Ben Pheloung (Mantis) - Media Tech Talk #005

(upbeat music) - G'day, welcome to Media Tech Talk, I'm David Richards. - And I'm Paul Lockett. In a world where algorithm still struggle to understand the difference between news and negativity, one of the companies change in the conversation is Mantis. The AI brand safety and contextual advertising platform partnered with nine entertainment, Lad Bible Group and Reach PLC, has been featured in MI3 Australia and Video Week as a pioneer in teaching machines to read the emotion and meaning behind content. - And at the helm is Ben Philong, a leader who's redefining what safety really means for broadcasters and publishers alike. So here's the question, what is it that Mantis offers that no other vendors don't and why that's resonated so well with the broadcaster like nine Australia? Ben, welcome to Media Tech Talk. How are you? - I'm good thanks, how are you guys? - Very well, how are you calling in from London? - How is it today? - Yeah. - Well, I'm watching Roselle has chased the pigeons outside the window and it's a traditional London gray. - Perfect, so all in all, not too terrible. - Excellent. Now, I want to delve into your media career. Can you tell us a bit about how you came into Mantis and what inspired you to get into media and technology? - Well, I studied chemistry and pharmacology at university and that really didn't work for me. I had a friend that managed to get me an interview at any media, which was the kind of very young network for the Daily Mail, the Metro, the whole host of other UK publishers. And that transformed from like a traditional old style network into one of the very earliest kind of programmatic buyers. We had one of the first seats on that Nexus. So I've been in the media and programmatic space, working with publishers and then in programmatic for quite a while now. Then working across ad networks and then data and media roles and then more recently at Mantis, working in brand safety space. Again, working with publishers and trying to help them make some more money. - How awesome is AI at the moment? With that data background, you must be loving it right now. - Yeah, yeah, I think there's a lot of incredible stuff we can do and also try to work and walk the tightrope of being able to do something cost effectively and sensibly rather than just trying Williams more. (laughs) - Three more dollars at the wall, doing something you can already do. - Yeah, awesome. Now, let's have a quick look into what you do. For those who are familiar with Mantis, do you wanna explain what Mantis is and why it matters to broadcasters and advertisers today? - Absolutely, so Mantis is a brand safety and contextual solution. We use AI to read a page as it is being published and provide brand safety and contextual signals. So advertisers can feel safe working with, when we largely work with really premium high quality publishers that are publishing a lot of content around a lot of different topics. So advertisers can feel safe working across those publishers and working with them really closely. Our goal for publishers is to make as much of their inventory as transparently safe as possible. - I'm gonna pull you up on that word, transparency later. (laughs) - Yeah, please turn it over. - So, when we're talking about contextual, and let's just take it back a bit 'cause there's a lot of content and we're talking about digital display, content and video. How, what are the differences and what is it to outside of brand safety? Just maybe dig into a little bit more about that and what advertisers are looking for and the reasons why they need it. - Yeah, I think contextual came to the forefront of the last three or four years. There was the upcoming cookie apocalypse, Google, Chrome, removing access to all of those, but it's still a space that is shrinking like your ability to target audiences across different platforms is harder and harder. The contextual compliments and supplements a lot of that. So having precise scalable contextual targeting kind of powered by our AI, which understands the page, understands the context of the page, allows for advertisers to actually reach what they're looking for and deliver those campaigns that have actually provide performance at scale without relying solely on cookies or on cookies at all. So that's at all part of that kind of media planning budget and diet that people need to have these days. - And what would be a case where a real life case where mantis will come in and save the day for an advertiser when they want to run an ad on a publisher page? - So some of the examples that we've kind of commonly bang our heads against, so we work with the independent in the UK as one of our kind of longer standing partners. They work with some advertisers that only really want to work across culture, lifestyle and environment sections. So they want to avoid hard news. And we kind of help them out in a few different ways with that. One, which is building our brand safety profiles that work just across those sections and they treat those sections differently than they do the hard news sections. We can have something that's talking about a true crime podcast in the entertainment section, which is still death and murder, but it's in a safer environment and it's in a safer tone than talking about more recent death and murder that you find on the kind of hard news section. So threading the eye of the needle there and helping publishers provide that safe space and explainable safe space for advertisers to operate in. So they can get that budget and deliver those campaigns. So what we find is publishers have this wealth of first-party data, this wealth of audience understanding. And sometimes this inability to actually serve ads to all of those high value users because this brand safety being applied on the buy side, and too much brand safety being applied on the buy side, and they can't get the ads in front of the right people at the right time. And that's where we come in. We're freeing up as much inventory safely as possible. We show our workings, the transparency