Turn CEO Bill Demas discusses what his company offers in terms of linking consumers across devices.
This is a continuation of a series that previously featured John Nardone, CEO of [x+1] and Omar Tawakol, CEO of BlueKai. On Wednesday we’ll post an interview on this subject with Scott Howe, CEO of Acxiom.
In terms of cross-device insights, what are your clients asking for and what can you deliver?
Marketers don’t want their data siloed and their channels siloed. The same is true for devices. You want cross-channel media strategies and a cross-device strategy. We’re investing in cross-device technology and we have elements of that today. We feel cross-device is an imperative, and it’s something you’ll see a variety of folks, including Turn, in terms of offerings in 2014.
What have you invested in and how does it work?
Yeah, so the problem is I can’t tell you today. There are some potential patents pending and we haven’t announced anything yet. When we get closer to market, then I’m happy to chat with you again. I can only talk in broad-level strokes at this stage.
So generally speaking, what are the technological challenges DMP providers see in linking mobile?
There’s a cookie-based world in display and a non-cookie world for devices. When you have a large mobile advertising business, you don’t target by cookies. There are other ID technologies we use. We’re good at the relationship mapping at the user ID level. Then you have to match areas of the cookie, areas of URL content and mobile information. We stitch that together so we know, anonymously, ABC123 is the same person on a certain device as they are on a certain PC. It’s a lot of user ID relationship mapping technologies that incorporate some of what you learn from cookies as well as other technologies to stitch that together, since cookies aren’t available on mobile devices.
How do you stitch this stuff together?
It’s very sophisticated technology because what if one ID source is different than another ID source? You have to figure out what if one ID says you’re a man and another says you’re a woman? I’m simplifying it, but the algorithms we use help us figure out the probabilities of what that user profile is.
If you’re on device tracking, say you’re on the mobile phone but your sister uses your mobile phone too. Say she’s buying dresses. You have to be able to infer: What is the actual usage pattern happening on the device? It’s about taking this data and creating algorithms and making inferences to create the best anonymous user profile you can.
What’s the accuracy rate?
The easy answer is it depends. We wouldn’t grow our business unless our probabilities is very strong.
What in the industry is considered a good probability?
I wouldn’t know, actually. I haven’t seen that as publishable information. They have to be reasonably high.
Drawbridge and a few others said 50-80%. Omar Tawakol from BlueKai claimed 70% is considered good in the industry. Is it possible to get 100% accuracy?
One hundred per—?! Okay, remember in our case we have 1 billion anonymous user profiles. Are we going a billion for a billion? No. We can tell by our customers, either they tell us we’re targeting well and generally we get very good feedback.
