B (2:44)

I think your career arc is really interesting because you have a lot of like tactical hands on experience on the buy side and you've worked at a lot of brands. You were in digital marketing at Kellogg, you were a brand manager at Goodyear, a media manager at the Hershey Company and then SC Johnson, and then you went over to Meta for A couple of years working on product marketing for brand safety before heading where you are now, which is Zephyr, a brand safety platform for walled gardens. So you went from being at brands, the decision makers for how budgets get spent, to being at one of the major walled gardens. Basically what people think of when they think of old garden, to being at a company that tries to make walled gardens safer. So with all of that purview, it's a unique perspective. What does that help? You know, that other people, I guess, don't know, who have only ever sat in one of those seats?

A (3:38)

Yeah, well, I think it creates a lot of empathy. It's really easy to oversimplify everyone's jobs and be like, I don't know what the value of agency is, I don't know what the value of brand is, but the truth of the matter is it's incredibly complex. Digital advertising, it's not just automated through these major platforms. There's a lot of decisions and a lot of expertise and strategy. And I think sitting in all the seats, everything from, like, the actual creative production of some of this content, to flighting and monitoring this content, to the empathy of working at a platform and understanding that even if a ton of stakeholders think that something's the right thing to do, it takes a long time to get that work done. So it's definitely created a lot of empathy and understanding. And I think when I work with our large brand partners, that empathy helps me to show up. And in understanding, you know, I think a lot of people don't necessarily understand why brands are really concerned around adjacency or running within content that's misaligned. And I get it. Especially when you're working for a brand that's been around for, you know, 100 years, you really have to protect that brand. And I think that helps me to understand all the different vantage points and to show up for people in moments that matter. You know, unfortunately, over this past weekend, there were shootings all around the world. And as a brand, you're obviously mourning with the people impacted, but you're also thinking of, like, I want to make sure our brand is showing up in the right moments. I don't want to monetize these moments because I want them to be reflective and to be able to, you know, be about, you know, know, the victims and the people going through these things. So I think it helps me to spring into immediate action as things are happening in the world. Just really thinking about, what do brands care about? How can we reassure them that in these moments we're being mindful of how that impacts their brand and swiftly taking actions for brands.

B (5:42)

I do like that take because there is the perspective that gets bandied about quite a lot that, well, brands don't support the news because they're so afraid. But in some cases you don't want a flashing banner that's promoting something fun and goofy to show up next to an article about something terrible and very sobering. So it's not always about not wanting to be near politics writ large and trying to be respectful of certain coverage while supporting other categories of news. If I hear you correctly, for sure.

A (6:18)

And I think a lot of times what we talk about with, with brands is there's a lot of monetization of news. It's just the news that you know, is right for brands messages to align with. So a lot of my partners will say, like when someone's reading about tragic things that are happening in the world, they're not in the right place to receive our brand message. And it might not be appropriate in that moment, right? If, if you're a big entertainment brand and you're talking about your movie at the box office, obviously, like that's great, but it might not be right to have, you know, when there's tragic information about people who have been killed, you know, while vacationing at the beach, like these are super serious things. So we help brands, you know, monetize news in a large way, but news that, you know, makes sense and makes sense to the alignment. We have a lot of brands that are just hyper focused on, hey, breaking tragic news. I want to respect that space. But there's a ton of news that is monetizable. There's a ton of positive politics that might be, you know, right for alignment for brands. So that's a huge part of what we do is we have a ton of customization in how we show up for brands. And you know, if you're operating under a framework where it's just like block all news, probably not a good idea, right? I think it's really important to have nuance to everything you're doing and that includes like just risk overall. And we've created a ton of nuance to help brands say like, hey, I want to block crime and harmful acts, but I want to show up in the entertainment depiction of that if it's a non realistic version of that, or I want to show up in the historical context of that. So we are able to create that customization so brands can maintain reach, maintain relevance, scale, but just be really pointed about the type of content that just isn't aligned to their message in that moment.

B (8:16)

I mean, just because our politics are polarized doesn't mean that our monetization strategies have to lack nuance.

A (8:24)

Exactly.

B (8:25)

Well, why move from Meta to Zephyr? What sparked that? And was there a problem or a challenge that you felt like you could solve better from the outside than from inside a platform?

