The challenges of counting heads to measure your online presence
By Jay Krall
Web metrics have long been crucial to proving return on investment for PR professionals who score digital ink with top Web sites and blogs. While new media metrics for measuring consumption of video, discussion in social networks and other multimedia content are coming to fruition, Unique Visitors per Month remains the gold standard for measuring a site’s reach. But what goes into calculating UV figures? And if one of your clients questions the accuracy of a particular number, how can you respond? The way these numbers are tracked is quickly evolving and will continue to do so in 2009.
Regardless of whether you get your Web metrics from a subscription-based provider like Nielsen NetRatings (whose data is available in CisionPoint), or you rely on a free service such as Quantcast or Compete, you should understand that there are two basic ways to arrive at UV figures.
● Panel-based research. All major providers of Web audience data include in their formula some form of panel-based research, which consists of monitoring the Web usage of a sample group of people, and then extrapolating the total audience size for a particular site based on the proportion of people in the group who visit that site. Here’s a simple illustration: if a particular panel finds that 2 in 100,000 people visit a site, and let’s say there are 200 million regular Internet users in America (data providers conduct surveys to determine the true number), that suggests that 4,000 people visit the site each month. When recruiting users for a panel, providers strive to ensure that their panels are representative of the general population with regard to demographics like age and gender, as well as mixing casual Internet users with always-wired, hard-core Web geeks. Panel-based research serves as the basis for most Web audience measurement.
● Web analytics. A wide variety of services offer Web traffic measurement and analysis services to Web publishers. Generally, this data comes from Web server logs. A caveat: server logs sometimes offer inflated numbers because they count Web spiders and bots. They also can double-count people who regularly clear cookies from their browser or access a site from multiple computers. This explains why figures offered by sites themselves tend to be higher than those that come from panel-based data providers.
Obviously, depending on the type of method used, audience figures from different sources can vary greatly for the same site. Even within the realm of panel-based research, methods vary, and mixing multiple methods has become common. ComScore calculates its audience metrics based on a panel of more than 2 million people, mostly recruited online. Quantcast incorporates Web analytics with panel data in a blended approach. NetRatings’ NetView product offers data from a panel of about 30,000 people selected through Random Digit Dialing, a standard phone recruitment method that is more costly and time consuming than online recruitment, but proponents say offers a better microcosm of all Web users than those recruited online. (comScore also includes some panelists recruited through RDD.)
The study of Web traffic for media relations measurement purposes now faces new challenges because of the niche nature of new media outlets like blogs. Many small blogs and sites that cater to no more than a couple of thousand people who are keenly interested in a particular topic have become desirable targets for PR professionals, who can be sure that coverage in these outlets is reaching the right people. But if a particular panel doesn’t happen to include anyone who is interested in a niche topic such as, say, travel for people with disabilities, the panel may not be able to produce any data at all for a site exclusively covering that topic.
That means that as some Internet media outlets become more narrowly focused, research firms are coming up with new methods to measure our habits. “Where we’re moving, and this is going to happen in 2009 in the U.S., is a methodology we call RDD Online, where you take this RDD panel and you supplement it with online recruitment, which is less expensive, because those costs get passed on to our customers,” says Scott Ross, director of product marketing for Nielsen Online. Adding online recruitment will increase the number of smaller sites for which Nielsen can provide data, he said. The company plans to place more weight on data from users recruited through the RDD method than those found online. This is aimed to account for the fact that people recruited online tend to be heavier users, and can skew the data. “The behavioral representation, the percentage of light users versus heavy users, is really essential,” Ross says.