Description
Commercial advertisers often seek to generate increased consumption of their products or services through "branding," which involves associating a product name or image with certain qualities in the minds of consumers.
Attention to Advertising
ATTENTION TO ADVERTISING
AKADEMISK AVHANDLING
vid Samhällsvetenskapliga fakulteten Handelshögskolan i Umeå UMEÅ UNIVERSITET
Carl Patrik Nilsson
Framlägges för vinnande av ?loso?e doktorsexamen till offentlig granskning vid Handelshögskolan vid Umeå Universitet, Hörsal G, Humanisthuset, Fredagen den 10 november, 2006, kl. 13:15.
Attention to Advertising
Carl Patrik Nilsson, Umeå School of Business, Umeå University SE-901 87 Umeå,Sweden
Abstract
In advertising, capturing consumers’ attention is imperative. Yet, how attention to advertising works in a World Wide Web context has not been researched extensively. This area is the main focus of this research. The World Wide Web is an interactive medium which has characteristics that deviate from traditional media channels. Considering the unique properties of the World Wide Web it was deemed important to study how attention to advertising is affected by various contexts as well as attention getting techniques. The study was conducted in order to develop an understanding of how advertisements in different Web task environments are being attended to as well as how attention to advertisements varies between different attention getting techniques that are being used in Web advertising. Furthermore, a model which describes the relationship between context, attention getting techniques and attention to advertising was developed and tested. In connection to this it was tested whether click-through or click-rate is indeed an adequate method when measuring advertising effect and thereby to what extent the pay per click pricing strategy is appropriate. In order to address the research issues a theoretical framework was assembled. Nine hypotheses were formulated deriving from the framework. To test the hypotheses and the model, an experimental research design was employed. Three experiments were designed to study the hypotheses formulated. 702 individual experiments were conducted. The conclusions from the study reveal that Web task environments indeed have an impact on attention to advertising. Increased complexity or a more dif?cult task environment demands more cognitive resources which in turn produces lower attention levels to the advertising stimulus. Reduced complexity, on the other hand, releases cognitive resources that can be spent elsewhere. Thus, a higher attention level to the advertising stimulus was recorded. This also con?rmed the properties of the proposed model; namely that the permeability of a ?ltering mechanism is dependent on for instance level of goal orientation and complexity of the physical task environment i.e. the complexity of the Web site. The study revealed that attention getting techniques such as pop-up advertisements increase the attention to advertising. However, it was found that pop-up advertisements are effective, not mainly because of their abrupt presentation, but because of the distinct properties of the frame. It was found that the frame has a negative meaning for Web users and when the frame comes into the visual ?eld it will attract their attention. At the same time attention will also be distributed towards the advertisement itself. This is recorded as an increase in attention towards the advertising message. The results show that the click-through measure is not an appropriate method when measuring advertising effect. The click-through measure may severely underestimate the advertising effect. Consequently, using click-through frequencies as a basis for pricing is inappropriate, unless the effects at play have been well understood and taken into consideration. Key words: Attention to advertising, attention getting techniques, Web advertising, Web environment, click-through Studies in Business Administration, B series, No 60. ISSN 0346-8291, ISBN 91-7264-189-4 Umeå School of Business - Sweden 2006
Attention to Advertising Studies in Business Administration, B series, No 60, 2006 ISSN 0346-8291 ISBN 91-7264-189-4 © Patrik Nilsson 2006 Stockholm 2006-10-10 Font: Sabon & Avenir Cover Design: Björn Hollström, Honeymoon Sthlm AB Graphic Design: Robin Bergström, Honeymoon Sthlm AB Printed in Stockholm, Sweden by Tryckbolaget All rights reserved. Except for the quotation of short passages for the purposes of criticism and review. No part of this publication may be reproduced, stored in retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior permission of the publisher. This material is protected under the Swedish copyright act (1960:729) and international equivalents.
ACKNOWLEDGEMENTS
This doctoral dissertation is the end product of a long voyage. It has been a privilege to cooperate with a number of organizations, colleagues, friends and loved ones. I would like to express my gratitude for their contribution and support. I wish to thank my main tutor, Professor George Tesar. I highly appreciate your commitment and help and that you set aside time to give feedback, whenever needed. I will, for instance, always remember our last early morning meeting at Arlanda Airport where you made time to discuss the thesis between ?ights. It was an inspiring meeting. I also wish to thank my tutor, Professor Anders Söderholm who has motivated and believed in my study. This motivation was particularly important towards the end of the project when there were still loose ends to tie together. My gratitude also goes to my third tutor, associate Professor Agneta Marell, who also served as opponent at the internal seminar. Together with two additional opponents, Vladimir Vanyushyn and Timo Mäntylä, insightful comments and feedback were given that have helped me improve my work. Timo was also very helpful at earlier stages of the project and contributed with important ideas and comments. Thanks to all of you. I am also indebted to my friend Thomas Laitila who has been available for meetings, discussed various topics and given fast feedback. Thanks for your very kind help and that I was able to disturb you every now and then, especially when I was working on the ?nal analysis. Furthermore I would like to thank Owe R. Hedström who inspired me to come back to Umeå and conduct research. I have never had a reason to regret my decision. Owe, thanks for your personal support and for being such a great colleague and friend. Even though I have designed all the experiments myself it is handy to get the help of specialists every now and then, such as system managers, art directors, designers and programmers. Therefore; Björn Hollström, Michael Lee Bjerkö, Johan Jönsson, Fredrik Lindgren, Mattias Johnson, Johan Tufberg and Magnus Lindgren, thank you very much for helping out with various parts of the experiments. Thanks also to all the respondents. Without your participation not much would have been accomplished.