A (8:39)

Jeez. It's a great question. So, first and foremost, I worked with Zephyr on the integration when I was at Meta. So Zephyr was the first partner to integrate with Meta on the social walled gardens, providing transparency for the first time. And so I had never heard of Zephyr, knew of Zephyr at all until I partnered with them. And honestly, it was such a great experience that, you know, that was a huge part of the draw. But my boss now, at the time, he was like, hey, Brittany, you could have a much bigger impact. Like, right now, you're working across Meta, but we're working across all of these platforms. And you, you know, the, the easier part about being at a company like Zephyr is I don't have to worry about what the platform as a whole is doing. It's really difficult. And again, this goes back to the empathy piece. It is really difficult to make standards and rules and community management for an entire platform where you're, you know, looking at things like speech and harm and all of these things that are very delicate. At Zephyr, it's much easier, right? We're just erring on the side of brands, and brands can have opinions. It doesn't have to dictate what every brand does. It doesn't have to dictate what people do. It dictates what's right for a brand. So the cool part is, is when I was at Meta, you know, you have to make a lot of really tough calls, and you have to walk partners through a lot of really tough decisions in order to do things like protect speech. Zephyr makes it a lot easier of, like, I don't, you know, we can do something different than what's the community standards for the platform, and that's really rewarding. But being at Meta really helps, you know, me in this role because as we partner with, you know, TikTok and Snap and YouTube, there's a lot of similarities and differences, and we can also help raise the bar across everyone. So every platform natively is doing things that are a little bit different in the brand safety space, and we can kind of use that to push all platforms to have more choice for advertisers, to allow more control through third parties. And so I'm able to go to our other partners and say, hey, like, this is what you can do to really elevate and this is what the industry is doing in this space. And it's much easier to have that platform experience to be able to, you know, give that kind of foresight.

B (11:00)

And, and I do appreciate that. Although I have a bunch of meta questions, if you don't mind. I'll put a pin in them just for the moment, because I want to talk about what Zephyr does, but this is a jargon free zone, so no jargon whatsoever as much as possible, because we do work in the ad tech industry, after all, so I can't ask for the impossible. But you guys call yourself. I'll just read what it says on the website. The leading global technology company enabling responsible marketing for social walled garden environments. You can use some of those words, but don't use them too much and explain it to me like I'm five.

A (11:36)

Yeah, well, this is a fun, because I like to sit and listen how my husband, who works in insurance explains this to other people. So, you know, I work with a lot of people and have a lot of friends that don't work in this space. So, you know, at the heart of it, advertisers, you know, spend a lot of money on these platforms. And the really cool thing is the platforms inherently have a lot of targeting where you can tell the platforms about, you know, who your customer is and they can go out and find them. Well, your customer may have a lot of things that are really aligned to your brand, but they might also have interests that are not aligned to your brand. So if, you know, if you're targeting, if you want to reach, you know, males and females 18 to 50, for example, they might have, you know, an interest in MMA fighting and that might be too violent for a brand. So what Zephyr can help do is say, hey, we want to reach this person, but we don't want to be near their MMA fighting content. We want to be by cooking content, we want to be by beauty content. We want to be around local news. We have all these areas of relevance that it's okay for our ad to serve next to, but we want to avoid kind of breaking news. We want to avoid crime and harmful acts. We want to avoid, you know, conversations around drugs, alcohol, tobacco. So we take those advertisers preferences and we have direct integrations with the platforms. And so essentially we Build out a profile, and that profile tells the platform where the ad can serve. So if we find content that isn't aligned, we block that content and the ad just shifts to the next available placement. So this really helps brands kind of avoid unwanted, what we call adjacency. So whatever's next to their content or from their ad running within a piece of content that is misaligned. So again, breaking news, you may not want the mid roll ad to be, you know, something about, you know, a car manufacturer may not want their ad to serve in that place. So we create that transparency. We have a UI that brands can see exactly where their ad served. They can see the exact content, they can see the violation rate. So anytime the ad did serve before we were able to block a piece of content, we're transparent in that. And they can just see how well, you know, their ads are performing and how safe their content is across Meta, TikTok, YouTube and Snapchat.

B (14:12)

Okay, is that too buzzy?

A (14:14)

Did that make sense?

