Video Targeting: Simplifying The (Still) Too-Complex
Online video is a lot of things, exciting and engaging certainly among them, but not, from an advertising point of view, very simple. The still far-from-completed challenge of online video -- if it wishes to become a truly integral branding channel (one really capable of commanding a solidly double-digit share of branding budgets) -- is to distill the still overly complex data generated by video usage online into a more transparently targetable channel.
To try to gauge the present state of online video targeting , and where it's likely to head in the next six months, I spoke with Waikit Lau, co-founder and president, ScanScout, one of the emerging video ad networks engaged in trying to simplify online video targeting.
"Twelve months ago the questions from agencies were still mostly about video ad formats," Lau told me. "They were about, 'Is it pre-roll or overlay?' Now, though standardization is still an issue, the conversation has shifted to the question of how you can scale online video while increasing value and ROI. That signals a real step-up in the level of seriousness to advertisers in this channel, but also huge challenges online video frankly hasn't lived up to until now."
The problem with online video targeting, Lau believes, is that it's been hard to do video optimization, primarily because of a lack of meta-data. Data has rarely gone decisively deeper than a title and a thumbnail description. The problem is, there's been no way to really understand context.
"There are three signals that are critical to understand video," he explains. " One is the semantic content within the video, the audio component. Another is the video content, what's specifically inside the video, the specific actors and genre as well as the subject. Perhaps the most critical component is the user and where and how they watch videos. That means the site they are watching on and who the viewer's are demographically, which can be provided by third parties such as Blue Kai and Quantcast."
By integrating these three streams of data advertisers, Scan Scout, along with other leading video targeting platforms such as YuMe and Tremor Media,are betting, can for the first time bypass the limitation of traditional video buying, which has been to buy context as a proxy of audience.
Lau gives the example of 18- to 34-year-old male car enthusiast. The question online video networks have been struggling with is how to differentiate segments meaningfully, such as between a casual car enthusiast and one who is an in-market auto intender.
A combination of enhanced meta-tagging and third-party audience can enable video behavior to now be tracked along several critical dimensions, he believes, especially the frequency with which auto videos are viewed over time (Is he watching more often than three months ago?) and the "intensity" of viewing: whether the viewer is watching specific kinds of videos for longer periods of time.
Another key difference between targeting online video and conventional online display, Lau believes, is that it's far more important in video than with banners to optimize both the creative and context of viewing.
"There is incredible variation in response to video creative based on time of day," he says. "So it's not only a question of what ad content and creative is shown, but when it's viewed. It's also really important when you're targeting consumers of longer-form content to test when ads within the video work. There's evidence, for instance, that ads at the start of the video work less well, perhaps because at that point consumers are anxious to get to the content of the video they've chosen. But we're seeing that ads introduced at various stages of the video watching, minute 7 versus minute 12 of a 30-minute video, can have dramatically different responses for different viewers and different videos. That's a whole other sphere of testing and analysis that's barely been scratched and that's highly relevant to the effectiveness of the online video channel."
One finding that's been consistent across the board is that engagement levels -- the degree of interaction with online video ad units of all kinds -- are much greater than with any other channel. The question has been how to effectively integrate the (for media buyers still-novel concept) of engagement into metrics and reporting. How can and should engagement be interpreted and measured as value?
"I believe there will never be one uniform metric for engagement," Lau predicts. "Rather, engagement metrics will evolve in very specific ways for each vertical. CPG advertisers, for instance, might think and interact strictly in terms of time spent viewing ads. Entertainment advertisers promoting a movie will look at the number of views of movie trailers. An auto manufacturer might decide that the best measure of engagement is how many times viewers spend more than two minutes on their car configuration page."
Beyond the engagement measure, there's still a wider need, Lau acknowledges, for agencies and advertisers to become more comfortable understanding how to read all the data generated by video usage online.
An example of this misunderstanding is that video advertisers, largely because they haven't been taught to know otherwise, still see their goal as getting viewers to "click through" to their sites. In fact one of the most valuable aspects of online video ads is that you can bring your site to users without forcing them to disrupt their entertainment to travel to your site. But video advertisers must realize that this tactic will bring their CTRs down.
On these issues, the growing but still (compared to its great expectations) nascent channel of online video has a window of opportunity to use research and education to enable advertisers to do what they're ready, but still somewhat confused about how to do: committing the next serious chunk of their budgets to online.