In the following is a list of those who have, in various ways, contributed to the ?nalization of this project. Please accept my sincere gratitude. Roland Carlsson, Umeå Kommuns Näringslivsservice Barbara Cornelius, Professor, Umeå School of Business - USBE Owe R. Hedström, Director of Studies, Umeå School of Business - USBE Anders Hellquist, System Manager, Umeå School of Business - USBE Ola Henfridsson, Associate Professor, Viktoria Institute Björn Hollström, CEO, Honeymoon Sthlm AB Linus Holm, Ph.D student, dept of psychology, Umeå University Jonny Holmström, Professor, dept of informatics, Umeå University Mattias Johnson, System Manager, Umeå School of Business - USBE Johan Jönsson, Mathematician and ICT specialist Thomas Laitila, Professor, Örebro University and SCB Michael Lee Bjerkö, Design Manager, Honeymoon Sthlm AB Per Levén, Assistant Professor, UCIT and dept of informatics, Umeå University Henrik Linderoth, Assistant Professor, Umeå School of Business - USBE Fredrik Lindgren, IT Specialist, Volvo Corporation Magnus Lindgren, Programmer - IT Specialist, Teknikhuset Agneta Marell, Associate Professor, Umeå School of Business - USBE Timo Mäntylä, Professor, dept of psychology, Umeå Univeristy G.M. Naidu, Professor of Marketing/Professor Emeritus, University of Wisconsin - Whitewater Karl-Gösta Nilsson, former U.N. envoy and former Lecturer in marketing Katarina Pousette, Executive assistant, Umeå School of Business - USBE Claudia A. Rademaker, Project Manager, Berghs School of Communication Anders Söderholm, Professor, Umeå School of Business - USBE George Tesar, Professor, Umeå School of Business - USBE Johan Tufberg, Webmaster, Umeå School of Business - USBE Vladimir Vanyushyn, Ph.D student, Umeå School of Business - USBE Rickard Wahlund, Professor, Stockholm School of Economics Thanks to my family and in particular to my father, Karl-Gösta Nilsson, who has always inspired and encouraged me to pursue my goals. Last but not least, my gratitude goes to my wife Claudia and our daughter Raquel. Your support has been invaluable. Claudia, thanks for reading innumerable versions of the manuscript and for coming with input at all times. Stockholm, October 2006 Carl Patrik Nilsson
In memory of Frank Rademaker
1941 - 2006
Contents
Chapter One Attention to Advertising .......................................................................1
Introduction .................................................................................................2 Why study Web advertising ..........................................................................3 A background and knowledge platform .......................................................4 Placing Web advertising in a broader context ...............................................5 The Internet, the World Wide Web and contexts ..........................................6 Advertising and attention getting on the WWW .........................................11 Modeling attention to Web advertising ......................................................14 Measuring advertising on the Web and pricing methods ............................15 Outline of the dissertation ..........................................................................19
Chapter Two Theoretical Framework.......................................................................20
The origin of advertising ............................................................................21 What is advertising and what is a Web advertisement?. ..............................22 The Web is growing important ...................................................................24 Advertising is taking place in a setting........................................................25 Outcomes of advertising.............................................................................28 Putting the parts together ...........................................................................31 Central concepts in advertising...................................................................32 Models describing advertising ....................................................................34 The World Wide Web as an advertising vehicle ..........................................42 Web advertisements are placed on the World Wide Web ............................43 The advertisement. .....................................................................................46 The Web dimension of advertisements .......................................................49 The concept of attention. ...........................................................................54 How the attention system operates.............................................................58 Advertising, attention and media................................................................63 Measuring cdvertising - click-through and pricing......................................66 This chapter ...............................................................................................69
Chapter Three Formulating Hypotheses .....................................................................70
First section - attention effects of context ...................................................71 Second section - advertisements and their respective attention effect ..........72 Second section extended .............................................................................74 Modelling attention to Web advertising .....................................................77 Third section – measurement and pricing ...................................................78 Other effects regarding Web advertising .....................................................79 Summary ....................................................................................................80
Chapter Four Research Method ..................................................................................81
Approach ...................................................................................................82 The essence of an experiment .....................................................................83 Overall design ............................................................................................87 The design of the experiments ....................................................................88 The main measures used.............................................................................92 Selection .....................................................................................................93 The experimental setting – the lab. .............................................................95 Instructions and procedures .......................................................................96 The survey. .................................................................................................97 The “exit poll”. ..........................................................................................97 Validity and reliability ................................................................................98 Analysis....................................................................................................103 Logistic regression analysis.......................................................................103 Analysis of variance - ANOVA .................................................................105 Operationalization of hypotheses .............................................................106
Chapter Five Results and Analysis ..........................................................................108
First section ..............................................................................................109 Second section ..........................................................................................111 Impact of external context factors on attention ........................................113 Third section. ...........................................................................................114 Extending hypothesis 5 ............................................................................120 Modeling attention to Web advertising ....................................................125 Fourth section. .........................................................................................126 Collateral effects of Web advertising ........................................................129 Control ....................................................................................................132
Chapter Six Conclusions and Further Research ................................................137
Attention to advertising on the Web. ........................................................138 Attention to advertising differs between attention getting techniques. ......140 A model of attention to advertising on the Web .......................................142
Collateral effects of attention getting........................................................144 Measuring advertising on the Web ...........................................................146 Pricing methods on the Web .....................................................................148 Future research.........................................................................................150
Svensk Sammanfattning ....................................................................152 References.............................................................................................154 Appendices ...........................................................................................171 Figures
Figure 1.1 .....................................................................................................5 Figure 1.2 .....................................................................................................8 Figure 1.3 ...................................................................................................19 Figure 2.1 ...................................................................................................23 Figure 2.2 ...................................................................................................24 Figure 2.3 ...................................................................................................28 Figure 2.4 ...................................................................................................33 Figure 2.5 ...................................................................................................34 Figure 2.6a .................................................................................................36 Figure 2.6b .................................................................................................36 Figure 2.7 ...................................................................................................38 Figure 2.8 ...................................................................................................41 Figure 2.9 ...................................................................................................47 Figure 2.10 .................................................................................................50 Figure 2.11 .................................................................................................59 Figure 2.12 .................................................................................................60 Figure 2.13 .................................................................................................61 Figure 2.14 .................................................................................................64 Figure 2.15 .................................................................................................65 Figure 3.1 ...................................................................................................78 Figure 4.1 ...................................................................................................87 Figure 4.2 ...................................................................................................88 Figure 4.3 ...................................................................................................97 Figure 5.1 .................................................................................................126 Figure 6.1 .................................................................................................139 Figure 6.2 .................................................................................................143
Tables
Table 5.1 ..................................................................................................109 Table 5.2 ..................................................................................................110 Table 5.3 ..................................................................................................112 Table 5.4 ..................................................................................................114
Table 5.5 ..................................................................................................115 Table 5.6 ..................................................................................................117 Table 5.7 ..................................................................................................118 Table 5.8 ..................................................................................................120 Table 5.9 ..................................................................................................122 Table 5.10 ................................................................................................123 Table 5.11 ................................................................................................127 Table 5.12a ..............................................................................................128 Table 5.12b ..............................................................................................128 Table 5.13 ................................................................................................130 Table 5.14 ................................................................................................131 Table 5.15 ................................................................................................133 Table 5.16 ................................................................................................134 Table 5.17 ................................................................................................135 Table 5.18 ................................................................................................135
Chapter One
Attention to Advertising
This ?rst introductory chapter will describe the area to be researched. Concepts and theoretical areas key to the study will be elaborated. Subsequently, the purpose of this study will be pointed out in greater detail. At the end of this chapter, an outline of this dissertation will be presented.
