According to the Internet Advertising Bureau's 2009 Ad Revenue Report
online ad sales in the US were $22.7 billion dollars in 2009. There are a number of channels and platforms by which these ads are sold. The IAB breaks them down as follows:
- Search (47%)
- Display/Banners (23%)
- Classifieds (9%)
- Rich Media (7%)
- Lead Generation (6%)
- Digital Video (5%)
- Sponsorship (2%)
- E-mail (1%)
While search advertisments, typically text ads served on results pages of major search engines like Google's Ad Words, is clearly the leading dollar bringer, but sales of Display, Rich Media, Lead Generation and Digital Video ads collectively compromise 41% of the market. These visually rich ads usually appear in and around content on publisher sites. Obviously advertisers seek to place ads for their products in front of eyeballs inclined to be interested in their products. There are two broad strategies for doing this:
- Contextual Targeting - Contextual Targeting works under the premise that ads should be related to the content in which they appear. As an example ads for minivans may be appropriate for content such as: reviews for minivans, stories about soccer moms or stories relating to families with many multiple school age children. Contextual Targeting is stateless in that it cares nothing about what a user may have done previously, it makes its inferences about what type of ad should appear solely based on the content that the user is currently visiting. Cookies are not necessarily (see special cases below) needed for this type of targeting.
- Behavioral Targeting - Alternatively, Behavioral Targeting works under the theory that if a person (or more precisely a cookie) has previously read a review of minivans, but is not reading a story the war in Iraq, the previous behavior may be used to target the ad to pages, such as stories about Iraq, which have little contextual relevance to advertisers. It is important however to understand that there are many, many practices which have all been placed under the moniker of Behavioral Advertising. There are indeed a wide variety of mechanisms by which data can be captured about a consumer from which ad targeting can be achieved. Just by way of example these may include:
- Offline Data Append - where the cookie is linked to an identity in the real world and the identity is known to have certain characteristics (available from many data brokers). If for example it is known that John Quincy Adams is cookie uid=abc123 and John Quincy Adams is an attorney, male, age 45, married with one child and makes $218,000 a year then that cookie may be placed in related targeting areas. Though Behavioral Advertising often claims to be anonymous, it is doubtful that players doing this type of targeting (and not putting in place some form of anonymization techniques) could truly be considered anonymous.
- Content Pixeling - where one by 1x1 pixels which are graphic images too small to be seen on the page but which transfer data and cookies typically to 3rd parties can allow an ad network to know that one of its cookies has been a given page. By way of example, if an ad network were to place a 1x1 pixel on all pages within a newspaper site dealing with "automotive" they could acquire a list of cookies they may conclude are "auto interest". While questions may be raised as to if this is a personally identifiable technique, it can at least be acknowledged that it is not strictly necessary for the ad company to know the identity of a cookie, but rather simply that the cookie had indeed viewed given content.
- Content Pixeling and Social Graphing - similar to standard content pixeling, this technique also uses pixels on pages but with an innovative twist. This technique uses two data sources, conversion pages for popular products and social networking. By placing pixels in both locations interesting assumptions may be made. For instance if a pixel is placed on pages within social networks it is possible to ascertain who's friends with whom. For instance all cookies visting a given page on a social network are likely either friends with each other or at minimum share a common friend. If one of this group should purchase something, for instance a new Ipad, it is a strong bet that others within that group would be good candidates to market Ipad's to - even if the individual cookies themselves have never demonstrated a behavior which is specifically indicative of being in the market for an Ipad. Again it is not strictly necessary to know WHO these people are to achieve this targeting but merely to know that they share a common social group.
- DPI or Deep Packet Inspection - is a new form of gathering data for behavioral targeting where data is gathered directly from the ISP. Typically this would work by an ISP monitoring areas of consumer web traffic to see what sites, what queries and what the actual content of web pages consumers visit. Obviously this technique raises a number of questions with respect to interception, consumer notice and choice and scrutiny over what content should or should not be monitored. While this technique has been broadly criticized as being privacy invasive and clearly has the potential to do so, it should equally be acknowledged that this technique too does not necessitate the use of identifiable data as it seeks merely to create groups of interest areas and does not need the identity of the individuals themselves.
While there may be very different ways by which the data is initially collecting ranging from data brokers to Social Networking, the mechanism by which behavior is maintained is cookies in all cases. Without cookies, none of these methods works.
- Special Cases - Cookies play a role even in areas like Contextual Targeting where they are not the primary decision engine in determining which ad is seen by a user. Cookies are also used for what we may consider to be intra ad decisioning. For instance cookies are used both for ad sequencing and frequency limitation.
- Frequency Limitation Marketers and Publisher both understand well that value often diminishes with exposure. Cookies may be used to maximize return by not over exposing a single ad to the same user too many times. This is of potential value to both publishers and marketers. Marketers can avoid showing the same creative to the same user too many time (when they could trade some of those repeat exposures to 1st time exposures to new customers for the same cost). Publishers can avoid irritating their valued customers by e.g. exposing a pop up or an interstitial to the same user multiple times within the same day when a cookie could be used to prevent this.