, I think the reason why we talked about about recipe
- 40:17sites, the reason why recipe sites are AI generated.
- 40:20Because they're just a list of facts, yeah.
- 40:23Right. There is no perspective.
- 40:25There is no nuance or opinion. Or even the recipe sites are
- 40:29going to be summarized in a summary and I won't even need to
- 40:32go to the AI generated recipe site.
- 40:35So content needs a POV and I think that's what will thrive
- 40:38and survive. That's a good point.
- 40:40So you look at like Free Press, they're getting acquired by CBS,
- 40:42what, 200 million, right. It's just basically like that
- 40:44actually happened. Is that.
- 40:46Yeah, it sounds like it is as of today.
- 40:47I mean, it may not, but like I was reading that it's going to
- 40:50Barry Weiss's organization, right?
- 40:53I mean, it's just basically like a bunch of blog.
- 40:56It's a podcast combined with, you know, some, some great
- 41:00content that they've been, they've been publishing on sub
- 41:02stack, right? And that's been acquired right
- 41:04for she's, I mean, said that good, good quality content,
- 41:08people value it. People read it.
- 41:09I mean, like, like, so I think that that proves there is a
- 41:12market for it, but it's probably like the format's just changing,
- 41:15right? And publishers need to need to
- 41:17embrace that. And and you know, I don't think
- 41:19that anyone knows what the feature is going to look like to
- 41:22your point, Mark, but I think that things are changing and
- 41:24there is there is an appetite for good quality content that's
- 41:27made by humans. Yeah, for sure.
- 41:29So, so maybe a little bit of a pivot is would love your
- 41:32perspective and you may not be able to say a ton on this, but
- 41:35scope three another topic, Brian O'kelly, o'kelly, you know,
- 41:40launching this this product initially.
- 41:42Speaking of pivots. Right.
- 41:43Yeah, yeah. To decarbonize the advertising
- 41:47ecosystem, which we all thought. How long is that going to last
- 41:49before he pivots back into sort of classical ad tech?
- 41:53It seems like his latest pivots are moving into the the ad
- 41:57verification IASDV space. What do you see of?
- 42:01That agentic ad verification, right?
- 42:03Yeah. No, there may be a third pivot
- 42:06here because I don't know if you saw today there were layoffs at
- 42:09Scope. Three saw that, yeah, it was a
- 42:11sales team, right? Yeah, and they want to go into
- 42:13agentic AI and or I'm sorry, agentic media buying.
- 42:19So I think there's a, you know, I think there's a path there
- 42:22where they're going to go into actually purchasing media based
- 42:25on the signal that they have. I mean, I wish, Brian all the
- 42:29best. We'd worked with Scope 3 when
- 42:31they initially launched for carbon, for decarbonization and
- 42:35had them as a part, as a signal coming into our system and as a
- 42:38reseller. And they obviously pivoted into
- 42:40other areas which were quasi competitive.
- 42:45You know, I think it's, it's we, we will see where that that ends
- 42:50up. Yeah, I think in this current
- 42:52administration, anything carbon is going to be tough, right.
- 42:57You know, that's for sure certain carbon intelligence.
- 42:59And you know, it was interesting.
- 43:01I mean, this is I guess this is another pivot and maybe they
- 43:03become an agency buying media, I don't know.
- 43:05But but I do think like the one point they made though, it was
- 43:08interesting Mark is that like instead of like maybe there
- 43:11wasn't that much of A market for carbon intelligence or carbon
- 43:13accounting, right? But but like there is some,
- 43:16there is a mark like if you take a lot of the crap out, you
- 43:18mentioned this before, it does lead to better outcomes.
- 43:21And you know, every, I think every impression is almost a
- 43:24gram of carbon generated. So if you're actually cutting
- 43:26out a lot of AI generated crap, the garbage can focus on real
- 43:29outcomes and real performance. You should.
- 43:31It should be better for the environment anyway.
- 43:33Yeah, no, for sure. I mean like, I, I think the one
- 43:36thing everyone can agree on is that efficiency is better,
- 43:39right? More efficiency, more
- 43:41effectiveness, less bend, higher ROI.
- 43:45Fewer middle men. Fewer middle men, you know, let
- 43:48and I think you know, that obviously is our thesis too,
- 43:51which is if you can find quality cheaper, that performs better
- 43:55and we can show you that it performs like that's hitting
- 43:59the, you know, the, the, the, the three legs of the stool,
- 44:02which is quality, efficiency and performance, right?
- 44:07And I think that's what every advertiser wants and that's
- 44:09good. And if, if and the output of
- 44:11that is, you know, economically positive and is environmentally
- 44:16positive, those are those are great things that that you can,
- 44:19you know, you can point to. Yeah.
- 44:21And I love how you've just simply met it down to it's like
- 44:23we had three, it's like 3 tent poles, 3 pillars.
- 44:26And your entire business approach over the last five
- 44:28years has been moving down in those three buckets.
- 44:30Really, right. So, so looking ahead and I'm
- 44:33sure you probably have to run in a little bit looking at,
- 44:35although I thought you said you had until 5:00.
- 44:37PMI did put AI put a suit on just for you guys, right?
- 44:39That's. What we appreciate.
- 44:41You came out of a client call, I think a client meeting.
- 44:44Somebody told me that. So looking at three years,
- 44:46what's your, what's your sort of bold prediction for verification
- 44:50optimization, you know, a unified sort of quality to
- 44:53outcome OS, which seems like you're really leaning into that
- 44:57theme, which is, it seems critically important, you know,
- 45:01kind of a more Federated ecosystem.
- 45:03What's what is what's your edge at diverse at DV and and where
- 45:06do what do you see the the? Future, I think, you know, I, I
- 45:10think there won't be an engagement without all three of
- 45:14those components as part of it like verification and optimizing
- 45:19the cost of that verification are just going to be
- 45:21interlinked. You won't even have a discussion
- 45:23with the customer because, you know, the idea of finding
- 45:27quality cheaper just makes sense.
- 45:29Like they're they're going to come together as one and that
- 45:33third leg of the stool, which is actually proving the
- 45:35effectiveness of the performance is just a no brainer.
- 45:38So when I look at the business three years out, I see, you
- 45:41know, a vast majority of our engagements in, you know, being
- 45:45across all three of those those areas through an integrated
- 45:49solution which we're building out right now.
- 45:51We launched our first product Authentic Advantage, which is
- 45:54our first optimization and verification solution together.
- 45:57I see all those coming together into one platform.
- 46:01You know, the one thing that we didn't talk about and we could
- 46:03talk all day is about, you know, the increasing rise of black box
- 46:07solutions right now, transparency and without, you
- 46:11know, any assessment of quality, only assessment of performance.
- 46:15And I think what we want to be is the counter to that, which is
- 46:18we're going to we're going to open the black box through
- 46:20transparency and verification. We're going to show what where
- 46:23you actually went and we're going to help you understand
- 46:25what worked and what didn't work.
- 46:27And we're going to do that in a very independent way because I
- 46:30think advertisers love the simplicity of the black box, but
- 46:34everyone I talk to fears of what that actually means.