1
Introduction
A number of media channels exist that can be used for conveying advertising messages. To the well-established channels such as TV, radio and print, the World Wide Web has been added. As much as there are differences between, for instance, print and TV there are reasons to believe that the World Wide Web has unique or special features that differentiates it from other media channels. The World Wide Web has characteristics that deviate from traditional media channels, making the World Wide Web a prime target for investigation. One signi?cant disparity, which is often pointed out, is that the World Wide Web represents an interactive media with a greater richness of information as opposed to what can be found in traditional media (Evans & Wurster, 1997). Researchers have found that consumers are more active on the Internet and the World Wide Web than is the case with other media channels (Hoffman and Novak, 1996). During a Web session, the Web user is active and interacting with the material presented on the Website, a behavior that may impact advertising on the Web. A main property distinguishing Web advertising from other kinds of advertising is that it has hybrid characteristics combining properties from print, broadcast, out-door, and direct response media (Dréze and Zufryden, 1997). In addition, Karson and Korgaonkar (2001) are pointing out no less than eight unique characteristics of which four1 will be mentioned here. As mentioned earlier, one of these characteristics is Interactivity, a second one is Complexity - Web sites exhibit a greater complexity than both print and broadcast media (Bruner & Kumar, 2000), a third is Flexibility – this type of media is very ?exible, a property making room for creative advertising. Web advertising can range from static to full motion audio-video, which cannot be accomplished in print advertising. A fourth characteristic is Usage - unlike broadcast media, the Web is being used in a number of ways such as a channel of distribution of digital products, creation of Web communities (Karson and Korgaonkar, 2001), for Web logging, Web shopping and Web chatting to mention a few. These four characteristics differentiate the Web environment context from the context dimension in print and other media channels. When summarizing the above, it is reasonable to state that the World Wide Web is distinctly different from other types of media channels with regard to the environment in which the advertisement is being placed. Properties such as greater interactivity, complexity, usage and ?exibility create this difference. Furthermore, the advertisements on the World Wide Web can be propelled to stimulate and attract attention in an additional and different way than what is the case in, for instance, print media. Eliciting or attracting attention can
1
The other four that are not mentioned here are Variability, Accessibility, Salience and Stand alone. These properties are not included here since they do not have the same signi?cance for the uniqueness of the World Wide Web today as they may have had in the past. 2
be accomplished through various kinds of strategies and techniques such as animation, ?ash, size changes or pop-up. Taken together there is an array of unique features whose impact on advertising on the World Wide Web is yet to be investigated and it is this knowledge gap that is the very target of the research at hand. Consequently, the over arching research problem is how attention to advertising is affected by various contexts as well as attention getting techniques.
Why study Web advertising
Studying how the context impacts advertising has been conducted previously (see for instance; Bruner and Kumar, 2000, Moore, Stammerjohan and Coulter, 2005, Dahlén, 2005, Janssens and De Pelsmacker, 2005) but never, to the best of the author’s knowledge, where structural differences between Web environments have been compared. Appel (2000) reviewed eight studies of the editorial environment in print in the United States of which none gave strong support to the environment hypothesis. According to Appel a number of investigators have studied the issue but none has “demonstrated the validity of the concept … to the point where it is generally accepted” (Appel, 2000, p.94). Even though Appel’s statement is somewhat discouraging, it was at the same time challenging and inspired to take on the task in a Web context to seek a convincing answer to the issue. Since the World Wide Web differs from traditional media channels with respect to its interactivity, ?exibility, complexity and usage it was considered plausible to expect greater differences between environments on the Web. Attention getting strategies and techniques has been studied earlier for print and broadcast media. At the time when this research project was initiated, in 2002, there were no known studies of this subject in Web advertising. As late as 2004 Diao and Sundar (2004) wrote that there was little or no empirical research that had been conducted in the area. Recently, studies have been done where attention getting in Web advertising has been investigated. In the Eyetrack III study (2004), attention getting techniques was studied. In this study, different backgrounds have been used when various advertisements were tested making it hard to speci?cally attribute the effect to either the difference in context or the difference in attention getting technique. Similarly, when comparing for example the pop-up effect with animation, advertisements of different sizes were used making it impossible to know whether the registered effect originated from a difference in size or a difference in technique. Consequently, there is still a need to scienti?cally investigate the scope of the special attention getting techniques that Web advertising offers. In an editorial by Kover (2000) he expressed concerns of how to measure advertising on the World Wide Web. His concern may have been related to the Web industry’s practice to use click-through as a main measure of advertising effectiveness. Even though this practice is in con?ict with how advertising is being measured in other media channels than the Web, the click-through
3
measure is the most widely used method (Chandon, Chtourou and Fortin, 2003). Given this discrepancy and the signi?cance of employing correct measurement methods it was deemed important to study this issue in this research. In connection to the click-through measure, a pricing strategy based on this measure has evolved namely, the pay per click pricing strategy. The adequacy of this strategy can be challenged, especially when taking the following logic into consideration; if the click-through measure is inappropriate when measuring advertising on the Web then it may be inappropriate as a basis for pricing as well. Therefore, this research issue was incorporated into the study.
A background and knowledge platform
To be able to comprehend the relationship between various phenomena in Web advertising, it is advisable to obtain a brief overview of the Web advertising area. Hence, a knowledge platform is presented in the following. Advertising on the World Wide Web has developed rapidly over the last decade. Initially, advertisements were designed as simple one colored banners for Web users to click on. The reason for why banners were uncomplicated was because the limited bandwidth of the Internet did not allow the use of sophisticated graphics. More elaborate graphics affect the size of an advertisement which in turn affects the amount of information that has to be transferred to an individual user who is sur?ng a Web page. With the modems that were frequently used just a few years ago it took excessively long time to download a Web page if the graphics were very sophisticated. As a result, most graphics on Web pages, including advertisements, were of a basic character. However, as the infrastructure of the Internet has been upgraded with a greater bandwidth, advertisers have re?ned the graphical design (Zufryden and Dréze, 1997) leading up to today’s full ?edged Web advertisements. The old Web banners typically carried a very short or incomplete message that often ended with the two words “click here” or “click now” that was blinking to call upon the attention of the Web surfer. The purpose with these banners was to generate click-through to a target Web site where the actual good or service was promoted in greater detail. The Web advertisements of today, contrary to the early banners, often have some basic properties such as an image or description of a product, a brand name, copy text and a headline put together into an advertisement entity – the Web advertisement. Modern Web advertisements are quite different compared to their older counterparts. They have a different visual appearance and also a different purpose since they are communicating a complete message rather than aiming at generating click-through to a target Website. These re?ned Web advertisements in particular, that are quite similar to print advertisements, are to be researched in this dissertation.
