Scaling The Video Opportunity
Until recently, applying behavioral targeting against video advertising suffered the obvious problem: scale. Even as metrics companies report upwards of 11 billion monthly video streams, only a fraction of that impressive number represents credible inventory for most advertisers. Last month, video ad network YuMe decided that a potential reach of 130 million uniques across more than 400 publishers offered enough coverage to start offering behavioral targeting. The network accesses inventory from MSNBC, MSN, and Fox News along with many more mid-tail providers of non-user-generated content. According to co-founder and President Jayant Kadambi, advertisers are asking to align campaigns against professional-grade content. But will such plans scale?
Behavioral Insider: What modes of targeting does YuMe already offer?
Jayant Kadambi: We can target demographically when we have the information, by geography, by contextual mapping of the content. We continue to find the technical ways to slice and dice the publisher-based content so we can meet the advertiser requests.
BI: Do we know that those layers of targeting increase ad effectiveness when it comes to online video content?
Kadambi: It depends. In general terms, yes. If you have done a buy and are an advertiser on our sports channels, we can say here are 50 Web sites and 100 pieces of content running on those sites or phones or downloads. We will monitor different metrics in the background and in our network, and in run-time we make modifications to where that ad is being placed based on how the ad is performing. If things aren't well contextually mapped and the agency says let me throw it on content A-B-C and we find out that people are fast-forwarding through the pre-roll, then we move it to content D-E-F to see if the fast-forward rate decreases or the click-through rates increase.
BI: On the tech side, where are the behavioral profiles coming from?
Kadambi: We have access to multiple sources of companies as well as direct data sources to get intent data from shopping, travel, and auto.
BI: Are you working with third parties for this?
Kadambi: It's not coming from one place. Our opinion is there is no one place where I can get all the data in order to make the overlap and the scale of the data worthwhile. Now that our network is very large, the intent data we are getting in these three sectors we believe has fairly large overlap. An advertiser will get a fairly large audience of in-market car buyers, for instance.
BI: Is that the level of segmentation we are talking about -- broad segments, or more granular?
Kadambi: It is much more granular. As you know, though, the more granular you get the smaller the audience becomes. If you say, I want a guy who has visited an auto site in the last day and got to this portion in the funnel and had a dealer call him and wanted a hybrid car, we can do that. We're still experimenting with how big the segment should be vs. how small the audience you get. The question on the table is, how small should we make the segment so we offer the advertiser a reasonable reach and audience?
BI: Explain the serving process.
Kadambi: We make all our decisions in run time. For example, we might do a behaviorally targeted campaign for in-market shoppers for refrigerators. When an ad request comes in from any user who hits our network, we will make a query at the time of the request to say, OK, what is the profile? Is this user an in-market refrigerator buyer? If that comes back, we will target against them on in-stream video whether it is an overlay or pre-roll or post-roll. And that will be across our whole network.
BI: What differentiates this then from other BT?
Kadambi: The two differentiators are, one, it is intent-based. And two, the dirty little secret about behavioral targeting and video is the networks are not large enough generally speaking until now to get a target that proves a meaningful reach to the advertiser. We believe we have that now. We can offer this at scale, which is very important to advertisers. Reaching one refrigerator buyer doesn't help you. Even reaching 100 or 1,000 doesn't help you.
BI: What do we know about different responsiveness between video ads and other rich or static media placement?
Kadambi: We have gobs of data for that. Regularly for the last year, we have been seeing probably an order of magnitude increases in click=through rates on video ads, even vs. rich-media ads. If it is contextually placed on overlays or general brand lift, all metrics across the broad seem to do very well with customers. Spends are going up. Everything is looking pretty good, actually.
BI: Video advertising tends to be aimed at brand advertiser, while behavioral targeting is a favorite of direct marketers looking for some kind of direct response.
Kadambi: We've got big car advertisers and big brand advertisers on our network. They are always interested in intent and behavior data. They don't want their car ad to go to someone who just bought a car yesterday, so that is an added feature that attracts them to advertise with us. So we believe there is a large contingent in the travel and auto and shopping category of brand advertisers that can make use of BT as part of their video campaign.
BI: But is their goal in this case different here from their banner campaign?
Kadambi: Their aims are brand related. We have been saying for quite some time -- and this is one way of proving this to the brand advertisers -- we will give you a strong correlation now between your $1 million or $2 million brand spend because you will be able to see how many people actually click through to your site and whether they bought something or not. You can measure the effectiveness of the brand campaigns with something other than a click or a fast-forward. We are just providing more data. If it is not effective on one site or one piece of content, I will start moving it around.
BI: Of those three segments you are starting with, shopping, auto and travel, which is the most interested right now?
Kadambi: We're getting lots of interest on both the travel and auto sides.
BI: Does BT targeting favor different types of video creative over others? If you are able to ID a car intender, you can assume they are at a different point in the buy cycle, and a typical brand pitch might be ineffectual.
Kadambi: We absolutely remarket as part of the buys to the auto and travel guys. If we have the intent data and the advertiser gives us three different creatives, we will use the three. And if we see you again we will remember that we gave you that previous creative and give you a third. That capability is there. The caveat here, to be frank, is that it isn't often that we get five creatives.
BI: Is there a premium for the BT layer?
Kadambi: We're still testing the market on the premium as we roll out this beta, but we expect it to give about a 25% lift.