B (14:15)

No, there was very little buzz in a good way. So I appreciate that. I do have a philosophical, semi philosophical question for you. At what point does brand safety cross the line from good media planning into overreaction? Because you're talking about necessary nuance. And I agree with that in theory and I would hope in practice. But over the years we've just seen brands pull spend from platforms or entire channels based on a single headline. We've seen buyers lean into overblocking that wipes out like perfectly suitable inventory. And it's kind of irrational. And we've also seen marketers adopt policies that really, it's more about avoiding, like screenshots, like the screenshot Industrial complex. A fear of that than implementing like a really smart strategy or being strategic. But then again, it's their money and brand safety is a valid concern. So they deserve to spend that money where they want to spend it. Just walk me through where you'd draw the line today. Like what feels like responsible caution versus maybe you look at a decision and you think to yourself. And maybe you express it too. Like this is actually hurting performance and not doing anything meaningful for brand safety. Like you're hurting yourself for sure.

A (15:33)

Well, I think a huge point of differentiation from Zephyr is that the industry by and large uses keywords and texts in a large way to block large swaths of content. And I think that's really misaligned. So we a lot of times will evaluate historical keyword block lists that teams are using and they have something like shot on there. Well, what about when you're, you know, have a FIFA partnership and shots on goal and you know, will see things like naked. Well, you're cutting out naked cake recipes, maybe nude nail trends like keyword.

B (16:09)

My favorite is chicken breast. Right. I mean, exactly.

A (16:13)

So I, I think you have to be, you know, we don't use keywords in that way. We have both, you know, visual computer vision models. We look at text obviously. But I think it's really important if you're using keywords as a huge anchor in your brand safety strategy, you're eliminating so much suitable inventory. I also think gone are the days where brands should be turning on and off media. I think that there are enough kind of controls and ways for brands to navigate these waters that turning on and off media is incredibly inefficient. I think that a lot of what we're doing is turning the conversation from just this pure play brand safety conversation into more of a quality media discussion. And I think when you put the lens of quality it brings forth obviously inherently safety is a part of quality. But when you're looking at the rise of gen AI content, Sora 2 content, a lot of this kind of low quality mass produced, it's kind of this made for advertising for social platforms. Right? And so I think that as a component of it you're thinking about not necessarily it's not just safety, it's how do I make sure that every dollar I have is as, as effective as possible. And if I'm serving against a ton of content that's made for these social platforms for purely just to be monetized, I would argue that it's going to be far less effective than other content made by creators, made by individuals on these platforms that kind of user generated content. So I think gone are the days where people just have a brand safety playbook or, or it should be a quality, you know, getting, gaining media quality for effectiveness on social platforms. I think the maturity model is very, you know, it's very different depending on the partner you work with. Some are totally bought into the quality media approach and are thinking about this. Other partners are just like, oh, I was just, I was just erring on the side of caution. I was just going to block everything. If you block everything Jenny I made that's going to be a really big problem in the next, you know, next year to be honest, as we're seeing more and more content around Jenny I being made. So I think it is very much like when brands have kind of a set it and forget it strategy that's where I get really worried because this is meant to be living and breathing. This is meant to be iterated on. This is not meant to have just really broad sweeping protections on the era of caution. We, you know, these platforms are our media engines and you have to, you know, precisely tune them around, kind of this safety and quality to get the outcomes that you're looking for.

B (19:11)

The SORA stuff is absolutely wild because it almost begs for a redefinition of the word quality because it's kind of crappy, it's kind of weird. Some of it is racist. You know, like I've seen some crazy stuff on Instagram and it's also kind of high quality in terms of production. It's really good. So your definition of quality has to get nuanced too. It's not just like, what is quality, what isn't quality, what feels like a social version of MFA and what feels like it was produced by a human. It has to get even more specific than that because the production value is going to get so high you're not going to be able to tell the difference unless you're really good at realizing that, like this is a joke. But people are also swiping so fast they're not really paying that kind of attention for sure.