4
Placing Web Advertising in a broader context
Marketing communication is an important part of marketing with the purpose to communicate with companies’ target markets2. Marketing communication can be described as representing “the voice of a brand and the means by which companies can establish a dialogue with consumers concerning their product offerings” (Keller, 2001, p. 823). The reasons for why companies communicate to their markets vary although often the purpose is to bring to customers’ attention information regarding new goods and services, to change attitudes toward a product or to remind consumers about products (Keller, 2001). The companies’ communication activities can furthermore be part of an information exchange between company and customer with the intention to maintain and develop customer relationships (Reid, Luxton & Mavondo, 2005). Regardless of the purpose, communication is an integral part of the marketing function and is executed by staff and functions within the organization and/or together with external organizations specialized in the area of communication. Marketing communication, frequently called promotion and being one of the four P’s3, is a broad concept that covers the sub areas of advertising, personal selling, public relations, sales promotion, direct marketing and areas connected to these (Duncan & Moriarty, 1998). The various areas are serving different purposes and aiming at achieving different goals. For instance, sales promotion is short term incentives to stimulate purchase whereas advertising can be used to build a brand image or to create a presence on a new market. Common for these sub areas is that they all include an element of communication that is directed towards the market. Communicating with the market and the consumers is important since it over-bridges the gap between companies and customers.
Marketing
Marketing Figure 1.1
Other sub areas in marketing
Marketing communication
Advertising
Personal selling
Sales promotion
Public Relations
Web Advertising Figure 1.1. Web advertising as a sub function of Marketing Communication and Advertising. Adapted from Duncan & Moriarty (1998, p. 9).
2
The communication aspect of marketing is explicitly formulated in the AMA de?nition of marketing, “Marketing is an organizational function and a set of processes for creating, communicating and delivering value to customers and for managing customer relationships in ways that bene?t the organization and its stakeholders.” (The AMA de?nition as reported by Keefe, 2004, p. 17 – 18) 3 Jerome McCarthy was the one who introduced the 4p concept into marketing in 1960 (Harvey, Lusch and Cavarkapa, 1996). 5
At a basic level, communication will make it possible to inform and make potential customers aware of a company’s business, its offering or to persuade customers to enter into an exchange relationship. One way to communicate with customers is through advertising in digital media channels such as the World Wide Web. This type of advertising, Web advertising, is an additional way to conduct marketing communication. In ?gure 1.1 above, it is clari?ed how Web advertising is related to marketing communication and marketing. Web advertising has not always existed. It is only in recent years that novel means to conduct marketing communication, such as Web advertising, have evolved. Technology development, and in particular the development of Internet technology and the Internet, has provided an infrastructure for marketers which can be used when communicating with customers. Despite the Internet’s relative youth, organizations have been fast with experimenting and adopting Internet related marketing communication. It has proved to be an ef?cient way to reach out to the market, or to, literally, reach into the customers’ homes or of?ces. For practitioners as well as researchers the development has meant an expansion of the current advertising ?eld.
The Internet, the World Wide Web and Contexts
The Internet4 is a computer based network that can be used for a variety of activities. Its inherent versatile nature makes it a formidable tool for communication and information transfer5 while it also sparked an interest for using it in a number of ways among companies6 and consumers. This interest propelled the growth of the Internet throughout the nineties and the early years of the twenty ?rst century. The diffusion rate of the Internet has surpassed any other media achieving global prominence in barely ?ve years (Sterne, 1997). Berthon, Pitt and Watson are arguing in the same way and they are highlighting the importance of the Web and also the swiftness of the
4 5
For a brief description of the Internet see Appendix A
The Internet can be used for many different kinds of communication and information transfer using applications such as: email, internet relay chat, ftp, telnet, blog, IP-phone communication, news groups, Web and many others. The Internet provides an arena for all kinds of communication. 6 As an explanatory note consider the following; From a company perspective it is recognized that the Internet can be used for various business activities such as logistics, buying, retailing, production, advertising, and the like. Companies have explored the Internet and its possibilities trying to exploit its properties to their bene?t. In this pursue managers saw great possibilities such as low costs for attracting customers, severely reduced costs for serving another customer, small companies competing on “equal” terms as larger companies, economies of scale, low production costs, low costs for distribution, a large scale network and infrastructure ready to use for any company and so on. All these potential bene?ts together with the business climate at the time (the late nineties) called for companies to take action. Companies just had to go “virtual” in order to defend positions vis-à-vis competitors and to be a part of what was about to come or simply, as Rossiter (1999, p. 1) formulates it, “…because it seems everyone else is doing it…” To the detriment of many companies and despite their efforts they often did not achieve the anticipated success. Recent research has presented explanations to this from different angels and most of all pointed at an unrealistic expectation on what IT and the Internet can do for organizations. The gained knowledge from the early days of the commercial Internet revealed that there are not only possibilities with the Internet but also obstacles and pitfalls that has to be understood if a company is to use it as an ef?cient marketing tool. 6
diffusion of the Web, “The Web is not a transient phenomenon. It warrants serious attention by advertising academics and practitioners. Statistics support this, although one astute observer recommends strongly that all estimates be made in pencil only, as the growth is so rapid. No communication medium or electronic technology, not even fax or personal computers, has ever grown as quickly.” (1996, p. 44). As a comparison, it took Radio some thirty years and Television some ?fteen years to become a wide spread media (Sterne, 1997). A great diffusion of the media also contributes to making it more ef?cient, from an infrastructural point of view, to the bene?t of the users (Rogers, 1995). Today the World Wide Web is part of everyday life and for many people the Web has become one of the major media for consumption (Morris, Woo & Cho, 2003). The Web is not only being used by consumers, it is increasingly being adopted by marketers as well (Prasad, Ramamurthy and G.M. Naidu, 2001). That the Web has grown important for both consumers and companies is undisputed. When discussing around the Internet and the World Wide Web it is important to differentiate between the two. What millions of consumers and companies world wide are using is in fact applications or services that are available on the Internet, such as email, newsgroups or the World Wide Web. This means two things; people are indeed using the Internet but only its infrastructure, secondly and also more importantly, it is the communication bene?t of the applications and services that has attracted the attention of companies and consumers and of most importance to this dissertation is the service that usually is called the World Wide Web, or just the Web. It is in fact the success of these services that has driven the growth of the Internet and in particular email and the WWW (see Appendix A). With the widespread use of the Web in society the phenomena has come to attract the interest of researchers. Even though the ?rst advertisement7 appeared on the Web (in the on-line magazine “Hotwired”) only a decade ago there have been a number of studies conducted that have explored various aspects of the Web. Research effort has been directed at comparing marketing activities on the World Wide Web with activities in the real world (see for instance Gallagher, Foster & Parsons, 2001) suggesting both that there are differences as well as similarities. Initial concerns about the fact that traditional principles of mass media advertising would not apply on the Web was disproved. Gallagher et al found that Web advertising was at least as effective as its print counterpart. They also found that some promotional material that was not classi?ed as advertising was less effective when presented on the Web. Gallagher, Foster and Parson’s research was undertaken in order to get an understanding for Internet’s comparative properties.
