Loading summary
Transcript47 lines
- [00:09]
A
Welcome to the Big Story, a roundtable
- [00:11]
B
featuring members of the Ad Exchanger editorial team. Every week we bring you an in depth discussion of key developments in digital marketing and media.
- [00:24]
A
Programmatic. It's automated, right? Machines, DSPs and SSPs talking to each other and picking the right impression. So can't we all just get along? Not quite. After some high profile disagreements and LinkedIn flame wars last year, the IAB Tech Lab is doing some peacekeeping. It wants to bring people from different companies and with different points of view together to talk it out before making decisions or LinkedIn comments that tear the industry apart. On today's podcast, we are bringing on Ben Hovines, Chief Media Officer of OMD Worldwide, who is the leader of this initiative which includes tech companies, buy side and sell side tech, as well as publishers. He is going to talk to us about how this new group, the Programmatic Governance Council, is going so far and what's ahead for the space. And if we have time, I'm going to have him bring out his crystal ball and share how he thinks Programmatic will evolve in the age of AI. I'm Sarah Sluss, Editorial Director of Ad Exchanger. And besides Ben, I also have our Managing editor Allison Schiff here today who interviewed Ben for her story about this initiative. And before we dive into our first question, make sure you mark your calendars for May 18th to 20th for programmatic AI. AI is transforming marketing and media as we all know. So are you ready? Attend Programmatic AI at the Park MGM in Las Vegas and you can learn how to make smarter decisions, be more effective and improve performance by using AI in marketing. And you can use the code POD10 or POD10 for an extra 10% discount and sign up in the link in the header of our website. So let's dive in. Ben, what is this council, the pgc, if we want to turn it into a three letter acronym? And what is it trying to accomplish? Of course we want to turn into a three letter acronym.
- [02:21]
C
Yeah, couldn't resist. Well, first, thank you for having me. So the PGC is a multilateral body that is housed inside of the IAB Tech Lab, wherein business decision makers from different corners of the ecosystem buy side and sell side, principally with ad tech also participating, can come together and define what we think the rules of the road should be for operating the ecosystem with an eye towards maximizing efficiency and effectiveness of the ecosystem for advertisers. Which in, in my view and in the view of some others who helped put this body together, is something that often gets lost in the LinkedIn flame war, as you mentioned in your intro.
- [03:05]
A
Can I.
- [03:05]
B
Well, can I ask a facetious question? Isn't this what the IB Tech Lab already does? How is this different?
- [03:12]
C
It's a great question. And it's, I think, best to think of the PGC as augmenting what Tech Lab has historically done. So historically, the Tech Lab has built and maintained the technical standards that underpin the programmatic ecosystem. OpenRTB is the most prominent thing that they've put out. That was the foundational text of the Programmatic Ecosystem. In the years since, it's been revised and expanded many times. They've also embarked on related standards efforts on the technical side. And I keep using the word technical because of what it describes and what it doesn't. It explains what the protocol technically supports, what type of information can be sent upstream and downstream. What's a properly formed bid request? What's a properly formed bid response? Again, from a technical perspective, what will these systems accept from one another? That's an entirely separate matter of how should these systems be used in the eyes of key stakeholders in the ecosystem. And that's the type of forum that the PGC is. PGC is setting up a forum for key stakeholders. Buy side and sell side, as I mentioned, to say, okay, well, it's a very flexible ecosystem by design. It was designed to be open and federated. That's what the open is for. However, that doesn't mean that there shouldn't be any rules at all because that creates more problems than it solves. So what do we think at the high level? Some rules should be, again, to make sure that advertisers are able to extract maximum value from the ecosystem, which is ultimately the best thing for everybody since advertisers pay for all of this.
- [04:51]
A
You, I can see why you're the leader, Ben, because you're making this very measured and calm and very, very peacekeeping vibes that I'm getting. So maybe if we were to rewind and imagine that the Programmatic Governance Council was in place before the TID or Transaction ID flame wars that I was alluding to in my intro, like how might it have played out differently?
