Commentary
Targeting Video From The Inside Out
- by Phil Leggiere , June 20, 2007
Behavioral Insider: Where do you see multi-media search and targeting today?
Thomas Wilde: Multi-media content till now has remained almost completely opaque, in that search tools cannot 'see' inside videos. Instead all they have to go on are titles or at best meta-tags provided by publishers, which in most cases only provide thumbnail descriptions. Imagine if Web search for text documents had only been able to deliver results based on titles without referencing content. It's as if you're in a clothing department store and instead of having clothing displayed in an organized way on racks by style, color, size etc. customers had to dig through storage boxes just marked "shirts" or "pants" to find what they were looking for.
BI: And how are you addressing that opacity gap?
Wilde: So what we do is dig into the spoken words of a video or audio file. Our text-based search results then include snippets from the files. and the ability to click on keywords from that file that are most relevant to search query. From a navigational point of view the technology allows consumers the ability to obtain entire transcripts of an audio or video or to navigate by finding the exact place in a file where a specific word or phrase or line is spoken. Up until now there's been close to zero ability for content owners to merchandise their materials or for consumers to know what they're getting.
BI: How can these discovery and consumer experience enhancements translate into monetization?
Wilde: As far as monetization goes, thus far only the coarsest types of targeting have been available for multi-media. If we take video, an advertiser may have a very general idea that video x is geared toward a certain demographic group but it's nothing granular. It's comparable to television's estimate of demographics but without the scale to work with.
The most immediate and obvious use of what we do is contextual. Advertisers can now run an ad in association with a particular video clip and have a very high degree of confidence about the relevance and appropriateness of the placement. But we see behavioral data as being the next and, in many ways, most important application.
Behavioral targeting is a classic example of the concept of 'garbage in, garbage out.' That is, the higher the quality of input, the higher the quality of the behavioral profile that emerges -- and of course, vice versa. So we see the ability to really classify video file content in detail as enabling advertisers to better understand just what content specific users are really interested in. An example is, if you have a viewer who looks at videos, say podcasts of technology news from the NY Times, Macworld and other consumer oriented downloads, podcasts etc. You don't now know, just from an inventory of the videos they've viewed what they're really interested in other than technology very broadly. But if you dig into the files they've been searching by keyword or phrase or specific topic, you can identify and differentiate people interested in learning about a new IPhone and people looking into computer security software or a new PC.
BI: Beyond content interests, what other types of behavior will be trackable?
Wilde: Once we understand the 'what' of video usage behavior, then it will actually become possible to get a meaningful look at the 'how' as well. For example, does this user view clusters of 30-second clips or full-length clips -- and, if both, when and why do they prefer one or the other. What formats of video- or audio-based ads do they actually pay attention to and which do they skip, which videos do they forward to peers, which ads elicit which kind of response.
Having a truly detailed knowledge of the multi-media content also means that ad serving doesn't need to be done in the dark as it has been. Video ads can be served within video documents -- not only based on the subject of videos but in terms of the other behavior of video watchers. Then, once you have a fix on what users are doing with video and how, you can follow a variety of behaviors over time.
BI: What's your near-term goal for moving clients into a more behavioral perspective on video?
Wilde: Over the next six to 12 months our goal is to push the process forward in educating both consumers and content providers. The more control consumers grow used to having in getting the exact video document they want, the more purposefully they will use videos. The more that happens, the more video publishers can grow truly merchandisable inventory. The more personalized discovery and usage becomes, the richer behavioral profiles and ad serving strategies can become. That's the virtuous circle we're looking toward.
advertisement
advertisement