7
October 1994 is often marked as the ?rst time when an advertisement appeared on the Web. It was in the on-line magazine “Hotwired” and the advertisers were Zima, AT&T, Volvo and a few others (Brandweek). 7
Research has also been directed to studying the Internet and the Web on its own merits, (see for instance Hoffman & Novak, 1996, Dréze & Zufryden, 1997, Rossiter & Bellman 1999, Gallagher, Foster & Parsons 2001, Zhou & Bao, 2002, Novak & Hoffman, 2003 and Novak, Hoffman & Venkatesh 2004,) which is an important step in determining how marketing communication, and in particular advertising, works on the World Wide Web. The study to be conducted in this research belongs to this category and is thereby not comparing the Web with other types of media. When studying the Internet and the World Wide Web from a marketing communication perspective it is obvious that the Web is one out of many media channels. It can be used to transfer information and messages to consumers. Other media channels that are frequently being used for marketing purposes are TV, Radio, news papers and the like. In this context the World Wide Web becomes just an additional way to distribute marketing communication to target markets. In ?gure 1.2 frequently used media are mapped with the Internet and the Web inserted into the media terrain.
Media Channels
Media Channels Figure 1.2
Media Channels
Tv Newspaper Billboards Radio Magazines Internet
WAP Entertainment Websites
e-mail News paper Websites
IM
News groups Search engine Websites
WWW e-mail Websites
Figure 1.2. Map showing media channels and the new digital media terrain.
The model above shows that the Internet is one out of many possible media channels. The Web, in turn, is merely one out of a number of Internet services. These Internet services are different communication channels that can be used to target consumers and to convey messages. Advertising can for instance be done through e-mail or instant messaging services and of course also on the Web. The Web is however not merely one media but can be divided into different Web site categories depending on, for instance, what activities the user can perform on a particular Web site. There are well known Websites that can be put into the various categories in the model in ?gure 1.2 such as Hotmail, Yahoo, New York Post, Le Monde, Google, Altavista, Disney or Lunar Storm (a Swedish Web site)8.
8
Notice that the New York Post and Le Monde merely have extended their business mission from print to Web whereas the other examples are entirely new enterprises in the Web domain. Disney is perhaps not entirely new since they have for a long time been in the area of entertainment. Though, the kind of entertainment they offer on the Web differs quite a bit from what they offer in their entertainment parks for instance. 8
In line with the above model, Hotmail and Yahoo belong to email Websites; NYP and Le Monde are newspaper Web sites et cetera. Of particular interest here is whether one and the same advertisement would have the same impact when presented in different Websites or Web contexts. Web context is here de?ned in a similar way as, for instance, De Pelsmacker, Geuens and Anckaert (2002), that is, the context comprises of the characteristics of the medium in which an ad is inserted. The impact of dissimilar media contexts on advertising effectiveness have been studied in for example TV and print. The context, in which a commercial or an advertisement has been placed in, has been found to be important for the effectiveness of advertising (De Pelsmacker, Geuens & Anckaert, 2002). Studies have explored the interplay between the advertisement and the media or media related factors, attempting to investigate which media factors that have effect on recognition and recall. Nowland et al and Politz (1962; 1962) have shown that different media vehicles can generate different communication effects even on the same target group using the same advertisements. The media vehicle is thereby central when deciding on which vehicle to use since different vehicles have different impact on the audience. Other studies (for instance, Harmon and Coney, 1982; Sternthal, Dholakia, and Leavitt, 1978; Sternthal, Phillips, and Dholakia, 1978, Hermes 1996) show that also content affects the impact of advertisements as well as credibility of the source. Advertising on the Web can be described in relation to how the advertisements are presented compared to other media channels. For example, TV advertisements - commercials - are presented sequentially so that program content is interrupted by a commercial break when the commercial is aired. This is also applicable for Radio where program content and commercials, with very few exceptions are presented serially. This practice clearly deviates from the normal case for Web advertising, since advertisements on the Web are usually presented in parallel with the Web site content. In print, advertisements are also presented in parallel with the program content. The reader will thereby be exposed to both the program content and the advertising simultaneously. In this respect, Web advertising and print advertising are similar. The differences between the Web and other media channels were pointed out in the previous but can be summarized as; the Web represents an interactive media with a greater richness of information than what can be found in traditional media (Evans & Wurster, 1997). Furthermore, the Web exhibits a greater interactivity, ?exibility and complexity than both print and broadcast media (Bruner & Kumar, 2000, Karson and Korgaonkar, 2001). The Web also has a visually distinct appearance, which differentiates the Web from other media channels. One of the visual features is that pages to additional information are presented as links that are clickable. As a result, the entire browsing system is functioning and organized in a different way as opposed to print or other media channels.
9
In addition, Websites are not homogeneous. On the contrary, there is a great variation in how they are designed and for what purpose they are being used. Websites can be categorized into groups that share similar functions or visual appearances such as search engines, portals, Web shops, news Web sites and the like. These Website categories represent different contexts in which advertising can be presented. Hence, visiting different Websites means visiting different contexts and the impact of context on advertisements may differ among Web contexts. On the Web sites described in the previous paragraph the Web user can perform many different activities. Various activities will call for more or less commitment from the Web user. In some cases the user can just surf around for entertainment and wander from Web page to Web page and be amused by various activities she is performing. In other cases the Web user is searching on Web sites and Web pages in pursue of certain kind of information or to perform certain activities. These two broad cases sketched are examples of different levels of goal orientation or internal drive guiding the behavior exhibited by the Web user. Janiszewski (1998) describes two broad types of behavior when searching for visual information; goal-directed search and exploratory search. Goal directed visual search is the type of behavior when the consumer is motivated and gathers information in an ef?cient way. Exploratory visual search is the type of behavior when the consumer lacks the motivation or experience to search ef?ciently (ibid). Janiszewski is taking as an example of exploratory visual search when a consumer is paging through a catalogue in an unplanned manner. This type of behavior is very similar to browsing the Web as is described above. However, the level of goal orientation and internal drive can also be seen as an additional contextual dimension that may have the potential to in?uence how advertising is attended to and processed. Since the Web offers a high level of interactivity, and other properties, this contextual dimension may be of great importance. Two aspects of the context dimension have been described in the above. The various Web contexts where advertising is embedded can be considered as mainly an external dimension whereas the Web users’ level of goal orientation may be considered as an internal dimension or more connected to the individual. Nonetheless, these two aspects of context may in?uence and necessitate different levels of attention from the consumer and thereby affect the impact of advertising. With the above information at hand, the ?rst part of the research problem is formulated as: Are there differences in effects regarding attention to advertising between different contexts and if so, how can these differences be explained?
10
The purpose of the ?rst study is to compare how the level of attention distributed to an advertisement varies between different Websites and different levels of goal-orientation. A number of experiments are to be set up to study the context dimension of attention to advertising. A more detailed description of the method and the speci?c design of the experiments are to be found in the method chapter.