piece. So the brands feel safe and the advertisers feel safe, actually appearing in these sections. It's pretty big call to make that inventory available at scale. When you find advertisers are critically looking at content, so their brands are obviously advertising up against or their campaigns are up against the right content. What are some of the challenges? And you've just sort of talked about true crimes. Some of the challenges that you have in this space and how do you manage that and some of the things like how do you manage in real time? Because I'd imagine the amount of data and information that you've got must be huge. And you've mentioned AI as well. Yeah, so we're always as close to the source as possible. When we're working with our publishers, and these are the top dogs, these are the guys that are making that, the writing of our stories and writing about absolutely everything under the sun. So we have to be pretty flexible in how we approach it. But we integrate directly into a publisher's CMS. We're trying to be as close to that source of truth as possible. So if something is published or something is edited or republished or it's a live blog or an evolving situation and a breaking news story, we're always there updating what we're sharing with the kind of media ecosystem as quickly as possible based on that latest article, that latest bit of information. Where we see kind of keywords being misapplied, we see a huge amount of content being blocked. We see things like the euros, England was in the semifinals. There was a penalty shootout, kind of 45% of the content we had across one of our publishers was blocked because it had shootout and shot appearing in that. We see recipe sites having content blocked because they talk about chicken breasts. We have, we have the Halloween content and everyone will be writing about today, like blocking things for horror and skeleton and things like that that are all fine. We kind of layer on sentiment and emotion scores to make sure that we're kind of weeding out bad news from good news and providing different ways for advertisers to feel safe in these spaces. - I think chicken breast is like the brand safety keyword list version of the pair of shoes that fall at Falley's year round all the time with retailing. - Yeah, absolutely. - The classic one that always comes up right? - Yeah, that's a classic for a reason. (laughs) - Yeah, luckily I haven't had that, to be honest. (laughs) Somebody's doing her job properly. (laughs) What do you think, like in your world with AdTech, have you, you've had experience obviously in contextual targeting brand safety? How have you found this space for you when working on the publisher side? - Yeah, it's been interesting like how it's kind of evolved over the years, you know, I think I remember meeting with - Oh shit, who was it? I'm trying to think of the company name now. This is like I'm back eight years ago and it was the first kind of company that I'd take business that I'd spoken to at the time who was doing contextual brand safety, looking at not just words on a page but the association between the words and whole sentences and everything like that. And that was kind of the first time this is back in 2017, and I think when I was meeting with Alan Jenner actually worked at that company at the time. Anyway, so it's obviously evolved quite a lot. And one thing that I've been kind of noticing and can you get your thoughts on this, Ben, is that you've seen in the past like contextual targeting solutions brand safety kind of being bundled up typically by those like measurement and verification businesses, right? Like your IIS and your DVs, but these days it seems like it is kind of branching off into its own little sector and you've got businesses like mantis that are offering these solutions independently. I'm also seeing, you know, DMPs that are starting to offer contextual targeting and segmentation and things like that as well and brand safety. So it's interesting seeing how the tech has kind of gone off in these different directions what you're going to take on it. - I think that's my read on the situation as well. I would say that the, one of the challenges we've had is we've kind of been built as mantis from the publishers side kind of tailor in what we're doing for individual publishers. And back in 2017, I think that's kind of my first experience with brand safety in that space as well was a lot of these technologies built to protect brands and sitting on the bar side and being applied to the open web as a whole, right? Everything from the kind of premium content you get on top tier publishers to the very, very long, long, long tail of the kind of crud that you have to navigate as well. And as advertisers and brands are moving back away from that kind of long tail space and back into having close to publisher relationships, premium publishers and looking at curated marketplaces and ways to kind of interact and buy in that space. But kind of balance shifting away from broad, broad be applied by side technology to something that's a bit more specific to understanding what an individual publisher is doing. I think way we're approaching, and yeah, a lot of the other guys as well is by leveraging AI technology to take what is, at the kind of breadth of human experience that you get written about across a whole host of different kinds of publishers and then translate that into something that the machines can pick up and actually buy against a scale and make decisions on. And I think that's maybe a challenge you've seen with DNPs and first party audience segmentation is that everyone's got a different taxonomy and that same applies with contextual targeting as well. So what can we do that we have lots of different kinds of publishers talking about lots of different things but turn those into signals that DSPs and buyers can actually interact with in a sensible way at scale to run campaigns across good quality and the trim publishers. - Hmm, and we could have a follow up on the AI point. David used this words, I think it was with Brian, when we met with Brian