• by Steve Smith , Staff Writer @popeyesm, May 12, 2006

Someday, when technologies like IPTV, personalized search, set top boxes, DVRs, and privacy concerns all finally plug together, we will be able to serve TV viewers ads according to what they've already watched and done in the converged mediaverse. One baby step toward that future ramped up this week, as rich media and in-stream ad network Tremor Network partnered with Tacoda's network to begin incorporating behavioral profiling in its ad targeting. According to Jesse Chenard, CTO, Tremor, BT offers a way to monetize some of the most general and user-generated video inventory that is otherwise difficult to map against highly targeted ads.

Behavioral Insider: How much in-stream video inventory do you have in the network?

Chenard: We have access to about 100 million avails per month. And that equates back to about 15 million unique users across a network of about 30 sites. But that's expanding rapidly.

BI: Explain how the behavioral targeting folds into the in-stream serving. Are you getting the profile data from users' previous visits to standard pages rather than from previously viewed videos?

Chenard: Exactly. That's part of the deal with Tacoda. We're leveraging the Tacoda data points. So across Tacoda's network of sites, if we get a match against a behavioral campaign we sold with video, we can leverage what Tacoda has there.

BI: So if a user comes into one of your networked sites and he has already been profiled by the Tacoda network, then you can follow him into the video player and use that profile to target the ad.

Chenard: Exactly. There are a number of different players out there, and some have done it in different ways in building closed platforms for doing ad decisioning. When we built our ad in-stream product we specifically worked with interoperability with the existing servers out there. Initially, we wanted to focus on the core insertion technology-- and in the end, it's turned out to be a great benefit, because advertisers want to use their own servers. Publishers want to traffic their site in tools they are used to. The other added benefit was that as new targeting technologies come out, we can leverage those.

BI: Behavioral targeting is often pitched as a solution for remnant or low-CPM inventory, but in the case of video many sites are already selling out their contextual pre-roll inventory. How does the BT pitch work in that kind of market?

Chenard: This is the big misconception, that the video inventory is extremely constrained out there. The mainstream video inventory is extremely constrained, the Yahoos, MSNs and maybe some of the tier-two sites. But once you get below that, and especially once you get to the user-generated content, we're talking about billions of impression that are going unsold right now. Our thinking is that these user-generated sites are going to need to make a little money.

The YouTubes of the world, who are burning millions of dollars per month just slinging their videos out there, are going to need to monetize. At the lower CPMs, even the sub-$10 some of these guys want to offer, makes it really hard for them to make their money back. The thinking is that if you can add some value, whether it is from the BT [profiling] or registration data so we can know that much more about the user, that's where we really see this being able to bump the CPMs up. In a generic news site, the news content may have otherwise gone untargeted and may have brought a run of network ads there at a sub-$10 CPM. If we know that user has been to an auto site or is an SUV enthusiast, we may serve them the right auto ad at that point, bringing in a $25 or $40 CPM.

BI: Are you specifically targeting the user-generated video sites? The randomness of user-generated video and its haphazard user tagging doesn't seem to map well against contextual targeting. It is hard to know what ad is relevant to many of these clips. BT would seem to divorce targeting from the immediate content, so you can ID existing interests of users and hit them with a targeted ad regardless of the clip they load.

Chenard: We have talked to some of them. I met YouTube. They're really still interested in taking on some more cash and not harming the user experience at this point. But later in the same discussion, they did say we really should talk in a month or two. With the press we have been doing, there is an amazing amount of inbound calling that we're getting from potential publishers, and a lot of them fall into either user-generated or user-submitted type sites that are popping up.

BI: Even using the much larger pool of Web page avails, BT segmentation can quickly diminish your reach for a campaign. With video, which starts with only a fraction of the inventory of static pages, it seems achieving acceptable reach on a BT campaign would be a real issue.

Chenard: That will be the challenge for us, obviously. As we get a little more historical data going through us... and we're seeing where our overlap is, we'll definitely be able to answer that question a little better. I do know, though, that from some of the anecdotal data matching we have done with them, that a lot of the people watching video for the most part are the ones in the right demographic of where most of our advertisers sit. So we're lucky there, that the early adopters of this medium are also the ones that people want to hit with their advertising.

BI: What sort of price premium are we are talking about over standard pre-roll buys?

Chenard: From anecdotally talking to the salespeople, we definitely see a 25 percent to 30 percent lift on the price, and some of the people we're talking with don't seem to be blinking at that.