Advertising and Attention Getting on the WWW
In a media world with increasing din and media clutter, advertisers are struggling to break through the noise and get the customers to attend to their messages. Advertisers are using different tactics to make their advertising to become attention getting or attention grabbing (Campbell, 1995). The quest for attention is part of the competition and re?ned techniques may result in a competitive edge. However, according to Pieters, Rosbergen and Hartog the main focus of consumer research has been on “information processing and on the effects of advertising on attitude change” (1996, p. 242). They are furthermore stating that “…little is known about processes of attention, in particular of visual attention to advertising” (ibid). This statement supports this study and that research in this area would be appreciated by the research community. In line with Pieters et al, the importance of the attention construct has been highlighted and emphasized by other researchers. For instance, Rossiter and Percy (2001, p. 167) are revealing their stance by stating “Before anything else can occur, you must ?rst pay attention to the advertising.” This indicates that attention is a prerequisite that is taking place prior to all other constructs, that advertising researchers aim to study9. The fact that not much research has been conducted in this area, as Pieters et al is stating reinforces the notion that there indeed are reasons to devote more research efforts to this interesting topic. Horace Schwerin (1967, p. 56-57) states in a somewhat similar fashion as Rossiter and Percy that, “The opening sequence of any commercial is of key importance, since advertisers must capture and hold the attention of viewers”. Janiszewski and Bickart (1994, p. 329) are equally concerned about the importance of attention when they argue that “Despite the tremendous amount of money spent on buying consumer attention, little to no research is done on consumer attention”. Additional reasons why attention is important is for instance that, “Catching the consumer’s attention can keep her or him from mentally tuning out, switching focus to an alternative activity, or zapping to another channel.” (Campbell, 1995, p. 226). Furthermore, increased processing attention has been found to lead to increased information processing and more positive
9
For example, memory decay, attitude formation, purchase intention or the like. 11
attitudes (Cialdini, Petty & Cacioppo, 1981, Petty, Cacioppo & Schumann, 1983). Increased attention and processing also make attitudes more persistent and more resistant to negative information (Haugtvedt & Strathman, 1990). Summarizing the above it is not far fetched to conclude that attention is imperative to advertising, no matter what media is being used. The strategies that are used to attract attention differ somewhat between media channels as a natural consequence of the inherent properties of a given media. In TV and Radio, attention getting tactics have been used for quite some time and there are many ways to attract attention. One small example of how to attract attention in the TV media is to delay brand identi?cation (and sometimes also product category) until the end of the commercial (Campbell, 2001). In doing so the customer will not know what product is being advertised until after processing the entire commercial. The purpose with this presentation technique is to create suspense (Fazio, Herr & Powell, 1992). Attention getting techniques in a World Wide Web setting is a relatively new business and the starting point is the Web advertisement and its properties. The modern Web advertisement of today has developed from the rather unsophisticated “banner ads” that were used in the early era of the World Wide Web. While many may have forgotten them it is perhaps in place to refresh the memory. The early Web advertisements can perhaps best be described as a one colored rectangular square blinking with a brand name (or the like) and a blinking text stating “click here”, “buy now” or some similar phrase. The old Web banners were an oddity in the advertisement family (print ads, TV and Radio commercials, outdoor advertising etc.). When comparing to advertisements in other media channels the banner ad was the most similar to billboards or banners used in for instance sport arenas or the like. Its shape and form were similar but the banner ad was placed in a Web environment instead of in a row of banners in an arena. This notion is supported by Lei (2000, p. 469) who describes banners in the following manner “At best, they should be thought of as mini outdoor billboards, with similar creative limitations.” By this Lei meant that the old banners could not convey complex messages. Instead, these banners were designed with the purpose to generate click-through to another target site where more detailed information could be gained. In comparison, the modern Web advertisements have a visual appearance that is more similar to print advertisements that can be found in traditional newspapers and magazines. They are square-like as opposed to the very rectangular banner advertisements. Furthermore, the Web advertisements have an elaborated and “complete” message compared to the headline/punch line oriented banner advertisements. Even though most Web advertisements are still clickable they are not primarily aiming at generating click-through to a target site. Instead, they are presenting the product or the like with the purpose to promote the relevant product by communicating an advertising message that is presented on the Web advertisement. It is worth to notice that an advertising message may serve various purposes. One advertisement may be
12
designed to increase awareness or enhance the company’s image in some sense, another advertisement may have the purpose to explain how a certain product is to be used and a third may be aiming at triggering purchase. Regardless of what purpose the advertiser has, it is no longer necessary to generate clickthrough. It is of course, by no means a disadvantage when the consumer clicks on the advertisement and learns further about the company and/or the product on a certain target Web site. When observing the attention getting strategy and/or the various techniques that can be used for advertisements it is obvious that the Web differs from print. Commonly, in print ads the advertisement designer can use colors, pictures, texts and size to impact the attention to the advertisement. On the Web, the advertisement designer can use the same means (colors, pictures, texts, size etc) but also the speci?c and inherent features that the Web can add to an advertisement, i.e. animation, pop-up effects, ?ash, size increasedecrease and other potentially attention getting techniques. This is apparently a signi?cant disparity that differentiates the Web advertisements from the print advertisements. The print advertisements are, by default, static since they cannot be animated, pop-up or the like. The effect that can be added to a Web advertisement in a Web context is an additional dimension of an advertisement which can increase the attention to the advertisement. How this dimension impacts advertising has not been systematically studied. With this information the second part of the research problem can be formulated: Are there differences in effects between different “attention getting techniques” and if so, how can these differences be explained? The purpose with the second study is to conduct a series of experiments in order to investigate how Web advertisements differ in their ability to generate attention in a Web environment. When studying various types of advertisements and how powerful they can be at attracting attention, it ought to be kept in consideration the fact that certain types of advertisements are not always appreciated by Web surfers. Some of the more aggressive attention getting techniques have been reported to stir up negative emotions among Web surfers. This is an unwanted collateral damage that occurs when the advertiser is eager to get the attention of its audience. How this is affecting the advertising has been researched but results stay ambiguous. Recent studies related to advertising on the Web have given contradicting results regarding people’s attitudes towards Web advertising. Ducoffe (1996) presented ?ndings showing that Web surfers did not ?nd advertising disturbing or interfering with the Web surfer’s activities.
13
From an advertising point of view, this was good news. Other researchers such as Brackett, Benjamin and Carr (2001) have, on the other hand, found examples where the Web surfers considered advertising on the Internet being irritating, annoying, and even insulting to people’s intelligence. The results are alarming, raising questions about what is being done erroneous. It is important to notice that Bracket et al’s ?ndings were presented ?ve years later than Ducoffe’s. It can be argued that Ducoffe’s study was conducted at a time when Web surfers were curious about every aspect of the Web while Brackett et al’s study was conducted when consumers had already gotten acquainted to the Web and when the novelty and initial curiosity had faded. However, attitudes towards advertisements, and the context where it is presented, are important factors since it is anything but positive when the advertisement is considered an insult by the target audience to which the company is trying to communicate with. The ambition is therefore to attempt to shed light on this issue. A second purpose with the second study is therefore to investigate if and how individuals’ attitudes towards attention getting techniques differ between different attention getting techniques, and what the consequences may be.