A (20:10)

I mean, I've definitely had friends and family members send me things that they thought were real content that are clearly made by AI. We think about it in terms of a pyramid. So we actually have an AI workshop, which is really cool that we do with partners. But essentially at the bottom of that pyramid is kind of core risk. So there are 12 standard risk categories that most brands care about. They're things like hate speech, crime and harmful acts, death and injury, to name a few. The AI proliferation of any of those categories are still going to be. So if a, if an advertiser doesn't want to be near racist content to be, you know, which pretty much everyone, all of our partners want to avoid hate speech. AI versions of hate speech are automatically going to be actions on top of kind of those standard categories. We have categories that are verticalized risk. So we work with a lot of CPG companies that they're worried about, you know, pushing over consumption of sugar, high sugar fat content. They're worried about, you know, kind of negative food ingredient conversations. Again, we're, we're helping them avoid that content because it's misaligned if it's original content. But if it's made by AI, we're also avoiding that. But then at the top of the Pyramid is this, you know, piece where it's, it's risky because it's made by AI. And this is where I think a lot what we're finding that sits in that, that doesn't hit one of the other two rungs is, is things like defamation of celebrities or.

B (21:48)

My dog Oliver also does not believe in the defamation of celebrities.

A (21:52)

Yes, yes. Wants to make sure. Excluding. Yeah. So, so anyways that the top of the pyramid is what we help brands to build out. And I think like you mentioned with the SORA content, what within there, you know, do brands care about? Because there's a lot of interesting things like there's dancing babies and obviously unless if you're my five and six year old who were like, how did that baby do that? Most people know that, hey, this is not a real baby. But, but you know, how, how should we treat this? Should we treat this like kids content? Should we treat it like adult content? Because there's a lot of brands, like alcoholic beverage brands, for example, who avoid kids content because it's a misalignment, which totally makes sense. They don't want to even have the perception that they're advertising to children under the age of 21. So it is that nuance that we kind of help to build out. But I agree, I think one, one man slop is another treasure. Right. I think, I guess found is slop is way too AI. Slop is way too big of an umbrella. We can't just call everything slop. There has to be nuance. So that's been really, honestly the highlight of the year for me is helping brands navigate this. And the thing is, again, we can't set it and forget it. We're going to take dust this off in Q1 and, and look at, we're looking. We have a SORA detector where we can understand content made by Sora. Specifically, we track all the models, but we're looking at what are the emerging trends and categories and what within those categories, from a quality or safety perspective, might brands be concerned about. And then we're iterating the strategy as we see the content shift.

B (23:41)

All right, well, we're going to take a quick break and when we're back, I have all those meta questions that I've been saving. So stick with us.

C (24:05)

Hello, I'm Sarah Sluice, editorial director of Ad Exchanger, and I have with me here Rick Irwin, CEO of adstrate. Hi, Rick.

D (24:13)

Hi, Sarah. Great to see you.

C (24:15)

So in a market dominated by very large identity providers, how do you see adstra Disrupting this space.

D (24:24)

Well, really simply we set out and every day we do the things that the bigger players in the industry either cannot do or will not do. And when I say cannot do, it's because we have technology that we architected from day one that allows brands and their partners to do things that they can't do with the larger partners, like manage all of their identity resolution in their cloud environment rather than in the provider's environment. And then we do things that larger players perhaps can do but won't do because it might threaten their business model that they've very successfully built up over time. A good example of that is challenger brands often need the ability to scale their business using identity solutions. And volume based pricing methods like CPM really don't allow for scale because they tend to charge commensurate prices as the client grows. And so we work with clients oftentimes to provide them subscription based pricing that's flattened level and they can plan around that for a year or 18 months or two years at a time. Those are two examples of how we disrupt.

C (25:38)

So more flexibility in terms of pricing and in how you use the identity product.

D (25:44)

Exactly.

C (25:46)

Let's talk a bit about transparency and drill into that. Ad exchanger listeners tend to love transparency. And where do you stand on that issue?

D (25:55)

Well, we think the two values that matter most in identity are transparency and control. One goes with the other in both directions. Transparency allows the brand to actually see how identity resolution is being performed because it can be adjusted to be more precise or less precise depending on the appropriate customer use case. And so we give our clients the ability to see how matches are made and then control rides along with it. They can change the level of precision for the use case that they are pursuing that particular day. So if they're pursuing an awareness campaign, they might actually want a little bit less precise identity resolution because they want to cast a wide net and reach a lot of customers. And in an application where they're offering best customers a special promotion, they would be more interested in having a very, very precise match so that they don't offer a best customer promotion to someone who perhaps accidentally isn't the best customer. So transparency and control are both extremely important.

C (27:04)

Okay. The opposite of a black box is what you get with Adstra.

D (27:08)

That's right.

C (27:09)

Let's talk about identity. It's such a broad term in the industry now, it can mean a lot of different things.

A (27:15)

How do you think people should be.

