Articles, Blog

NOAH12 San Francisco – Proximic, Rodney Mayers

January 21, 2020


My name is Rodney Mayor’s. I am chief revenue officer of Proximate. We are Data company We are now based here in Palo [Alto]. We’re actually founded in 2006 back in Munich Germany And and you know one of the things that we focus on is Kind of maybe a little mundane, but it’s the most basic fundamental asset of online advertising And that’s the page and so today what I want to share with you is a little bit about how we approach Pages How we look at pages and what it actually means to our to digital advertising We’re afraid that I go Um so headquartered here in in Palo [Alto] We have about 40 customers right [now] all across all across Europe and russia will be moving in or back in to ASia actually in Q4 of this year And what we specialize in is highly granular page level data Most of you may know that as contextual data, but something interesting happened You know back in 06 when we were started we were trying to understand. How do you actually? Get meaning in Data out of unstructured content. How do you how do you sift it? How do you find it and we looked at semantics? we looked at kind of the things that were around at the time, and it didn’t quite work for us, so we pioneered a brand new opportunity with both non linguistic Non semantic structures we use pattern recognition and what that does and allows us to do is to actually take a page Break it down into its component parts and provide very granular data about the about the page itself What this [also] does is it allows us to handle languages in a very unique way So most partners are partnering Russia video international came to us and said we cannot find anyone to help us With our advertising can you help us? Understand and do contextual analysis and brand safety And we throw our machines at it and in about a week or two We came back with at least the base understanding of the russian language, and we’ve been working with them ever since in in handling the Nuances [they’re] from to investors have been with us from the very beginning Wellington partners and Holt Sprint ventures also, based in Europe So what is this thing the the advertising environment? What is rodney talking about you know in in the early days? And I’ve had the good fortune of being in the industry for quite some time going back to traditional media roots is It was so easy back into in the late 90s. You kind of picked your portal you picked your destination site You said this is great. This is safe. I know what this is if there was a traditional media component. You’re like yet I know what that is offline? so it must be the same online and then this funny thing happened call RTB and Digital programmatic buying and the hell that broke loose from there was that you are now buying Impressions at the impression level you are flying around you were buying everything then social came in then meet then mobile came in and then Of course this video and brand safety took on a different a different element. It was no longer about Just the site people don’t consume sites they consume pages it was now about What is the environment [that] I’m actually entering and so you know you can say yes? I’m buying forbes or I’m buying Cnn, but when there is a story about name your category a cruise ship that sinks off the coast of italy or Mexican drug lords cut off the heads of people and drop them on the highway that is not brand safe Brand Safety is not just about porn. It’s about the environment that you’re entering Also, you know we talk a lot of we’ve heard a lot about Viewability well for us looking at page quality is also something that You should factor into your ad impression that you’re looking are you around seven ads or you in an environment with three ads? Are you in an environment with 14 links versus original content? What does that page look like that’s what we focus on. We neither buy or sell inventory We are a pure data provider and we work with with partners across the industry Our technology as I mentioned earlier is based on something called Pattern Proximity profiling in a nutshell What we do is our spiders actually will go out much like a search engine or [Riva] page And we will analyze it for all the elements. I discussed earlier. So we look at context we look at the environment We look at whether it has a certain brand safety flags We follow the I abs on quality assurance guidelines So we look for porn and terrorism and weapons and everything else that we can find on that page And we relay it back to our to our partners one of the things we found with with this particular approach Is that it was very good at handling user-generated content not just user-generated content at the bottom of a page? but in strict UGC environments How do you tell the nuance of the conversation when dealing with the blogroll how do you parse that and actually understand? What is happening on that page at that time within a certain view and as we discussed earlier? It helped us handle new languages as well With that approach though something very interesting happened. Most categories and let’s say the iab categorization handles about 372 about 390 categories what we wound up with in our data structure was about 350,000 categories no one in their right mind would try to target on anything that deepened that granular But that granularity is important for targeting and for understanding exactly what’s happening? So what we had to do then was develop a second part of our technology called dynamic Category matching and all that says is we take all of that unstructured information We take all of that data sitting in our nodal database And we transfer that and map that into iab into Google categories into ebay Categories are into your own Taxonomy that’s very very powerful because as we found There is no one singular taxonomy that defines the entire scope of every language or the entire scope of categories But what’s important is if there is a category? That you cannot necessarily find or you need to deal with nuance you need to deal with that