Modeling Attention to Web Advertising
The core of advertising research can be traced back to the very end of the nineteenth century. It evolved from personal selling in general and from the AIDA model in particular, which in fact was a model of personal selling. It was adapted by researchers in advertising and became a foundation for further research within the area (Strong, 1925, p.76). The AIDA construct (Attention - Interest - Desire - Action) was the ?rst formal advertising model (ibid) and others have followed such as the well known Elaboration Likelihood Model of persuasion (Petty, Cacioppo & Schumann, 1983). Many of the advertising models are including the attention construct but how attention to advertising works in a greater detail (for a comprehensive review of advertising models, see Vakratsas & Ambler, 1999) is seldom elaborated on. There is however a neighboring discipline, experimental psychology, where attention has been studied extensively. A number of researchers (see for instance Broadbent, 1958, Deutsch & Deutsch, 1963, Treisman, 1964, Walley & Weiden, 1973, Kahneman, 1973, Moran & Desimone, 1985) have modeled attention elaborately. Common for these researchers’ models is that they provide a richer understanding for how attention works than what advertising models do. This is perhaps not unexpected but given that attention is often treated, in advertising models, as a box amid other boxes it is perchance in place to intricate on how attention is modeled in advertising (see for instance the ELM or the AIDA model and others). The bene?t of such an approach would be to extend the knowledge of one area into another.
14
Amongst the theories on attention that have been put forward, Treisman’s model has received more positive reception than others’ (Johnston & Heinz, 1978). Treisman has been one of the proponents of an early selection theory (Broadbent’s theory also belongs to the early selection category but is more in?exible). On the other hand, Treisman’s view on attention is that there is a sensory register allowing stimulus input to pass an attenuator that reduces or attenuates processing of information outside focal attention (Treisman, 1960, Eysenck and Keane, 2000). As a consequence, limited amounts of information can reach short-term memory where information is stored temporarily. Crucial in the selective attention process is how input is transferred from the sensory register to the short-term memory. Broadbent suggested a ‘?lter’ that mechanically screened out non-attended stimuli (Keane, 2000), which was shown to be incorrect by Treisman (1960). Treisman, on the other hand, referred to an ‘attenuating’ mechanism. In her view this mechanism may signi?cantly reduce the incoming input, but does not necessarily block information in a similar conduct as in Broadbent’s ?lter (Treisman, 1964, 1960). Elaborating around the concepts and the function of the attention system, rather than treating it as one box, and to build a model infused into an advertising setting could provide greater insights and understanding concerning how attention operates in advertising. It provides a deeper understanding of how contextual factors or attention getting techniques are impacting the attention to a Web advertisement. Thus, a supplementary purpose is to test an adapted model, originating from attention models in experimental psychology, in an advertising setting in order to provide explanations of how attention to advertising works. What is to be focused on, in speci?c, is the relationship between the Web environment/ context, attention getting and the attention system that may allow advertising input to pass through or to be attenuated.
Measuring Advertising on the Web and Pricing Methods
Hitherto, the focal areas in this dissertation have been outlined and described. The research problem has been presented and the core purposes along with it. Still, when studying attention in a Web advertising context there are a number of supplementary areas that can be of interest to investigate. In addition to the main studies, an opportunity was given to add two minor areas namely, “how to measure advertising on the Web” and “pricing strategies for Web advertising”. The ?rst area (how to measure…) is a study of an old problem in a new setting (the Web setting). Nonetheless, it is important to empirically con?rm and replicate how advertising is to be measured on the Web in view of the fact that it is a different media channel as opposed to the traditional ones, which has been explained previously. The results from the ?rst area, however, may have signi?cant implications for the second area “pricing strategies for Web advertising”. If a different measurement method of advertising effect (recognition & recall instead of click-through) is more appropriate than the
15
method that is currently being used, then perhaps Web advertising pricing strategies also need to be revised. In other words; if click-through is not accounting for the entire advertising effect, then the Web advertising pricing strategy needs to be changed since the foundation of the current Web advertising pricing strategy is based on the click-through measure. In connection to these two studies it is in place to emphasize that it is Web advertisements that are being targeted and not hypertext links, keywords or other kinds of Web marketing. It is advisable to keep in mind that keywords generated from search engines and the like often are not quali?ed as advertisements when using a modern de?nition of advertising. Next, the two above mentioned additional research issues will be elaborated on further. The ?rst one to be described is; Recognition & Recall versus Clickthrough as a means to measure Web advertising effect. During the ?fties some inspirational research took place, which Darrell Lucas (1960) described as, “probably the ?nest example of experimental design ever reported on advertising research”, when referring to the PARM study by the Alfred Politz organization (Krugman, 1985). The PARM study (A study of Printed Advertising Rating Methods) was a study to investigate issues regarding measurement of advertising, in particular differences between recognition and recall of advertisements. This study showed that a signi?cant difference between the recall and recognition measure existed and that recognition was more sensitive than recall. At the same time, the study gave support to the small samples that were used by leading practitioners at the time such as Starch and Gallup & Robinson (Krugman, 1986). Ever since the heydays of the PARM study, the advertising industry as well as academia has been working on conducts to improve the techniques to measure advertising effectiveness. Nonetheless, in the case of how Web advertising is measured, the inherent properties of the Web could automatically provide a response measure. This measure, i.e. the click-through measure, is measuring the number of individuals that click on an advertisement. When this method to measure responses was identi?ed, the click-through measure soon became widely used as a measure of advertising effect without much re?ection over its appropriateness. In this context, it is valuable to remember the fact that there is a number of means to measure advertising, for instance recognition, recall, attitude and opinion measurements, comprehension measures, believability measures, persuasion measures, buying predisposition, ad ratings and behavioral measures (Leckenby and Plummer, 1983). These various measures are targeting different effects that advertising can achieve. For instance, recognition and recall are often used to measure the initial attention phase in the advertising sequence. Krugman states that the special virtue of the recognition measure is to elicit information of brief low-attention exposures to advertising, whereas
16
recall is for advertising that has been more closely attended to (1986). Since recognition is more sensitive than recall the recognition measure will comprise individuals who are able to recall the advertising as well. Those who recognize the advertising will also recall it (Krugman, 1986). Recognition as a measure of advertising effect refers to a respondent’s subjective estimate that she or he looked at or attended to an advertisement that has been previously shown. Recall re?ects an advertisements’ ability to register the sponsor’s name (Wells, 2000). These measures that were included in the PARM study were relevant rating methods for printed advertising. Despite the well established position that these methods had, the click-through measure became widely used as an advertising measure when the World Wide Web appeared. In fact, some proponents of the click-through measure were discrediting the previous methods and before long the most frequently used measure was click-through, a position which is still holding (Chandon, Chtourou and Fortin, 2003). In studies of Web advertising, the focus has often been on the effects of banner advertisements with click-through frequencies being used as a means to measures effects. The click-through frequencies have however dropped to below 1% making it an unreliable measure when using small samples (Khermouch and Lowry 2001). There are reasons to assume that an advertisement has effect albeit an absence of a “click on the advertisement” (Interactive Advertising Buereau, 1997). This makes it logical to shift from click-through studies to study advertisements on the World Wide Web using traditional methods (for instance recognition and recall) as measurement. In doing so, it will be made possible to extend the measurement procedures that have been applied for decades in traditional advertising and to transfer them into the Web advertising domain. Using recall and in particular recognition, seem to be a natural option that is also in line with what, for instance Krugman advocates (1986, 1988). Krugman, is one of the researchers that has most credibly pointed at the necessity to use recognition and recall in, for instance, TV media. Researchers have in fact been looking into the Web measurement issue but still there are disbeliefs and most of all, the industry is using the click-through metric frequently and referring to it as a means to measure advertising effect. Briggs and Hollis (1997, p. 33) are stating that “The practice of evaluating Web advertising on the basis of click-through is like evaluating TV-ads for automobiles on the basis of how many people visit the showroom the next day”. Based on this, the purpose with this ?rst minor study is to investigate whether there indeed is an effect of Web advertising as measured by recognition and recall and if so, how big this effect is in relation to the click-through measure. This may enlighten somewhat regarding the appropriateness of using the clickthrough measure as a means when measuring advertising effects. The second additional research issue to be addressed is; Pay per click versus Pay per view as a basis for pricing in Web advertising.