- [05:21]
C
It's an interesting counterfactual because of course part of the reason the PGC was created was because of how the flame war about TID burst into the public view last year. I would actually describe that as the catalyzing incident that directly led to the formation of the pgc. But it's a fun thought experiment. What I think would have happened is it would have come up as an agenda item for discussion under the broader umbrella of dealing with bid request duplication. I know you both know what that is, but for any listeners who don't bid request duplication is when one impression opportunity is represented in the bid stream many, many times. For reasons that go above and beyond commercial necessity, it can be considered a standard practice to duplicate a bid request a handful of times across key supply partners. Like let's say a seller is integrated with a magnite and an OpenX and a Google. As supply partners, they need to duplicate the bid requests in order to populate those supply paths. However, what actually happens in the marketplace is we often see bid requests being duplicated dozens or hundreds or even thousands of times. And that is destructive to the ecosystem because the cost of processing all of those duplicated requests and then the compute of trying to deduplicate them and resolving them back down to an individual impression ultimately, like I said, gets borne by the advertiser and drives down the efficiency and effectiveness of the ecosystem. So I do think the tid, which is a piece of technology that the Tech Lab actually developed and incorporated into the OpenRTV standard 10 years ago, believe it or not, it's not a new OpenRTP component, enables people on the buy side, especially DSPs, to de duplicate these various bid requests by mechanically, deterministically resolving them back down to that single impression opportunity. So I think that's the answer because that's actually and we'll talk about this more in the course of this conversation, I'm sure, but that's actually Agenda Item 1 for the PGC is figuring out how to tackle bid request duplication overall and then addressing TID as a subcomponent of that. Because even if you had perfect TID adoption, not saying we do, but even if you did, that wouldn't comprehensively solve the problem. There are other elements that need to be in place and so that's what the group has been endeavoring towards.
- [07:49]
B
One thing I think is interesting though, that Tony Katzer, the CEO of IBTech Lab, said to me when we were talking before I wrote the story, is that even just a small change is not all you want to do, but good enough for now. And then if you make a few small changes, it gets better and better. So don't let perfect be the enemy of good. If you're going to cut down the anti tax just by a bit and then by another bit, you've made a lot of progress and you've saved a lot of money.
- [08:20]
C
Yes, I couldn't agree with that more. I'm a big fan of the concept of progress at the margin, which is can you in meaningful incremental steps move towards a long term goal? Andrew Casal, who's founder of Index Exchange, has often and publicly said that he thinks that the North Star for programmatic transaction costs should basically be credit cards, like getting the cost down to 3% or 5% of media. Now, if you believe the latest ANA Programmatic Transparency Study, which is a great piece of work, then you would believe that programmatic ad tech costs are about 30% of every dollar spent. The programmatic ecosystem wild. And I say if you believe it because that study is reflective of the ANA's membership, which is disproportionately large and sophisticated advertisers that can hire an advanced agency like Omnicom with advanced programmatic capabilities in order to try to drive those costs down, both through expertise and also commercial arrange that allow us to get lower rates for our clients. Right? But a very small or mid sized advertiser that is maybe not hiring such a capable agency wouldn't have access to those advantages. So for them, their total ad tech tax could be more like 30%. Excuse me, could be more than 30%. It could be more like 40% or 50%. So if anything, if you're looking at the broad swath of advertisers, not just very large advertisers, the ANA report probably understates the problem.
- [09:51]
A
So for the tid, do you have any particular solutions in mind that you think might be fair for all the stakeholders involved? I know you have a lot of interesting insights around bid duplication, so I think I'd love to hear more about that.