around AI, democratizing certain capabilities that in the past would be kind of unique to certain people or skill sets. I think that in the same way, I think it was also Brian, we talked to about this around how AI is also kind of democratizing certain capabilities for organizations like publishers, for example. And there's now, it's kind of providing that ability for publishers to do their own contextual segmentation and analysis of content. Whereas that in the past would have been kind of relegated to independent ad tech and ad tech businesses. So obviously AI being a powerful tool in allowing you guys to kind of build your products and make the products better. Do you also see it as a potential kind of threat to the organization in that publisher starting to use that tech? Will there be a need for independent contextual businesses in the future? - Yeah, good question. I think our parent company is rich PRC, which is the big largest newspaper company in the UK with national titles and local newspaper titles as well. So we're pretty close to how rich is approaching the application of AI across the business. And a whole host of that is more on the editorial side. Like, how can you support journalists? How can you understand the content they're writing and what is resonating with your audience? Because there's no secret that publishers are seeing a big change to the way that users are interacting with their content at the moment. So there's pressure to do more to less and to give more of these tools, this democratization to the journalists and the editorial teams to kind of build out and strengthen what they're able to do. And I see more movement in that side of the kind of media ecosystem than on their contextualization side. And I think we sit and the whole host of our kind of competitive sit as well as understanding how you can take something that's being written about, our biopublisher, and put that together in such a way that someone actually wants to buy against it through the program plan scheme and through the existing kind of media plumbing that we've got in place for everything. And I think that's where we kind of come into play. And that's where we're all going to be playing it as well. - I like the integration with CMS that you're talking about before, obviously, for people who aren't familiar with the way content's published in digital media. I don't want to take it back to 101, but-- - There are dozens of us. (laughing) - The publishers are literally editors and the team publishing content, literally using a templated based environment and the publishers are quite simply WordPress, right? Or Microsoft Word document where they create that content. It gets published, but your integration has the ability to look at that content and to contextually understand what is in that content and drive a scored based outcome that the advertiser can then identify and then align its campaign for brand safety to say how safe this page is for their brand to be advertised or to align with. - Well, that's obviously in a digital static print 2D space. Where are you guys in the video space? 'Cause it's a big piece and I think this leads into one of the big opportunities that you've capitalized on here in Australia last year with your partnership with the Nine Network here. Do you want to take us through a little bit of that and have video played a big part for them and yourself? - Yeah, so at the moment we are, we're looking at a whole host of information for a video but it's still largely text based. And videos are really hard space to work in. Well, certainly for brand safety and we're making some big strides again, powered by AI and AI's and proved abilities who understand what it's looking at. In the past, we've seen people make attempts of video brand safety that have been relying on metadata for a video and the transcription of a video which can be effective if you've got a high quality piece of video content such as this. But if you're looking at a bit of mobile phone footage and it doesn't have any audio track and it's off something horrible happening, a car accident or a terrorist attack, you don't have much for the technology to sink its teeth and to make a good judgment about. That's changing and that's where we are going is having AI make clearer data judgments about what's going on in a video based on it, watching the video itself. At the moment, we rely on quality publishers like Nine Having An Associated Article with a piece of video content. And we're making judgments on that which we get a much clearer read and understanding of that video by looking at the associated article piece and that allows us to provide brand safety across the video content and that approach gets adjusted depending on the publisher we're working with and where their library of video content comes from. Yeah, so we're always making a read based on the largest kind of pool of information that we've got but it is at the moment text based. - Ben, I want to ask, so you're talking about companion articles for video, that makes sense and obviously from a scoring perspective that makes it easier. But how are you going into content that's made for broadcast video on demand in an archive library where content is literally pulled out and played on demand? Do you go into the scoring piece there or are we just talking more digital, quick consuming content? I think when we're possible, a lot of that kind of beyond is associated with a description or an article or something at some point and it's a life cycle and that's where we work. So we've got the original kind of publishing piece of content that people are based and if we get the kind of associated text with that, we can provide those brand safety signals for the video there and they can trail along after that piece of video content unlike an article which might be edited and re-published and updated that video content is pretty static. So once we've got a score on that, it can look like it's like cycle. - Yep, okay. It's a big deal to have someone like Nine for Mantisk and obviously with the changes that we saw