Targeting Ads on TV
Since Google changed the face of Internet advertising, many companies have adopted solutions based on Contextual Advertising, where ads are selected based on the contents of the Web page/Email/IM. In some cases, the user’s profile, if known, is also used to better match ads.
Mobile Advertising is also based on the context but usually relies heavily on the user profile as each handset is directly associated with a specific user. Since mobile operators can collects a lot of information about their users, the generated profile greatly assists in the matching process.
Unlike Internet and Mobile domains, it’s not trivial to understand the context of a video stream and to identify the person watching it. First of all, how can we know whether the TV is turned on when even the Set-Top-Box (STB) is unaware of it? and even if we knew that the TV is on, how can we identify who is watching the program?
In the early days of Interactive TV (iTV), applications used to require users to ‘login’ to the service however this approach was not accepted by users and was neglected along the way. Solutions that attempt to build user profiles for TV watching habits must find a way to bypass these constraints, either from the STB side or with assistance from a server-side application.
But the questions still remains – is it really worthwhile to target ads on the TV domain? In their article “Using Data Mining to Profile TV Viewers“, Spangler, Gal-Or and May show that targeting TV viewers creates a lifting effect that increases the changes of having a household person that matches the ads’ intended segment.
“The table here outlines the gains from viewer profiling for five target gender/age segments defined by Nielsen Media Services, Inc.; for a discussion of segmentation strategies. If an ad is sent to everyone, 25.18% of the recipient households will include a female aged 18 to 34. If an ad is sent only to households selected by the profiling system, 58.06% will include such a person. A household selected by the
profiling system group is therefore 58.06%/25.18%, or 2.3 times more likely, to be of the desired type than one selected at random. This ratio is the model’s “lift”; the lifts in the table are different for different demographic groups because some groups’
characteristics are easier to predict based on their viewing patterns”.
Several companies have already started to work on such solutions, although most of them are in stealth mode. An example of such company is Invidi, which already provides tools for various sorts of targeted advertising.
Permanent link