- [10:09]
C
I can definitely talk about the overall contours that the conversation has had since we started discussing this. The first thing that we've been getting our arms around is just getting rigorous definitions around the various forms that bid request duplication can take. Because the first thing I said when I was giving that practical explanation of it earlier is that there are legitimate reasons to duplicate a bid request. So we have to allow for that from the outset, right? Like, okay, there are legitimate use cases for this, let's stipulate what those are. And then here's a whole raft of illegitimate use cases for it. And what do those look like? We're still going through that process now, but we're coming to a consensus pretty quickly about it. There are also some nuances that come down to the channel level. CTV has different requirements for bid request duplication. For example, compared to something like open web display. That's because CTV has fragmented ad sales rights. Right? Like if a major CTV streamer is selling ads that are delivered on a Roku device, for example, Roku has a portion of the ad sales rights. They are not all controlled by the streamer themselves. A lot of things like that that have to be accounted for. We're working through all of it. And that's one of the reasons that we have multiple sell side co chairs. We have sell side co chairs, one per each major programmatic format. So we have one person representing open web, we have one person representing CTV and we have another person representing the in app space. So that's where we are right now. Where we think this is headed is we're going to come to a rules of the road document or what one of our best contributors has said we rules of the playground document. And the reason that he likes that analogy, and I agree with this, is that if you look at the sign on a playground, it doesn't try to enumerate every single thing that's illegal that you can't do. It just tries to enumerate the obvious things like if dogs aren't allowed, if playing music isn't allowed, if you're not allowed to bring in outside food or drinks, things like that. So that's where we're going to come to. And then from there it's going to be on major buy side players, inclusive of Omnicom and our competitors and also inclusive of the major DSPs to decide something along the lines of okay, effective date X, maybe it'll be six months from the date on which we finalize that list. But effective six months from now, we are going to shut off demand to any supply sources that are not conformant with these rules of the road. And that gives everybody a reasonable adoption, period. And I threw six months out because that's probably about where we'll land. But it's going to be defined based on what we hear back from the sell side participants about a reasonable adoption timeline. Because different things are going to have different levels of complexity associated with adapting to them based on their tech stacks and their integrations and things like that.
- [12:56]
B
I chuckled to myself for a second when you said rules of the playground because I was thinking, oh, like little kids that fight with each other so. So trying to keep them apart so that they're not having like a battle in the sandbox. Not privacy sandbox.
- [13:09]
A
Of course,
- [13:12]
C
yeah, we don't talk about privacy sandboxes.
- [13:14]
B
We don't say those words on this podcast anymore.
- [13:16]
C
It's too painful, certainly for me, I assume as well. Yeah.
- [13:24]
A
To talk a little bit about the problem of vid duplication. It's something I've thought about a lot and talked to people like you about a lot as well. And I kind of wonder if we go from a world with the 30% ad tech tax to a 3% to 5% ad tech tax. I feel like there's a lot of companies in the middle that potentially get squeezed out and if they don't add a lot of value, if they were, I don't know, reselling mislabeled video or the kind of thing that was happening in the past, good red ins, I guess. But I also think that there's a lot of publishers who are worried about a world where they get like three bids or like one bid. Right. Like there's actually surprising low bid density for a lot of publishers. And here you have companies like Amazon, Google and Trade Desk where they might be representing the demand of hundreds of thousands, millions of advertisers. And then a publisher gets one bid. So that doesn't feel fair to a publisher. And I think for sure that was the reason why, you know, prebid kind of came out with its own competing standard in the interest of the publishers. So is there a solution at all? Like, how do you. Is there an in between? I guess between getting like one to two to three bids and like getting a million bids without using bid duplication?
- [14:53]
C
Yeah, there are many roads lead to good bid density and we can talk about what a couple of those are. So one road is kind of the stick and twisted world that we live in today, where a lot of publishers have decided that the only sure route to bid density is to jam the DSPs with hundreds or thousands of duplicate bid requests. And it's a way of brute forcing it, because the DSPs have a lot of countermeasures in place to detect such activity. But since they're not deterministic, meaning they can't mechanically deduplicate them via a tid. If a TID is not present, or if it's been scrambled, then some duplicates are going to get through and generate bid responses. So in a narrow microscopic scale, it's economically self interested and rational for a publisher to do this, because even if 90% of their bid request duplication gets caught and filtered out, 10% gets through, which means that they're getting a lot more bid responses than they otherwise would, which means that their bid density is going up. So they just keep doing it and that's why we are where we are. There is another route. There is support, and has been for years in the IB Tech Labs OpenRTB specifications around bid responses for something called multibid. This is a term that a lot of people don't understand properly, so I'll just define it really briefly. It means that a single bid response coming back from one DSP can itself encapsulate multiple bids from multiple advertisers. So if there is an impression opportunity that comes through from, say, the New York Times, there's nothing to stop a major DSP like the trade desk or DV360 or Amazon, at least from a technical perspective of looking at their demand set, saying, we've got 10 different advertisers with active campaigns for whom this impression opportunity is directly relevant and it meets all of their inventory screening requirements. And they could package up 10 different bids inside of a single bid response and send that back downstream. I'm not going to get into the details of how multi bid adoption varies today from one DSP to the next, just because it's pretty in the weeds and also because it's changing rapidly. But. But the point is there is already technical support for that route and there are already major DSPs today that have solid multi bid implementations. The challenge is just how do we get from our current state to that state? Because it's dangerous for any one party to move first, right? If a DSP moves first, it could actually make things worse by just further encouraging bid request duplication on the part of publishers. And publishers are worried if they move first, they'll see a collapse in bid density because there won't be any change in bid response construction on the DSP side. And going back to what I said a few minutes ago, this is why this directly led to the creation of the pgc. When you need coordinated movement in order to realize a positive change, there hadn't really been a way to do it, because the way that the tech lab constructs their technical work, and this is not a criticism of the tech lab, by the way, this is just how technical standards bodies operate. Each technical standard is developed in isolation. There's not really a consideration of, well, okay, but then this would cause this second order effect, which then we need to address with some other measure. Whereas the PGC by its construction can directly address that. And that's why Tony has been such a believer in the overall concept.