over the last 18 months with Oracle moving out of the market with Moat. What was the position that you pitched in? Are you able to go into that detail? I'm trying to look at see where the Mantis piece that made you guys better than any of the other contextual targeting and brand safety vendors in market? - Yeah, I think it was an interesting opportunity and quite a sad one. So I know lots of people will work to Oracle over the years and it's a big change. But we worked really closely with Nine to find out what they were worried about, what they were working with. They're a really technically competent group. So they really liked the, a few different things of the pitch, they really liked the AI understanding of the page and being backed by IBM Watson to kind of make a read of what's going on on the page and use that to power the brand safety and contextual piece. They liked the transparency. So you're kind of asking me about this earlier. We show all of our working. So we've got a scanner, you can put in any article, you can have a look at why it's being blocked and not blocked. We break down the page itself into concepts, categories and entities, which are the kind of AI outputs. They don't need to explicitly appear on the page. We can have an article talking about a bomb going off and police being called an emergency service being on the scene and we might identify the concept of terrorism and then say that's unsafe because of it. But we don't need to see the word terror or terror attack appearing explicitly on the page. So we can pick up stuff like this, which allows for them to manage and maintain their own brand safety profiles for broad use, but also for specific clients themselves. So they can get into the system. We've trained them up on it. They've got the skills to kind of do this on the fly. And then we provide a backstop and support to make sure that if there are any edge cases in weird situations, we can help out with that. But they've got the keys and they've got the training and they've got the understanding of how the technology works, which means that they can tailor things really quickly to the situation on the ground. They're one of the first clients that we've worked with in this way. Like we're in the past because it's easy to get things wrong when you're looking at a complex system and adjusting AI prompts and the like to get things wrong. But they were really keen to work this way and they know what they're doing now. I've been working with it for more than a year and so I think they're pretty happy with it. I think we've got this really coming out. So I'm lifting a lot of marketing stuff from that. So kind of real off for you guys. Yeah, okay. Just on broadcast live, how do you go in that space with TV? Yeah, it's hard. It's hard. I think we are still trying to unpick that one. The AI that we're looking at applying to video content is fast, but it's not extremely like real-time fast. And if it was to be real-time fast, it would be probably too expensive to apply as quickly as it needs to. So we have to look at what kind of lag we're happy with and we can get away with for that space. Or looking at overall trends for a given show over a longer period of time and basing it probabilistically on previously aired content. So there's a couple of different approaches there. That's probably the path we're going to go down. But it's hard to get video to work in real-time given the technological constraints. I like that technological restraint. I think that's an indication of you riding at the front of the change at the moment. And live TV, as you said, is highly unpredictable. And the amount of information that you'd have to review to get an honest review that's safe for a brand safety piece is important. And then you want to throw a programmatic in there where you're trying to do real-time decisioning at the same time. My brain just exploded. I just want to take a step sideways for a quick set because the amount of data, and you're being a data man, we'll have some great value in what it is that Mantis has access to. When you're scoring the page or the content, that information is valuable. Obviously, the advertiser has access to a point. Can the publisher get access to that data? Meaning, the point I want to make is, or question on ask, is can that data get pushed into their own data management platform or other assets that they've got where they can process data and profiling? Yeah, absolutely. So we've got basically two feeds of all the data that we gather from a publishes page. The first one is the actionable stuff. It's the segments that sit on the page for them to use. And then we've got the wealth of data that sits behind every page. So when they publish, publish that content. We get all of the metadata back from the system and we use that to build out the segments. We treat all of the metadata that we get from their pages as their data. And we send it with that needs to live. Sometimes that's informing DNP segmentation and first-party data segments. So enriching that with something we do with the independent is building out specific foundational contextual segments for them to enrich their first-party data sets. And we have other use cases where we help publishes sales teams. So if they're doing a competitor analysis or someone like Reach, they are publishing news about British Airways and Jet2 and all these different airline brands. Same thing with supermarket, same thing with pharmacies, same thing with banks. How often are those different brands and companies spoken about with which tone of voice? Is it positive sentiment? Is it negative sentiment? And how does that compare to their competitor set? So we kind of take all of this wealth of raw data that we have. And we send it where it needs to live to be as most use for a publisher as possible. Sometimes that's audience. Sometimes that's kind of back-end analytics. In other times, it's how much of our content is viewed as safe for the open exchange. If you've got someone a random buyer sitting on a DSP somewhere just applying a keyword block list, how do they see