ABSTRACT:
Exposure models focusing on reach and frequency where an audience is exposed to an ad message are the foundation of audience measurement. Given that a certain fundamental assumption of conventional exposure models is not relevant in the Internet advertising context, this paper suggests that the audience's on-line banner exposure can be best analyzed by its Web site visits. The negative binomial distribution (NBD) model, having long been applied in analyzing repeat behaviors, is proposed to serve as a banner ad exposure model. An empirical validation indicates that the model performs well in both data fitting and prediction.
(ProQuest Information and Learning: ... denotes formulae omitted.)
As market players take a more realistic view of the Internet as a medium for advertising nowadays, an array of metrics has been introduced specifically for Internet advertising (Bhat, Bevans, and Sengupta 2002). However, given these Internetspecific metrics, at least in the nonproprietary domain, there is a dearth of modeling work shedding new light on how Internet advertising can be planned more objectively and efficiently. In comparison with conventional media, the Internet opens windows of opportunity for advertisers to track an audience's behaviors that were unimaginable before, but the issue of reach and frequency of banner advertising-how the audience is exposed to an on-line campaign-has not been rigorously addressed.
In terms of advertising media models for planning (Rust 1986), prior research studies proposed that models developed for conventional media can be satisfactorily applied to the Internet advertising context (e.g., Leckenby and Hong 1998)-in this we do not agree. Since the Internet audience's possible exposure to a given advertising message is no longer limited to the upper bound defined by a schedule in the conventional sense, we argue that exposure models developed on the assumption of message "insertion" cannot serve to describe, not to mention predict, the actual exposure of an Internet audience to on-line advertising. As detailed clickstream user-centric and/or site-centric data are currently available to advertisers, it is neither a scarcity of data nor the accuracy of record that is at issue in efficient on-line ad planning, but rather a lack of a simple but robust model that relevantly guides the measurement of on-line audiences for the sake of advertising planning. This paper proposes such a model for bannerlike ads
We actually borrow the development of the model from other marketing research contexts in which the model's applicability as well as its robustness has been long and widely demonstrated. As a simple model for the description and prediction of the Internet audience's extent of exposure to on-line campaigns, the model is expected to serve as a stepping-stone to future modeling efforts in this respect.
In the following, the background of both Internet advertising and related advertising media models is first briefly sketched. The problem of conventional exposure in the online context is then discussed. Thereafter, the proposed model is introduced, for which performance, characteristics, use, and limitations are also empirically investigated.
BACKGROUND
Banner Ads on the Internet
Banners, pop-up ads, pop-under ads, interstitials, and so forth, are current variants of popular Internet advertising. Given their growing acceptance by agencies and advertisers, however, there is an array of problems that have not yet been satisfactorily resolved. Among them, the most frequently mentioned are (1) confusion concerning the effectiveness of bannerlike Internet advertising, (2) confusion concerning appropriate pricing models for bannerlike Internet advertising, and (3) confusion concerning appropriate advertising metrics (Bush, Bush, and Harris 1998; Chatterjee, Hoffman, and Novak 2003; Drèze and Hussherr 2003; McDonald 1997; Shen 2002).
These issues are closely interrelated. Since practitioners hold different views as to how Internet advertising can be most effectively applied to their integrated marketing communication efforts, various metrics and consequential pricing models naturally coexist in the marketplace. The confusion actually results from Internet advertising's dual promise of one-toone targeting and global reach (Drèze and Hussherr 2003) or, to put it another way, from the dichotomy of Web marketing goals: direct response versus branding (Broussard 2000).
At the early stage of Internet advertising, compared with conventional media for advertising, the Internet was taken as a very different medium. People were inclined to highlight the interactiveness of the Internet (e.g., Ephron 1997). The belief that Internet advertising is unique and that interaction on-line is what Internet advertising is for makes click-through, the most obvious behavior that represents interaction between the audience and the advertiser, a frequently talked-about metric in gauging performances of on-line campaigns. During the course of Internet advertising growth, this emphasis on interactionrelated behaviors has gained a deeply rooted popularity in practice. In a survey of media directors of interactive advertising agencies, Shen (2002) found that a majority of the respondents still use click-through to measure the effectiveness of on-line advertising. An interesting but rather confusing finding from Shen (2002) is that in terms of pricing, more than 90% of the respondents frequently used conventional CPM (cost per thousand) to price banner ads, whereas only 33% of the respondents have a click-through rate as one of their pricing models.