C (27:17)

Thinking about identity and how is identity sometimes misunderstood?

D (27:22)

Well, the biggest thing I would if I could wave a magic wand. I would love for people to understand the real and important difference between identity graphs and identity resolution platforms. Astra is an identity resolution platform, and every identity resolution platform produces a graph, which is a snapshot. I use an example, I use a metaphor of a movie. What's your favorite movie? Can I ask you your favorite movie, Sarah?

C (27:48)

Right now I'm feeling Melanie Griffith in Working Girl.

D (27:51)

Okay, Working Girl. So imagine if you were, if you were a production assistant or a camera person on the making of that movie when that movie was being shot, and you made a scrapbook of everything that happened during the movie shoot, you'd have this big scrapbook and page by page it would sort of tell a story, but it's not the same as watching the movie. And that's the difference between an identity graph, which is like a scrapbook, it's just a point in time, and an identity resolution platform, which is actually like a motion picture in that it is running matches perpetually all the time, updating to make sure that the association of one form of identity to another is accurate. And that requires software. Hence why it's a platform. And a graph can just be a bucket of data, but an identity resolution platform is actually a working, breathing piece of machinery, if you will, that performs the work of identity 24 7.

C (28:51)

I love that distinction of identity resolution and identity graph. It feels like an ad explainer article. I'll file that away. Thank you, Rick, for educating us on identity and for supporting our podcast.

D (29:05)

Great to be with you, Sarah.

B (29:22)

All right, we're back. And before getting into the promised meta questions, I wanted to ask you one more about keyword blocking. Like, is blunt keyword blocking still a really big problem? Because when I talk to brand safety and verification vendors, they claim that everyone's pretty much moved on and that it is all AI powered and fancy these days. But that's not what you hear. At least some of the time when you actually talk to publishers, they still complain about over blocking and having legitimate revenue be wiped out by these really just kind of irrational approaches that are way too blanketed. So, like, what gives? What is the reality?

A (30:05)

So I would say the reality of what I know of working with our clients is as they've moved from other providers, keyword blocking still tends to be a big part of what I hear from a strategy perspective. What a shame that to be said, there are, there are uses for keyword blocking that I actually do believe in specific, specifically for search. So TikTok has some great, you know, sophisticated search terms that you can block your ad from serving in that search environment. I think that that makes a lot of sense for the placement. I think keywords can be really helpful when there is a current event that's happening that is unfolding in real time where you want to put like a very specific name or something for a, you know, 48, 72 hour period because there's so much content being created that it's hard to keep up with. The problem is, is a lot of times those things happen and then they never get updated or changed. They just.

B (31:10)

Ariana Grande in Manchester. That's the classic example.

A (31:15)

Exactly. So, you know, I think that that is, I think that they live on. I think they are a problem for a lot of brands. It isn't part, you know, meta, meta, TikTok, YouTube, Snap, most of them have very limited keyword blocking. YouTube has some, like I mentioned, TikTok has some for search, which I think is really the right tool for the right space. But I, for the most part, the walled gardens don't enable keyword blocking in a large way. I would say it's a much more legacy open web thing. But a lot of times when we have new partners coming over, they're like, hey, let us to get a, you know, V1 of our blocking profile. Let us send you our keyword exclusions from open web. So I do know that they exist and I do know it's things that people use. We really try to educate that it isn't best practice and, and shouldn't be used in, in that manner.

B (32:16)

Nuance, nuance, nuance. I mean that's the takeaway I think everyone should probably bring from this episode. We'll talk. I want to get into some recent ish news which is Meta's exit stage left from media Rating Council audits. So for anyone who doesn't know the news, in October, Meta decided to stop participating in MRC brand safety audits for Instagram and Facebook. And that led them both to lose newly granted brand safety accreditations for those feeds. And that move basically means that one of the biggest ad platforms is stepping back from independent standards based verification of how safe its environments are for advertisers, which is a pretty big deal. It raised a lot of eyebrows, even though some people can be skeptical of the value of MRC audits. But I want to know how marketers should interpret this like, is Meta just saying trust us on brand safety, you know, salute emoji and why should marketers accept that?

