AdExchanger: What kind of data science does EveryScreen Media provide and where does it fit in the ad ecosystem?
LAUREN MOORES: The best way to think about it is, say you want to find your desktop audience on mobile. We both know there aren’t cookies or a universal ID that we can use to find that audience on mobile. So we’re using our data science to find lookalike audiences. We do that going against a longitudinal database of location and user scoring that we have a patent on. We call this the Geo-Social Network.
What we do is take the information that we get from an exchange and create a connection between the IDs sent by them. It could be any ID, like an IFA or Android ID. What we’re hoping is that it is as consistent as possible for that particular exchange. Even if it isn’t, that’s okay since we’re essentially using a deterministic probability to determine whether or not me coming through exchange A on device ID C is the same person coming through exchange B on the same device.
In mobile, unless you’re using IFA, which gives you the ability to go across exchanges and apps because it’s not being hashed, every exchange has its own ID and it’s being hashed differently. You can’t rely on the ability to create a single profile across apps or mobile, let alone try to do it cross-channel. That’s what we use the data science for.
Does your approach take into account concerns about consumer privacy?
Any information that we’re using, we’re doing in a one-way hash such that you can’t reverse that information. We also realized that we needed to make sure the system we built could use whichever data is acceptable at that point in time. You could use location as your main variable in creating geo-location networks. However, not everybody agrees that is in the clear in terms of not being PII [personally identifiable information]. The exchange doesn’t need to send me the actual data, either. We could use some sort of identifier that indicates a consistent pattern that could be hashed.
What is the social component of your Geo-Social Network?
I’m not using social network data at all. What I’m looking at is the data that’s sent to us through the exchanges from the publishers, so when somebody uses their tablet or smartphone and the ad request is coming from the publisher and is being sent through an exchange, that’s the data that I’m using to create a location distribution profile for devices.
Are you also working with CRM data?
The DMP side of the business is very flexible. We can use first-party data and third-party data. For first-party data, we can do a couple things. One is hyperlocal targeting since clients often already have some segmentation and we will create an audience for them based on that info. Another thing we can do for customer segmentation is instead of getting a pixel file from a desktop, we can get the customers’ info as it’s defined in a client’s CRM database and we can match that to our mobile audience.
What type of trends are you seeing in mobile?
Location is key in mobile. You can infer a lot of behavior through location and a lot of our clients’ campaigns are being done through hyperlocal segmentation. We can take that data and use dynamic ad serving to find somebody in a particular area of a store and target them with a creative for another store in that area. You can also include weather, so you’ll serve up creative based on what the weather is at that time. We’re getting asked to do a lot more of that [weather-based ad targeting].
What has also changed over the last four or five months is there is more competition on the RTB side. We have been bidding through our different exchanges for over a year and many times we’ve been one of the few buyers. Now I’d say that the competition for the winning bid has doubled or tripled in some cases. I think brands understand that for mobile, going with mobile RTB is not going against the grain, because that’s where it starts. Whereas on display, you still have questions like ‘why would I do RTB or why would I allow my inventory to go through an RTB channel when I have all these direct relationships?’”
What’s on your roadmap? What will you be focusing on in the coming months?
The idea of EveryScreen was to give our clients the ability to do a cross-channel campaign starting on the desktop and hitting other devices, including digital out-of-home and smart TV. We want to work closely with some desktop partners so we can develop a more comprehensive cross-channel campaign system with the ability to measure that.
That’s the hardest question out there—how do you provide attribution for desktop versus mobile versus any other channel? My goal specifically on the data side is to not only get closer on the desktop side with our partners, but to also start using more third-party data to build a stronger audience. As the bid requests that come through publishers change and as the exchanges add more premium information, it provides you with more signals to create better look-alike audiences.
Also, we’ve been working with different sources out there that could help you create a tighter connection with users. For instance, if you’re using a TPID [Trusted Preference Identifier] from Truste, which is not a behavioral ID, but an opt-out ID, seeing where the industry goes in having that solution that crosses channels and devices could also be interesting.
So reports of declining PC sales don’t faze you?
Online advertising still has a huge budget and that revenue is not going away anytime soon. We do have a particular approach on the data science side that would allow us to create look-alike audiences if cookies went away. We don’t rely on desktops but we need to have a complete solution. Not only is it about working across channels, but working across media. Adding to the mobile impressions by making sure that you’ve got desktop impressions, video impressions and some of the more niche data like out-door home and Smart TV, even though those are still in beta form. That’s where we want to head. Being able to read that data and provide a comprehensive audience. It’s what everyone wants to do, right?