- [18:24]
A
Do you think that multi bidding is a way that this is going to be solved tid or are there other proposals in the works.
- [18:33]
C
I think TID and multi bidding are certainly part of the answer. Just because there are efficient ways to solve for some of these problems. Right. If you have tids, then you can deterministically resolve multiple bid requests down to a single impression opportunity. That's a clean efficiency gain. And it also, if it's universally adopted for one reason or another, we can talk about theories of change on this if we want. If it's universally adopted, it's going to mechanically, meaning structurally, drive up efficiency in the marketplace. Right. Just because the economic incentives for bid request duplication beyond what's commercially necessary for ad sales rights and multiple supply paths are going to disappear. And same with multi bidding. Right. It's more efficient to send multiple bids in a single container than in many different containers. That's another way that efficiency can be driven up.
- [19:24]
B
Amazon knows it when they send you packages.
- [19:27]
C
Exactly. Add tomorrow's delivery. Great feature.
- [19:31]
A
The programmatic version of that, perhaps. Okay, are there any other items on the agenda lower down, the second, third, fourth that have been brought up that might be taken up in the future that might kind of give us a window into where programmatic is going?
- [19:49]
C
Definitely. A key part of the charter actually addresses agentic applications in programmatic. That's going to be a major area of focus. No one can predict exactly how agentic is going to materialize over the next few years, but everyone agrees that it's going to lead to the opportunity, at least for major workflow and ad effectiveness improvements. And so we want to make sure that again, this is being put in the service of advertisers. Another couple areas of discussion that have come up have been ID bridging, just establishing what the rules of the road should be when you need to take an ID of one type and transform it to another. Because as you can imagine thinking about that, there's opportunities for great efficiency if you can resolve multiple different types of ID down to a consist identifier. There's also opportunities for great mischief if you're taking, let's say, a very liberal view about what IDs might match in your system to the ID that you're getting on the inbound side could get up to no good if someone was doing it in an improper way. So that's another one on the agenda. And. And another major one is incrementality. Okay, how can we better highlight the incremental effects of programmatic advertising?
- [21:00]
B
Ben, what if you solve all of this? What are people going to Talk about on LinkedIn?
- [21:08]
C
Well, I don't think I Have to tell you that one of the best things about this industry is just how dynamic it is. So even if we solved all of today's problems, tomorrow we'll present new ones.
- [21:17]
A
No doubt. Yeah. And you brought up agentic as kind of the second order of business. And that's another area where there's a competing standards body similar to the TID scenario. Is that an extra incentive to tackle that? What does the competing standards body mean? It clearly means that the IB Tech Lab couldn't solve it. I guess I would say where it's has enough dissent that there's a competing body.
- [21:48]
C
That's interesting. Well, yes, certainly ADCP and IAB Tech Labs, AANP efforts have a lot of conceptual overlap. I think there are major differences in how they're designed and how they've been implemented. And a lot of these differences are beyond my level of technical expertise. But I think we've seen throughout the history of our industry and others that whenever there's some technological revolution, standards wars often take place. I don't know if either of you remember Blu Ray versus HD dvd.