A (33:15)

Yeah, so as a good former Meta employee, these opinions are Purely my own. They're actually more so first and foremost, I don't want it to be forgotten that the platforms, four or five years ago, no one ever thought there would be these types of controls for advertisers on platforms. Something like MRC wasn't even, you know, it was like, that is for open web. So I think the pendulum has swung, you know, a lot of different ways. But I actually think Meta's announcement was pretty loud actually in January about their stance on brand safety. And I think a lot of platforms had similar things that they did quietly. If you, you know, don't work in this industry or, you know, aren't in, you know, thinking about brand safety 247 like I am. In January, Meta announced that they would, you know, changing moderation of hate speech.

B (34:18)

Yes, the content moderation pullback.

A (34:20)

Right, exactly. And I honestly, it was my last day of vacation and I was like, oh, back to work. I actually think that platforms and I think that some of this is, you know, a product of the political climate we're in. But platforms all throughout 2025 made very bold statements that brands were allowed to have preferences, but they needed to work with independent third parties to do so. And I think it makes sense, if I'm being honest. Like, one of the most challenging things when I was at Meta is when you are again, making rules and controls and things for the entire universe. They have so many different partners that care about different things from a performance lens, from a safety lens, from a targeting lens. And when you try to boil those down to something really blunt across an entire industry, you can get it wrong. So I might, I might be. A lot of people might have different opinion, but I think because I worked at Meta, I look at this as Meta is firmly saying, like, we have made all of these connections and partnerships available. You can work with any partner you want, will show you all the content, they can show you all the content they are serving against. If Meta turned off the pipes to not let third parties, you know, kind of share this information with their clients, I would think that that's a really big issue. If they turned off their first party controls. Really big issue. I think what they're saying is, you know, the Media Reading Council, it's a, it's a very expensive accreditation, something that Zephyr, you know, newly minted. And I think that it's just really, you know, they're just shifting and saying like, you can do everything you want to do, you can customize with third parties, and we're going to push the Ball in another court. I honestly think it's more important than ever for brands to have that independent oversight. When Meta signaled that their policy changes were changing, we had a lot of partners say, like, hey, I know Meta has changed their definition of hate speech, but that definition has not changed for me. And so I think that, you know, I do think that marketers need to be diligent. I think they do need to have a strategy. I think the platforms have a North Star of making money, monetizing content, all of that. They've made these controls and things available, but I don't think, I think it is very clear what, what court, who owns kind of making sure those, you know, settings, controls, things happen and, and who's responsible ultimately for brand safety for advertisers. And I really think if you care about more than just kind of monetization standards or community standards from a platform, I really think you should be working with a third party.

B (37:15)

Well, how important though are MRC audits just in general? And I'm also wondering whether Meta's move could be a cat is for other platforms to do the same or maybe they have quietly and I just didn't realize it.

A (37:30)

Yeah, well, you know, I do think it's really important that there is some oversight in open web. We've seen a lot of things around harmful content serving and people serving, servicing that up. I think it's important to have some sort of oversight on how, you know, how these controls, how these things are being done. And I think the MRC is that governing body right now. I do think it's really interesting because it is a very expensive process.

B (38:01)

So I think time consuming too, in many phases.

A (38:04)

Yes, super time consuming. When I worked at Meadow, we, I was working, you know, the team was working on the very first MRC accreditation for kind of in stream and those types of placements. And it took a number of years. And then you think about all the people dollars on top of it. So. So I do think we have to figure out a way for these audits to be faster, to be more nimble, to not be as expensive, to be more scalable so that you can have more. Because they're all voluntary, right? Like if you're doing kind of shady stuff, you can just choose not to pursue an MRC accreditation. There's other accreditors too, like Tag and not always are those, you know, interchangeable. Like sometimes you can leverage some of the work done from one for the other. But you know, different countries and different companies prefer certain certifications. So I do Think that there's an opportunity to figure out a better way to do that. I would guess that Meta making this swift change, it wouldn't surprise me if other platforms follow suit or like you said, maybe they already have. This is something that you kind of have to watch the website and watch some of, you know, some of this happens. But. But Meta lost accreditation. I do want to say, like, Meta didn't lose accreditation because they did something nefarious. Meta lost accreditation because they are not going to continue to, you know, pursue re accreditation. Right.

B (39:38)

They just, they decided to pull back.

A (39:40)

Right.

B (39:42)

Well, one more MRC question. In light, in light of the news of Meta pulling back and not pursuing it anymore, are there any specific questions that like a CMO or a head of media should be asking their Meta reps this quarter and into next quarter in light of that decision? Like, what would you ask if you were still working at Goodyear or Hershey or SC Johnson? What would be on your mind? What would you want to know?

