The Journal of Advertising Research 50th Anniversary Edition

The Internet has outpaced every other medium in terms of the speed of consumer adoption. At this writing, more than 1.3 billion people worldwide age 16 or older have online access via fixed ISPs or mobile devices1. And they use the Internet regularly for communication, entertainment, commerce, and as a timely source of information about a broad variety of issues. In fact, online activities in the U.S. now account for about 25 percent of consumers’ media consumption time2.

With such extensive consumer-based online activity, it’s not surprising that marketers are using the medium aggressively as both a sales and marketing channel. This activity also has driven the need for researchers to provide marketers with metrics on how consumers use the Internet. At the same time, the market-research industry has adopted the use of online surveys as a fast, inexpensive way to obtain consumer insights that can be used by marketers in both the online and offline worlds.

The emergence of online research has not been without controversy. Described as “the most measurable medium ever created,” the Internet initially was believed by many to be easily and accurately measured using website servers that recorded every visit to a site or every ad delivered. By placing a cookie (a small piece of computer code) on each visiting computer, it was thought that the number of unique cookies would reveal the number of visitors to any website.

It was only when the market-research industry used using time-tested approaches (continuously tracked behavioral panels, for instance) to show that cookies were being deleted by about 30 percent of all Internet users - and that these so called “deleters” were doing so about four times per month - that it became apparent that counting cookies had led to dramatic overstatements of the actual number of unique visitors to websites. Similarly, it was concluded that using ad servers and placing a cookie every time an ad impression was delivered to a computer in an attempt to measure the reach and frequency of an ad campaign simply led to an overstatement of the reach and an understatement of the frequency actually delivered.

At about the same time, market researchers using behavioral panels were showing that online advertising was not just about direct response and that the click on a display ad (or lack thereof) ignored the latent branding impact of advertising on subsequent brand choice. Put another way, consumers can respond positively to online ads even without clicking on them. This was especially surprising to the many new Internet advertising practitioners who had no previous experience with advertising research. This finding, however, was less surprising to market researchers who had toiled for years in the advertising arena prior to the emergence of the Internet and who understood the many different and complex ways in which advertising can affect consumer behavior.

All of this is not to say that web servers don’t have a place in online market research. Just that one needs to carefully consider what the servers are actually counting. They do indeed capture all of the visits to a website and all of the ad impressions that are delivered. But they also capture “bot traffic” (i.e. computer-driven) site traffic and non-requested pages (sometimes called “push traffic”) which must be filtered out of any visitation metrics, and they are unable to determine which specific individual is using a computer at any point in time. For example, if a single person visits a website from his/her work computer and also from his/her home computer, the website server will count this as two unique visitors, whereas in reality it is one unique person visiting the website twice. Or, if two different people in the same household visit the same website at different times using the same computer, the website server will count this as two visits by one unique visitor whereas in reality it’s two unique visitors. These are obviously important distinctions that any marketer or researcher needs to understand.

Interestingly, leading edge research today is increasingly centered on the use of integrated website server data and continuously measured behavioral panels. By leveraging the strengths of each database and taking into account their respective weaknesses, it’s possible to provide faster, more accurate and more granular insights into online consumer behavior. This is particularly important as the research industry seeks to understand how marketers can best utilize the Internet in a world where social networks are changing the manner in which consumers seek out, receive and use information that affects their brand choice.

On the survey side, substantial research has been conducted by individual survey research companies as well as the ARF to demonstrate that online panels - even if not built according to strict probability sampling rules - can be used to provide fast, accurate and inexpensive insights into consumer attitudes, sentiment and even behavior. Today, research is ongoing to determine the limitations of online survey panels and the situations or applications where probability-based samples are still required.

To a certain degree, what we’ve observed with the emergence of Internet research is the reality that, when attempting to measure a computer-driven medium, one needs to carefully consider what metrics the computer is counting and providing. As Albert Einstein famously said some time ago: “Not everything that can be measured matters.” Ultimately, marketers need to understand the behavior of people and not just computers.

1 Comscore, 2010
2 Research conducted by Knowledge Networks for TVB