- [22:21]
A
Oh man, I love that example. That's a great one. So we're in a Blu Ray versus dvd.
- [22:27]
C
Yeah. But like the inciting technological event was that scientists invented blue lasers that could be miniaturized into consumer hardware, and then that touched off a format war because people couldn't agree on how exactly the technology should be deployed and what the disk should look like and how they should be encoded and things like that. So basically, I don't necessarily see the existence of ADCP as a sign that anything is wrong with what the tech lab's doing. It's just that there's an incredible new technology that everyone agrees is going to have a major impact on the space. And people have different interpretations about the best way to take that technology and implement it for advertising use cases and programmatic in particular.
- [23:04]
A
When you think about what's going on with agentic AI right now, what gets you really excited about the possibility or even what you can do today? And what are you more concerned about?
- [23:21]
C
Well, I think today we're already seeing within the four walls of Omnicom huge efficiency improvements enabled by creating agents that can undertake their own workflows inside of the organization or even something as basic as serve as expert consultants. We have many, many agents that we've programmed within Omnicom by a variety of different SMEs. And so if we have a generalist planner that wants to get a deeper understanding of the mechanics of the paid search Marketplace, for example, we have agents that have been programmed by the relevant SMEs who have unlimited availability and unlimited bandwidth and they have free calendars. Right. So even things like that that might sound trifling actually have a huge impact inside the agency in terms of our ability to proliferate expertise and best practices through the company. I think down the road I'm really excited to see the proliferation of natural language interfaces across the different tools that we use. I don't know about you, but I don't think dropdown menus and radio button pickers and multi select boxes are the end of the line in terms of human computer interfaces for advertising. I think we're very quickly going to be able to collapse the gap between a brief from a client and the agency brief to an actual active campaign configuration in a DSP or something like it, or other buying consoles outside of the programmatic ecosystem. I also think we're going to see huge improvements to measurement because I mentioned incrementality a few minutes ago. One of the challenges with incrementality is just that it's quite complex and there's a limited supply of people who are qualified measurement experts who can design, validate, measure, a study. And so there's been a scarcity the whole history of incrementality in the industry because even though the science is settled, the experts that you need to do it properly are in short supply. Well, since the science is settled, it's something that agents can be trained on. So I also expect an explosion of high quality measurement where instead of you can have these same SMEs just overseeing huge portfolios of measurement efforts which will enable advertisers to not just have a better read on what their media is doing for their businesses, but also do much more well informed budget optimization. Because measurement won't be things incrementality. Measurement will no longer be done in pockets, it'll just be considered business as usual.
- [25:43]
A
I feel like whenever we write about measurement and incrementality the stories are like so popular. It's kind of the thing that, yeah, it's the elusive Rosetta Stone translating all of that. And certainly platform based measurement has a lot of flaws that I think a lot of agency people want to stitch together and marketers want to be able to have a single source of truth.
- [26:12]
C
If I'm going to parrot that, yeah, shameless plug. I've actually been working with the same team that I worked with on my MRC auction transparency standards effort, which Alison and I had a chat about a couple of years ago now. And we're going to be putting out industry endorsed guidance on best practices for incrementality measurement. Just to demystify the underlying science and methodology around it can be a little bit overwhelming. And so what we're going to be doing is putting out a pretty practical guide to the different forms of legitimate incrementality measurement and what advertisers should be asking their agencies and platform partners. And this is something that we're working with, excuse me, this is something that we're working on with the ANA forays, wfa, MRC and of course Tony from the tech lab.
- [26:58]
B
So we've been talking though about all of these really interesting and sophisticated things that people will be able to do. But I'm always wondering about the junior level people and where they get their experience. Like what are the jobs going to be like for entry level people in the, in the agency world. I mean, of course they could use natural language interfaces to do things that would require a lot of training in the past. But to really understand outputs and to know what you should be asking, you need to basically be hazed. Right. Like I feel like I'm decent at my job because I did really unpleasant at the time, yet formative and important grunt work for years like that. Grunt work is just being automated away and ask my former self, I would have loved not to have done those things, but I see the value in them now.