our articles in our content? So we're flexible with it. All the stuff that we've got belongs to the publisher. So we send it where they get the most use from it. And that use case is kind of constantly involved as well. Do you guys work with Lucky Products, the E-selling, or working with publishers directly, or do you work with brands and agencies as well on the buy side? We started out solely working with publishers. And now we're working with agencies and brands on the buy side as well. Through a lot of curated marketplaces and that space, providing access to all these different publishers we work with through curated marketplaces and working with the big hold coves and their market place offerings to do that. We find that the people in those spaces really want to be working across good quality publishers and a task with that. How can we deliver the ad campaigns across the publishers that we've got arrangements with and partnerships with. And then all they're always struggling to do that because brand safety is blocking too much and blocking for the wrong reasons. They need to have those controls in place. So anyone running an ad campaign for a decent brand these days needs to have brand safety in place. And we're trying to give them access to these premium publishers with better quality brand safety that blocks less over. Also, they can actually deliver stuff. So that's how they make their margin as well. Yeah, just a quickie on a quickie. Oh, this is a little Australian euphemism. Just pop out every now and then. So anyway, I'm not going to go down that path. That's some rabbit hole. I'm sorry. On, I just want to go down and have a look at some of the ways that your identifying content. Because we're not just talking about video and texts and images. You can actually identify the logos as well. Am I right? In that space? Yeah, we've got a little bit of technology that can look at images and pull that information out. It's sometimes useful and sometimes not. But we will look at some of the images on a page if there's not much other information there. And we can pull out things like recognisable celebrities and logos as well. Yeah, which is great, obviously, to make sure that the brand is not advertising up against something that they don't want to in a number and say for the others. Absolutely. And that's where we gather information on the sentiment of a page as well. So if someone is mentioning Apple, for example, but it's a negative press, we can exclude their content. Because we're looking at the overall tone by taking the sentiments and emotional scores for that text or balance that out across the entire article. So you can see how you're doing. But you can also look at the sentiment based on individual brands that are being spoken at about as well. So if it's a comparison article, you can actually say, oh, yes, our competitors are coming off worse on this. So let's run something against that piece of content. And the way that this information, like once a publisher has integrated your tech, how does it typically that data get kind of flow through? Is it key values that are on the page in terms of, is it contextual segments or brand safety segments? Or are they actually seeing the raw data values of what categories of the page keywords of the page, what the sentiment is, any brands that have been identified? Is that raw data made available to the publisher or are you just sending the segments across or both? It's a bit of both. So the actionable stuff, the segments, they get built onto the page as part of that CMS integration. Or they can be loaded onto a page via a tag as well. But they would appear as brand safety profiles, red green amber for safe, risky and unsafe. And they get baked onto the page, and they get picked up by your ads server or whatever technology using to read that and serve your ads. Same with the contextual segments and the emotional sentiment scores. And then the back end full-fat data feed, which involves the number of brands being spoken about and the individual concepts and categories and entities that we pick up from a page. What we could, what we prefer to do is just plumb that into whatever data visualization tool that a publisher has for looking at the larger performance for business. So sending that feed where it is most useful, Domo or Tableau or wherever else it needs to live. You can look at them, you can pair that with your Google Analytics data or Amplitude data and see, OK, these topics are getting the strongest yield and performance and people are spending longer. Prince William versus Prince Harry or these particular sports or whatever else like that. So you've got that data, the more granular data, where it's useful and can be married up to these other other data sets that we don't have of you on in terms of your yield or performance of self-written rate. Ben, do I mention this word transparency? That's going to come up. How do you manage transparency in your platform where we've got publishers who are publishing content and as we've just discussed in great detail want to make sure that they can commercialize their content appropriately. And if we've got advertisers who are avoiding specific words or keywords in a world that we live in today, that could be quite a lot of content and make it difficult for the publisher to make some money, right? How do you provide the transparency that your scoring is accurate and is scoring the right page with the right score? And so the publisher can make a query on that to see if it is doing it appropriately. Yeah, really good question. I think this is one of the things that publishers like about us is that every publisher that we work with has their own separate environment set up for them. And that stores all of their unique brand safety profiles and then some standardized ones that we use across the industry as well. And they can put any of their pieces of content into that at any point and see how that scores for all of their different advertisers. And they can see the logic behind it as well. So they've got, oh, this is being flagged for this concept, the