at a structural level So a case in point is again with our partner in Russia Porn as we see it here in the united states is defined at the data level in a very unique way but dealing with russian porn and it sounds kind of funny, but dealing with Russian porn It’s actually a far more subtle game and what we did with our partners They said look here are the [key] [words] that actually make sense? Here’s what we see when we look for porn here the environments We are trying to avoid we took those keywords we put it into our system And we remapped it back into a new structure that allowed us to identify in Russian environments that they wanted to Avoid That’s very very powerful and very unique and as we look to go back into China in Q4 It allows us to handle those environments Especially for brand advertisers that are looking to avoid certain certain environments This for us is what makes us unique is what makes it special? And [in] all the the large agencies and their analytics teams as well as dSPS and others that we work very very closely with This is one of the key differentiators that they find extremely powerful as well so what does this all mean, so I come to a page and You know in the old days or even today you can say hey, I’m a xyz site. Here’s my channel. It’s called Sports And here’s my subchannel let’s call it outdoor sports well sometimes what we’ve seen is in many times one page Just one page can qualify for three or four categories The self-declared category by the seller of the inventory is great for a very high level understanding of the page But on the opposite side the actual data level we may see that that one article that one piece of content is Associated with news sports outdoor and even travel map in that environment is important, not only because yeah, you can do it but if you’re trying to bring demand to your inventory if you’re trying to Target and Bring more advertising to that one particular page you may actually qualify that page from more than one advertiser With all the other attributes of brand safety and page quality associated with it as well Taking that page and breaking apart. We do the contextual Analysis we also look at brand safety as I mentioned and then we also look at page quality What does this page look like one of our our large agencies came to us and said you know can you help me define? The environment can you help me understand are there more than three ads on the page will I get high share a voice is The quality of the pay actually a good environment for this particular brand and each brand has completely completely different standards for what that means an Interesting thing and you hear a lot about audience targeting Today and also just in the industry is you know I think we sometimes forget the audience’s show up Because of the content into pages and so how do you relate all of that audience information? Actually to content and so what we do on behalf of our partners is we actually will take their log file data we will actually process it and cluster back for them the connections of Their own audiences to the content those audiences were engaged with that that mapping of content to audience is extremely important for one reason if You’ve got an automotive at and you’re targeting in an endemic channel like automotive because you think it’s an auto entender But you actually find performance in a different category You may want to do look alike content targeting you may want to go after that audience member in an environment that matters to them? But somehow there’s a connection there that you did not see we try to surface that intelligence We try to surface that data and provide it back to our partners so that they can improve their advertising This is extremely important for our brand advertisers. I’m retargeting direct advertisers find this interesting Kind of more in the periphery, but brand advertisers find this very interesting because it reminds them of what they buy in traditional Media when they buy television they know their environments when they buy a radio they know their environments And when they buy print they know their environments And they know which environments perform we’ve been very successful in bringing this to market our business model is straightforward It’s a combination of licensing as well as usage All of our page level products start at about four cents cPM Our audience or user level tends to be more of a licensed product and is very customized to each each one of our partners Everything we do is dropped into our own unique analytics cube. We also again. We looked at hadoop We looked at all the the typical frameworks that were out there, and we actually have to build our own System to actually handle this much data. Especially on a daily basis and provided in a customized Ui The key to our our business model though Especially on the usage side is you only pay when your ad is actually delivered to the page? So we don’t just charge you for processing we tend to do that on an ongoing? Basis and in order to do this well you have to process lots and lots of impressions We are quickly approaching half a trillion impressions a month processed This is globally. We think once we bring China online and continue our growth in Europe We will probably hit about 500 billion impressions a month processed The the actual Data that comes off of that Is measured yet in terabytes per day, but what we actually do is Louie meunier? Who is our chief scientist some of you may know Louis? From his days at Alta Vista. He was a co-founder of Alta Vista He helps us tackle, not only language, but also some of the heavy heavy lifting on on processing to that end It’s a very seasoned team We do not divulge revenue or Profitability numbers as a small private company, I will be around however to to speak with with anyone [that] is interested Again, thank you [for] your time and attention I know there’s been a lot [of] us up here already this morning But we are proximate and we look forward to working with many of you in the future. Thanks very much

No Comments

Leave a Reply