- [27:52]
C
Yeah, absolutely. It's something that we're thinking about a lot as a company as we're looking to make our operations more efficient and take advantage of AI at every level of the org, because it's something that we need to do to stay competitive. Right. Like all of our clients are reasonably expecting that we're getting more efficient and more productive over time. So there's an implicit question in what you're asking, which is how much of the grunt work is just work and how much of it is valuable training for their next role and the role after that. And the answer is going to have to be that we're going to need to tease those pieces apart and make sure the valuable components are trained in another way. But I think it's inevitable. The history of the industry, at least one way of looking at it, is that we're going to higher and higher levels of abstraction. Right. Like that your relationship with the underlying media buy when you're sitting at a DSP today, even with, let's say there's no AI in your workflow is already highly abstracted and there's a lot going on under the hood. And so to help people understand that, because they're not actually seeing the mechanics of everything we've been talking about, right? Like the bid stream, bid requests, bid responses, those are all totally abstracted away. If you're using a dsp, they're literally invisible to you. That's why we have an extremely robust training program for all of our programmatic practitioners and why we think that enables us to have the most competitive programmatic buying practice in the industry.
- [29:15]
A
Yeah, like what's like cursive and what's not. I have a kid who I think may be learning cursive and part of me is like, do they? I don't even know that it's worth learning cursive. I still can't do the gene. How will they be able to read old, old letters, you know, but is there value beyond that? Right. I think that's really, really interesting, especially as I find that AI chat, it kind of functions often like more of an entry level employee where sometimes, like it just gives you ridiculous things back and you're like, no, don't do that, do that. Right. The same way you might give direction to someone junior. So I think it's been very beneficial to people who are more mid level and do have really good judgment about what output is good and what output is not.
- [30:09]
C
Yes, absolutely. Another analogy I've heard, which I like, is it's an incredibly precocious intern, super smart, has read everything about the subject and has no life experience.
- [30:21]
B
Bit of a butt kisser, but.
- [30:23]
A
Yeah, yeah, yeah, that too, yeah. Okay, Ben, any final parting thoughts about TID incrementality? And there was two more that we talked about.
- [30:34]
B
TID incrementality, bid duplication, the demise of bid duplication soon.
- [30:41]
A
Yes, parting thoughts.
- [30:44]
C
Well, I think you mentioned at the top we talked about gazing into the programmatic crystal ball. So I will offer you what little I've managed to glimpse when peering through the haze. A lot of people are predicting the death of DSPS and the death of programmatic as we know it without really walking through their theory of why that would come to pass. And it's kind of circling a lot of the themes we've been touching on in this conversation, Right? Like the idea that each impression opportunity generates a bid request. So basically what we're saying is programmatic enables for vast pools of inventory to be deconstructed down to the impression level, or maybe multiple impressions bucketed into one bid request, but deconstructed into very, very small slices so that very sophisticated systems on the receiving end of these bid requests can parse them out and chop them up and do the best allocation possible based on the various needs of advertisers and what the system knows, what these systems know about the underlying inventory, et cetera, et cetera. A lot of what I hear about the death of Programmatic essentially amounts to doing away with all of that in going back to a world of basically hand sold buys or IO buys that are being facilitated by AI, to use a very old expression. To me that seems like throwing the baby out with the bathwater. It seems crazy. It seems like you're giving up the best thing about the ecosystem, which is addressability in the service of very unclear gains. So I think that Programmatic, at least in the basic sense of it, like this exchange of requests and responses in the service of doing the best matching possible between supply and demand to get the best outcomes, isn't going anywhere. It just needs to be reformed in terms of how it's practiced day to day, such that advertisers can realize more of the promise of the ecosystem. Because hopefully through vehicles like the PGC, we can stamp out some of this behavior that's become destructively adversarial as opposed to just normal buy side, sell side tension. And so to that end, if anyone who is listening is an IAB TechLab member and has not already joined the PGC, I encourage you to join. It is Open to all TechLab members. Shoot a note to Tony, Shoot a note to me. We'll make sure that we get you on board. We'd love to have your voice as part of the group.
- [33:08]
A
Okay, so join the PGC and Programmatic will stick around as an automated matchmaking service, but without the mischief or the sick and twisted forms of getting enough bid density, as you said. So well, Ben, thanks for coming on.
- [33:28]
C
Thanks for having me.