Modeling the Clickstream: Implications for Web-Based Advertising Efforts

Advertising revenues have become a critical element in the business plans of most commercial Web sites. Despite extensive research on advertising in traditional media, managers and researchers face considerable uncertainty about its role in the online environment. The benefits offered by the medium notwithstanding, the lack of models to measure and predict advertising performance is a major deterrent to acceptance of the Web by mainstream advertisers. Presently, existing media models based on aggregate vehicle and advertising exposure are being adapted which underutilize the unique characteristics of the medium. What is required are methods that measure how consumers interact with advertising stimuli in ad-supported Web sites, going beyond mere counts of individual "eyeballs" attending to media. Given the enormous potential of this dynamic medium, academic investigation is sorely needed before cost implications and skepticism endanger the ability of the medium to generate and maintain advertiser support. This paper addresses advertiser and publisher need to understand and predict how consumers interact with advertising stimuli placed at Web sites. We do so by developing a framework to formally model the commercial “clickstream” at an advertiser supported Web site with mandatory visitor registration. Consumers visiting such Web sites are conceptualized as deriving utility from navigating through editorial and advertising content subject to time constraints. The clickstream represents a new source of consumer response data detailing the content and banner ads that consumers click on during the online navigation process. To our knowledge, this paper is the first to model the clickstream from an actual commercial Web site. Clickstream data allow us to investigate how consumers respond to advertising over time at an individual level. Such modeling is not possible in broadcast media because the data do not exist. Our results contrast dramatically from those typically found in traditional broadcast media. First, 3 the effect of repeated exposures to banner ads is U-shaped. This is in contrast with the inverted Ushaped response found in broadcast media. Second, the differential effects of each successive ad exposure is initially negative, but non-linear, and becomes positive later at higher levels of passive ad exposures. Third, the negative response to repeated banner ad exposures increases for consumers who visit the site more frequently. Fourth, in contrast to findings in traditional media the effect of exposure to competing ads is either insignificant or positive. However, carryover effects of past advertising exposures are similar to those proposed in broadcast media. Finally, heterogeneity in cumulative effects of advertising exposure and involvement across consumers captured by cumulative click behavior across visits and click behavior during the visit was found to significantly contribute to differences in click response. This has implications for dynamic ad placement based on past history of consumer exposure and interaction with advertising at the Web site. Response parameters of consumer visitor segments can be used by media buyers and advertisers to understand the benefits consumers seek from the advertising vehicle, and thus guide advertising media placement decisions. In sum, our modeling effort offers an important first look into how advertisers can use the Web medium to maximize their desired advertising outcomes.