An embarrassment of data

In spite of the ceaseless business hype surrounding the Internet, it can be easy to understate the shift it marked in terms of the quantity and variety of data generated about audiences. As noted previously, traditional media operate with a relative paucity of “natural” information. Without conducting surveys, a TV or radio broadcaster has no direct indication of the number of people watching or listening. Print publishers have more natural information at their disposal. Newsstand‐driven publications record the number of copies of each issue distributed and returned, but have to estimate total readership. Subscription‐based periodicals also have basic demographic information about their readers. (Controlled‐circulation titles know much more, and use reader surveys and free subscriptions to actively shape an audience profile desirable to their advertisers.) Source: authors’ research On the Internet every action by a reader generates a data trail in a chain of computers running, at a minimum, from the Web site being visited, to the ISP, to the user’s own browser. Each of these three tiers has become the basis for a competing approach to audience measurement: Server‐based (or census‐based) analytics software runs on publisher servers; ISP‐based estimates collect traffic data from major ISPs such as Verizon and Time Warner; and panel‐based metrics use tracking software installed on the computers of a panel of Internet users. Ad servers and advertising networks introduce additional layers of data collection: a single page request by a reader may result in calls to the publisher’s in‐house ad server, to individual third‐party ad servers, and to an advertising network. Server activity at each of these layers can be aggregated and analyzed. More importantly, both content and advertising servers at each layer may introduce a “cookie” to identify the reader’s computer in the future as he or she returns to the current site, or visits other sites in the same editorial or advertising network. As a result, a single user’s actions may be simultaneously tracked by multiple categories of observers. Over time, then, the number of parties who can produce meaningful information about online audiences and audience behavior has increased. The vocabulary of audience‐related statistics has increased with it: To note only a few of the most basic metrics, publishers and advertisers today must be conversant in “page views,” “click‐throughs,” “unique visitors,” “usage intensity,” “engagement time,” and “interaction rates,” in addition to the demographic and behavioral profiles of their audience.