A (40:09)

Well, so, I mean, I think the teams, I worked so closely with, the sales teams at Meta, and there's some incredible sales leaders there that I think do really great work. I think at the end of the day, I wouldn't be asking for something, I'd be pushing for something. So there. Meta is one of the only platforms where there are still some placements that aren't truly transparent. And Stories would be one of those placements. So Facebook and Instagram Stories. And I think it's really important, you know, we see on TikTok, on YouTube, you know, like they've opened up all placements for transparency. And I think in a world where there are, you know, they're pushing brands to, you know, kind of make strategies and customizations with third parties that it really has to include all of the inventory. So that would be, if I were a cmo, if I were, you know, a marketing director, media director, I would be pushing Meta to, you know, continue transparency controls actions across every surface of the platform, because brands deserve to have that control in not just defined spaces, but everywhere. I think that Meta's done actually a great job at creating a lot of controls that didn't exist. You know, when I worked at Meta four years ago, there weren't controls on any surface besides in Stream. So they've come a long way, but the work isn't over. And I think as we're looking at more automated formats like ASC or Smart plus or whatever the equivalent is across where AI is making more decisions in the media, buying and Execution, having transparency to all of that decisioning is really important. Even if those placements make up a really small portion of the overall buy. So that's what I would be pushing for.

B (41:59)

How good is AI at handling some of the nuances of content moderation? It can handle a lot of stuff, I think. But what about the multitude of edge cases and then that creates a dotted line. There's a dotted line between that and trying to protect for brand safety.

A (42:17)

You mean like at scoring content?

B (42:19)

Like the accuracy at scoring content? Yeah.

A (42:24)

So it's, it's incredibly nuanced and we, we do a process called discriminative AI and it obviously all has human oversight, but you can look at content and it's incredibly good at, at identifying the context. So if you, an example that comes to mind, if you look at a piece of content where somebody says something like I pulled weeds from my garden, it would come up as no risk. I harvested weeds from my garden. Our systems would be like, yep, I harvested weed from my garden. They instantly know that this all of a sudden is drugs, alcohol, tobacco. But you know, it gets a lot more nuanced than that. People are using different language, saying unalived or using the water.

B (43:12)

Like segs. I think I've seen like S E ggs. Yeah, yeah.

A (43:16)

For sex. So we actually train our models off of this content and figure out the nuance. So when there are things that are happening in the world, we actually pull that content, we look at how the models are scoring it and we pivot if needed. So when kind of like the Charlie Kirk assassination happened in real time, you know, like there is additional context that's needed in those moments to be able to like we have a new understanding of something that happened. So there are going to be times that are incredibly vague where something is being hap, you know, being talked about of like, you know, if, if somebody's saying what happened to Charlie in Utah all of a sudden like AI isn't going to automatically know overnight that, that you know, is regarding an assassination. We have to teach the models that we actually have an in house content moderation team that with all current events with things happening, they're working to retrain these models all the time. We train models every single day. So it's again that kind of constant practice. But they aren't perfect. You're right. And I think human oversight is incredibly important. We also have a high priority queue where we look at content that are our AI is unsure of. And we also check things that our content has labeled to make adjustments as needed. I don't think it's ever going to be a perfect art. But what we're, you know, what keeps us up at night is we want to make sure that we're not over labeling content to decrease performance or under labeling content to increase risk. And we have a lot of very performance minded advertiser, lower funnel some of the largest brands in the world. And so it's something that we have to be really good at because it impacts media outcomes. We're making a promise from a safety, suitability, quality perspective. And so we have to make sure our models are really good. So that's kind of the process we go through.

B (45:21)

Well, I have a question about this is my penultimate question because this episode has flown by, but it's about access. So how a walled garden, say Meta, works with the MRC or works with a third party partner. So marketers essentially rely on the self reporting that platforms provide, right? Because even their closest partners don't have direct unfettered access to the full backend data from walled garden. So you rely on APIs or feeds or some form of integration that the platforms provide and they essentially control that and it's limited to and subject to the platform's terms, whatever they are. So with that in mind, like where does that leave brands in terms of getting like real true transparency? There is a certain measure of just you got to trust what you're being told, right?