concept of death or the, it's mentioning Harvey Weinstein or Donald Trump. So that's unsafe for these brands that are concerned about politics or any bad association or any of these people that have been in the news. So all of that information is the Airfront and Centre for them to look at. And in the case of publishers like Nine, they've got the tools at hand to make those adjustments. Well, this isn't safe or we've missed a block against shark attacks or we've got something there that we need to work on and change really quickly. They've got the tools, they can make that adjustment and that gets built back onto the page. And because we're stuck into the CMS, that gets reapplied really quickly. On the flip side of that, when we are working with publishers and they have got some premium brands that they want to run ad campaigns for and those brands have got monstrous long keyword block lists. We often, what we call like a translation service, we let's take all those topics that you can see that those brands are worried about in those keyword block lists. And let's translate that into the mantis technology. Let's instead of having 15 different words for, like everyone that's been involved with the terror attack over the past 15 years, let's just block the concept of terrorism. If they're worried about, this is often the case with car brands and motoring brands, if they're worried about climate change, let's block that using our technology rather than having 25 different words, all trying to approach the same thing from a different angle to keep it safe, but blocking too much. Same thing with violence, same thing with knife crime or sexual assault or anything like that. And because we are working with, and understand that publishers were working across, we can kind of tailor that to the content they're publishing. So that piece as well, we will actually show all of our working, the kind of publishing sales teams to go back and say, look, Apple, you don't need to block all of these different keywords to run safely across these sections of our site. Same thing with Sky News and all these other publishers and brands that have got these big concerns. We can show all of the detail there and the missed opportunity. There's often, if you've got these big brands, there's often the cases where they're running technically complex campaigns, where they've got their own internal data set that they'd like to marry up to a publisher's data set. And you've got that traditional programmatic Venn diagram of disaster, where you take a little bit more, have a bite out every time and you're left with five impressions that you can possibly serve against. And that's what we're trying to help, how can we make as much of that available as possible? This is interesting because your, with Reach PLC, the legacy is, as you've said, in the print, or newspaper, digital media space as a publisher. And we've seen Mantis come out of that space. So it feels, and correct me if I'm wrong, that you've got the publishing and media industry at the heart of what you do. Is that, am I right in that approach and how it sort of comes out to market for you guys? - Yeah, absolutely. And I think that's the core reason that we were built was Reach is a publisher. We're seeing a huge amount of safe content being blocked with good intentions and fair technology. So that's why we built Mantis to kind of solve that particular problem for publishers. And that continues to be there, the kind of driving force behind what we're doing. How can we help publishers make more money than an increasingly challenging environment for publishers? - A bit of an interesting question, just randomly sort of, is it able to differentiate between, say, if you have two articles being written about somebody, it's the same name, but it's different people. Are they able to differentiate between, figure out, let's say I was blessed with, you know, the probably the most common name to ever exist in the human language, which was, I was born Paul Smith and my middle name is John. So Paul John Smith, which I've literally filled out forms of my life where that's the example name that's in the form. That's how common it was. And I changed my surname to like it in my 20s, which that's a story for another day. But Paul Smith, obviously, there's probably thousands of articles out there written about a Paul Smith. Would it be able to differentiate between all the different Paul Smiths out there? - Yes, short answer is yes. Longer answer is sometimes it's complicated. Like, yes, so we're looking at a couple different data points we're getting into the weeds here, but we have a concept score and we have an entity score. So an entity is a little bit simpler to explain. And that's something that is being spoken about and it doesn't look at the larger kind of universe of understanding that we base our judgments on. It's like, this could be Donald Trump. This could be Paul John Smith. It could be whoever is in a local newspaper or in an obituary, just people that are being spoken about in the text and the subject. And then we've got concepts as well. So concepts are backed up by Wikipedia. So it's using that as a source of truth and making judgments based on that. So if there are other things that would link the name of the person being spoken about to other topics that are appearing in Wikipedia that kind of reinforce that linkage, then we'll make a distinguishing judgment. So we're able to use concepts in a different way to entities. Because the entity is sometimes there's a specific person that a brand is worried about and it's got a negative association with. Tesco in the UK always had Mel B from Spice Girls on their blog list because they had some sort of disastrous influence a deal with her at some point. So there's stuff like that, but then there's the concepts which allow us to be a bit more high level to provide deeper linkage based on what is available on the Wikipedia page. So yes, and like I've said, the complex as well. - Yeah, yeah, no, fair enough. That'd be a good test to add to the list. When you go into the