Advertising Week Experience Blog – by Brooke Aker, ADmantX CMO
Over 70 billion dollars was spent worldwide on digital advertising last year. This large number covers publishers, agencies, media placements and ad creation. It also includes a phenomenal area of technology that helps find an audience for the advertisement and optimize the match between ad and content and exchanges for ad buyers and sellers. Optimizing the match between content and ad is a fundamental step in making the economics of digital advertising work for buyer, seller and consumer.
This optimization is more true for newspapers than most other types of publishers as the transition and overlap between print and digital causes tension.
A variety of targeting approaches seek to improve the ways in which an ad is selected for use on a content page. A 2011 research study conducted by Yahoo suggests “people spend 25% more time fixating on ads that are personally relevant to them”. Yahoo also found that consumer ad “fixation increases by 15% when ads are contextually relevant. This increases the chances that the ad will be stored in long-term-memory and ultimately leads to higher recall.”
So, targeting can significantly improve on consumer objections to advertising and therefore optimize the media and advertising spend. There are many types of targeting. These are listed below as a matter of introduction, or perhaps review for those in the industry.
– Demographic: Traditional targeting based on data such as age, gender, income and ethnicity.
– Geo-targeting: Targets a consumer in a certain geographic area using location data mined from the ISP or IP address. A powerful tool for local business objectives and marketers, it allows the display of local product inventories to customers in the case of new rich-media ads.
– Behavioral: Tracks the actions of thousands of users as they surf the web and aggregates them for trends. These patterns become the basis for targeting and can include purchase history. One example: A visit to Edmunds.com can be the basis for serving an auto ad even after a consumer moves on to a non-auto-related site.
– Contextual: Matches the advertising to the content being consumed (whether a text, audio or video file). For example, an article about the outdoors brings up a camping ad.
– Site-targeting: Similar to contextual, this tactic matches the ad to the theme or genre of a publisher. A high-definition TV marketer may choose to advertise on sites dedicated to consumer electronics.
– Day part: Just like offline, day part focuses on people’s work/life schedules. One example: targeting ads for Egg McMuffins from 8-10 a.m. Day part targeting also works well for impending product releases and roadblock campaigns.
– Purchase-based: Tracks the purchase history of users to establish trends, much like behavioral. People who bought one brand’s shoes might be interested in more of the same from another brand.
– Retargeting: Aims to locate consumers who dropped off midway through the path to a purchase and serve them a new ad in the hopes they will complete the purchase. This is called remarketing in the offline world and is sometimes classified as part of behavioral targeting online.
– Semantic targeting: in the ad ecosystem, semantic technology occupies an increasingly important niche. Semantic technology improves the current method of “contextual targeting,” which is the practice of understanding page content as one factor in selecting an appropriate ad to be placed on the page. More subtle uses of this technique includes detection of the mood, tone or emotionality content is likely to evoke in the reader. For brand advertisers who desire to associate their brands with a feeling this is particularly effective.
Furthermore, there are several sub-classes of semantic targeting. One uses shallow linguistics to establish a base understanding of the meaning of words on a page, followed by various statistical manipulations to enhance this. The other sub-class of semantic targeting uses a fully linguistic approach. Here, there are no statistical means to establish understanding but rather the use of a rich set of rules and a semantic network (e.g. a database of the words, their meanings and the relationships between them). This method mimics with surprising accuracy how we are taught to read and comprehend as a child.
There has been a long debate about the merits of one semantics sub-class over the other. Those advocating a statistics approach do so by acknowledging the superior accuracy of a fully linguistic method but noting the longer time to process and limitations in scaling. With the advent of cloud computing, these limitations no longer exist and today it is possible implementing and using these technologies to solve these problems.