A (46:16)

Yeah. So I mean, I do think at the end of the day, I think that we have seen time and time again, the platforms share with us all the time places where they've gotten things really wrong, where they've monetized really awful content. And this isn't just one platform, this happens everywhere because this stuff is hard. So if we never saw what we call floor violating content, I would be really leery of like the connections and the APIs. But the truth is we find this content each and every day and we action it each and every day. And you're right at a certain point, like I worked at Meta, I really feel like the team working on this really cares about it. And we do have to take a leap of faith and believe that those APIs and connections don't have any cleansing of any kind in it. I think it would be really, really, really problematic to advertising as a whole if there was some sort of layer of where data was dropped. We also do compare what they send us in terms of that API to. We have access to the ads manager to see what, what content actually delivered. And we can reconcile those two. So if there was missing data, we would have a way to account for it. But you know, I think, I think this, this sort of question, if you like step it back to anything, it doesn't have to be out about safety. Even if you think about just like media delivery or anything at a certain point, you do have to trust that if I put in $1,000 and Meta says or TikTok says or Snap says, I ran $1,000 worth of ads like we, you can't, you can't see each and every one of them. So I think this layer of trust exists in everything that we do and I think it just feels a little bit higher stakes in the brand safety and suitability space. But I feel really confident that we get everything that we see these violations. We also have, you know, the platforms do want to make this stuff better. Harmful content they take very seriously and they action pretty quickly. So we have a feedback loop with the platforms where when we find this high risk and floor content, we send it back to them and say, hey, this shouldn't be monetized as a whole. Which actually helps the entire advertising ecosystem. It helps to stop that content from being monetized even from partners who don't use us or use a competitor. And I think it's a really important layer is that we're actually getting better. The platforms are also interested in us sharing more and more labeling because there are teams. And I, you know, working at Meta, I know this like there are engineering developing teams that really, this is all they care about and they do want their models to be really great. And I think when you're working at a major platform on these major engines, sure there are people that their North Star is only profit, but having work there, there are also people that their only North Star is safety and their only North Star is advertiser protections. And I think that some of these recent decisions make their jobs infinitely harder. But I do believe that there are people at each and every platform that are advocating for this. So yeah, so I think we'll. If there's some bad actors in some way messing with that, it would be hard for us to tell. And I think it would be really difficult even for a governing body to, to determine because I think it would have to be pretty deeply seated. So yeah, I'm sure There's a conspiracy TikTok thread all about this that we could probably hop into and, and no doubt.

B (50:03)

I mean there's one of those for pretty much everything. Yes. Well, I want to zoom out and Also bring us home. So last one, if a listener has made it to this point, I thank you to everyone who still has their AirPods in and I want them to have one action that they can take after this episode other than listening to it to get smarter about brand safety and suitability, the nuance there and verification. Like what should that one thing be?

A (50:35)

Well, if you're a marketer first and foremost you just look are ask the teams responsible, are we using keyword blocking? Because I think that's something really easy to just if you are in a large way and that is like definitely hurting your performance, your reach, all of those things. But I think let's zoom out. Let's pretend you aren't working on a direct brand or media team or you're just more in this space. Go into any of the platforms and search Sora and start to look at the content and find a piece of content that if there wasn't the watermark, if there wasn't, you know, the hashtag that you could actually be fooled by and think about, you know, does that matter? And if it matters, why does it matter? So is it like the celebrity defamation that I mentioned? Is it the misinformation of a public feature figure that doesn't necessarily cause real world harm? Start to just educate yourself on the content that's being made and then also find a piece of content that you're like, hey, actually this is fine for my brand and what are kind of the key components of that? Because we are going to have to find places to monetize in this brave new world of AI developed content and we have to figure out what the tenants of that content is and have a strategy around it. And if you want help, you know, here as ever, we're ready to help you kind of create that strategy and playbook.

B (52:03)

I think I'm pretty media literate and I'm having trouble telling what's real and what's not real already and it's not as good as it's going to get. So stay safe and stay suitable and stay nuanced and smart out there.

A (52:18)

Yeah. Thanks so much Allison.

C (52:35)

This podcast was brought to you by Adstra. Adstra's leading identity resolution network enables precise targeting, scalable activation and richer customer understanding. Whether you're looking for person based accuracy or massive scale or only Adstra has the transparency to give you both. Learn more@adstradata.com that's a d s t r a data dot com.