checklist of like our pain for a contextual solution, can it tell which pulse myth is the right one? - Jesus, you know, that just happens up a whole can of worms, Paul. - So I want to say get into a little bit more on like the programmatic side as well. Like there's a lot of talk around like signal loss obviously with cookie deprecation and everything happening in that space. And so, I mean, one thing that we used to talk about at newscorp was, you know, the signal loss, but improvements in signal fidelity because the number of hops between the by side and the cell side has actually gotten quite smaller. And the tools and products in tech that we have on the publisher side these days is able to like mantis tech is able to create these, you know, really high fidelity contextual signals and things that can be passed through the programmatic pipes. Are you seeing publishers use the tech that way to kind of enhance their programmatic options and is that kind of the e-seeing benefits there in terms of revenue if that's what is happening? - Yeah, in a couple of different ways, I think. There's the kind of direct and programmatic and PMP deals that publishers using enriching those with mantis contextual signals in the same way that they use their first party data segments to do that. And then there's the other area that we work in a lot more over the past 18 months or so, which is the kind of curated space and then collating multiple publishers together using mantis data signals to provide at one single point of access for brands. And that's really powerful for standardizing approaches and for things that are quite difficult to actually deliver against on any one individual publisher. We find that with women's sport, for example, I know that there's a lot more content being published about it but still nowhere near the same level as kind of men's codes. But targeting that and distinguishing women's sport from men's sport is often quite a challenge. So we kind of build out segments based on that. We've got lists of the entities, the player names for the individual teams and individual sports that we're looking at and finding all of that content across 10 different publishers that are covering that sport that allows for someone like Guinness to find and actually deliver a campaign of scale, which is nearly impossible to do otherwise. - You must be doing something right, Ben, because you've been featured in MI3 video week and other trade presses, Pioneer and Contextual AI. I want to ask you which moment felt like your biggest milestone so far? - I think mine was a really big one in the past, kind of 18 months was so like it's really cool for us because we've been largely UK focused before and hour to work globally, but also work with one of the largest brands in Australia and do something different and kind of technically exciting in their space as well. So customizing for a whole new market and bringing the technology to a whole new market and changing your approach in that sense as well. Yeah, that's the most recent one. I think that's quite exciting. - I think coming up with podcast, right? - Yeah, of course, absolutely, really take your time out. I just want to get my backdrop. - Yeah, we'll send one over to you. You've got to come back on the show, though. Now, just want to get your leadership style and we've had conversations before this podcast and I've really been interested in the way that you present and you're really engaging. What would you say as far as leadership styles that has been the most valuable for you in your journey that's led you to where you are today? It's actually quite fascinating. - Well, personal questions ain't far out of them programmatic ones. Leadership style, gosh, I don't know. I think it's just paying attention to people. I think being able to understand whatever it is that you're talking about and adjusted to people's needs on the fly, I think, is I've been lucky enough that I've got a brain that kind of gets into the weeds with the systems and understands that aspect of it and lucky enough to have lots of exposure over my career with lots of different people who have made me force me to improve how I communicate and how I get ideas across. - Yeah, discipline's a big one, isn't it? - Yeah, like just, just, just, I think discipline but also it's a constantly moving space and it's easy to sometimes get a bit disillusioned with it but also the ability to keep engaged with it and keep interest being interested in it. I think a large part of it is the people in the industry and the complexity of it appeals to me but being able to find enough fascination and to keep being interested in power that kind of dedication to it. - Yeah, and on that note, I mean, everything is changing so quickly. It's the thing I always enjoyed about working in AdTech and kind of falling into that space originally is that it's always something new. It's always something to learn, some new thing to try to push the limits on. What does the next kind of 12 to 18 months look like for me and to us? What's on the roadmap that you're allowed to speak about and what's the focus for you guys? - Yeah, there's a whole host of interesting stuff that we're doing with AI, less focused on the, the kind of programmatic side, but more focused on the supporting publishers in other ways, supporting editorial teams, understanding what's happening on a publisher's site, how they can engage with their audiences more cleverly, supporting journalists with new tools. So that space is really exciting. We've spoken about the video piece and what we're gonna be able to do there by actually watching a piece of video content rather than relying on other associated signals to video and in a larger sense watching and trying to keep up with how AI is gonna change everything, providing the signals that we've got in search of the big old codes as they build out their kind of agentic AI solutions. How can we offer them something that is interesting and how can we make sure that we actually get paid for it and the publishers get paid for it as well. So the whole space is gonna be changing really, really quickly. A lot of it, it's just hold onto your butts and try and keep up. - Yeah, so it's gonna be interesting. I thought about this the other day when we were talking to I think it might have been Ari Papara, I was rewatching the episode that goes live next Tuesday, every, at time of recording for this show anyway. And so we were talking about the, who the major players are in the media industry in the next kind of like five to 10 years. Is it gonna be Google? Is it gonna, is it still gonna be Google? Is it gonna be, you know, Microsoft gonna take a bigger play in a place than there? And I mean, there's one company I didn't even think to mention at the time, which is OpenAI and with the launch of their browser Atlas and also getting into the advertising business as well. And, you know, starting to actually, you know, have OpenAI advertising platform, which no doubt, you know, that's all that stuff is gonna tie together. And, you know, that could fundamentally shift the whole industry completely. If Atlas takes off and they, you know, their ad business is kind of integrated in some way to the browser and the data. And I'm sure that, you know, effects, you know, things like what contextual analysis they'd be doing just built in with their models and then, you know, feeding that directly into their ad buying platform that they no doubt will, you know, build. So I think that's definitely one for us all to keep an eye on. - Absolutely. I think the, you can see the way that Google's search results have been adjusted and the impact that's had on publishers. The way that people are engaging with stuff online through AI means that there's less, fewer eyeballs on other parts of the internet. There's only so much attention to go around. The advertising needs to follow that. I think the, personally, I'd be quite disappointed if OpenAI launches that have kind of ad supported product. But I think it's also inevitable if they're trying to pay their extreme and extraordinary power bills with their no doubt racking up generating so-called two videos. So I would be very unsurprised if there is an ad product in that space and brands are gonna have to keep up with that and kind of work out how to engage with that in a way that means big shifts in how they're planning. - I guess the benefit that you guys have is kind of going back to like the viewability and brand safety theme. My earlier question around publishers building their own contextual tools. There's always been a benefit to that element of being an independent player in that space, right? When you've got brands and publishers and then you've got the companies in the middle who are providing the independent analysis and measurement and things like that. So even I would like to believe that even with something like at OpenAI ad business that there's still a role for those independent companies like a man just to play in that space. - Yeah, I think it's gonna be, as you were saying, everything's gonna move very quickly. So adjusting the approach to whatever new universe turns up in two weeks from now and then six months from now is gonna be key for all of us. - Yeah, that leads me into probably a wrap up question for the show. If you could leave viewers with one thought on what inspires you most about working digital media, what would that be? - For me, we're all at the forefront of a whole host of really exciting technologies. That's the part that appeals to me. The way that systems are talking to each other, the way that AI is being integrated into our workflows, the way that can't and how we're supporting news. We get to touch on so many different industries and technologies and behind all of that, there's a whole lot of very nice people to actually deal with and interact with as well. So that's the stuff that keeps me interested and engaged with it and that's the appeal. I don't think anyone has come into media and said, oh yes, this is my first career choice. I'm gonna help sell ads but lots of people have stayed because it's like I could do my head under the table right now. (laughing) Oh geez, there's always one sorry. - Yeah. - It's Ben, well thank you. I just again, it's just fascinating. I'm really keen to see what happens and evolves through the mantis here in this region and obviously throughout APEC as well. It's such a big space for you to play in and good to see someone like yourself at their home helping steer the boat through these amazing uncharted waters to recite that. - All right, thanks for having me on, guys. I really appreciate it, a lot of fun. - Thanks man. - Thanks for being here. - Thanks man. All right, enjoy your, well it's both morning. Enjoy your coffee. - Second coffee, yeah. - Good idea. - And Felon, thank you so much. General manager at Mantis, look forward to seeing you again next time on MediaTekTorp. - Thank you, see. - Like it, that was really interesting. I've found Ben's approach to how they're working with nine in this country really fascinating. Obviously with Oracle and Moat moving out, the way they did leaving publishers in the lurch really, and an opportunity for Mantis to come in and sweep up a big publisher like nine is huge. Especially when you've got IS and DV in the market, so click to try and get these guys. - Yeah, absolutely, it's a shame what happened with the Oracle shutting down that area of the business and it's really cool to see organizations, like Mantis come out there and really fill that gap and push the envelope and really kind of innovate in that space, contextual and brand safety. It's been something we've been dealing with for a very long time now, and the vendors in the market have changed from, as you said, IS and DV, measurement of verification companies to now data companies getting involved and dedicated businesses like Mantis who are really solving those problems and innovating. So it's great to have Ben on, great to get a look behind the curtain and have that conversation. So, and for everybody else, join us next time for the conversation that's inspiring Media Minds. Here on Media Tech Talk. (upbeat music)