17
Mangàni (2003) argues that pricing methods on the Web has evolved along a dissimilar path compared to broadcast or newspapers. Mangàni is attributing this slightly different evolution to the technological possibility to track and analyze the behavior of Web surfers. On the Web there are mainly two ways to charge for advertisements, pay-per-view or pay-per-click. Pay-per-view (also called CPM10) is a model where the advertiser pays for the number of impressions whereas pay-per-click is a method to charge for the number of Web surfers that actually clicked on the Web advertisement (Mangani). The click-through measure is a straightforward and quick way to measure click frequency on a Web advertisement since the statistics are easily accessed and produced in the same moment as the customer clicks on the advertisement (Berthon et al., 1996). Consequently, publishers and advertisers started to use this measure as an indication of the advertising effect in the early days of the World Wide Web. Without much contemplation or any studies of implications, the click-through measure also became a foundation for pricing. The logic for this is based on the following train of thought; if click-through is measuring advertising effect then it is reasonable that advertisers pay for what they obtain, i.e. clicks. From the advertisers’ perspective this is an appealing idea since this would mean that they are only paying for explicit action in the form of a click. The entire idea rests upon the assumption that there is no effect beyond the click and therefore the advertiser should only pay for the number of individuals that has “clicked through”. The answer from the ?rst area will provide an answer to whether there is an effect beyond the click. If there is no effect beyond the click then it is reasonable to have a pricing strategy which is based on clicks. If there is indeed an effect (in addition to the click) then there are motives to state that an advertiser who is merely paying per click will get an additional effect that is not being paid for. Briggs and Hollis (1997) have compared the practice of assessing Web advertising based on click-through by evaluating TV-ads for automobiles based on the number of people visiting the showroom the following day. Consequently, the purpose of the second study is to elaborate on the aptness of using pay-per-click as a pricing method for Web advertising.
10
Cost per thousand – M is the latin letter representing one thousand.
18
Outline of the dissertation
Chapter two of this dissertation provides a theoretical background and sets the stage for the study to be conducted. It also features the theoretical framework of the dissertation mainly examining theories of advertising and attention. Chapter three contains a summary of the theoretical argumentations leading to the formulation of hypotheses to be tested. In addition, a restructured model that originates from the attention research ?eld is inserted into an advertising setting. Hypotheses are formed to test the model. Chapter four provides the methodology, the overall design and way of conduct of the various experiments. This chapter is also addressing issues regarding how and what type of analysis that will be conducted on the collected data sets. A presentation of gathered data will kick-off chapter ?ve and initiates the analysis of the data. Various methods for analysis are being used. Chapter six comprises of conclusions, a general discussion and guidelines for further research. Following this a section with references and appendices is presented.
Outline of the Dissertation
Chapter 1 Chapter 2 Chapter 3 Chapter 4 Chapter 5 Chapter 6 Introduction, Research problem and purpose Theoretical background and theoretical framework Summary of theory and hypotheses Research Method Results and Analysis Conclusion and discussion Figure 1.3
Figure 1.3. Outline of the dissertation.
19
Chapter Two
Theoretical Framework
In this chapter the theoretical framework will be presented. The framework will address theoretical issues associated to the research problem. While in the previous chapter a brief portrait of the relevant theories has been given, this chapter will provide a presentation of a more elaborate theoretical overview.
20
The origin of Advertising
From a historical perspective, advertising can be seen as originating from the public criers on town markets (Dyer, 1982) in ancient cities such as Ur, Babylon and Jerusalem. The public crier served the function of providing timely, and some times persuasive, information about goods and services available on the market. In a similar fashion, the modern advertising of today is communicating to customers about available offers. The means of the public crier and those of modern advertising may differ but the basics remain the same. Daniel Starch argues that advertising in some form probably has existed “since the time when men lived in communities and competed with one another for the necessities and luxuries of life” (Starch, 1923, p. 17). Without going any further into the archives of advertising, it can be concluded that advertising is not just a recent phenomenon but has been an intrinsic component of society for quite some time. Advertising as a scienti?c area for academic study shares similar characteristics as the marketing subject. Advertising is a subject that is often applying theories and ideas from other disciplines. For instance, the in?uence of advertising on consumers is understood in light of theoretical perspectives brought in from linguistics, literature, psychology, sociology, anthropology, economics and other subjects (MacInnis, 2005). Methods like ethnography, survey research, experimentation, quantitative modeling and the like are being used to assure a diversi?ed understanding of the advertising phenomena. “Through this multimethod, the multi-theory perspective, we (the “blind men”), gain insight into the “elephant” known as advertising” (MacInnis, p. 14, 2005). Albeit this research does not span over all the areas and use all the methods stated above, it is clear that this research is an intersection where multiple subjects unite. In this sense, this study is following a traditional path in marketing and advertising research. Advertising, being one of the means which can be used to market an offer by communicating to an audience, is intimately connected to primarily two areas. One area is marketing, which has the purpose to create, communicate and deliver value to customers (Darroch et al., 2004). Another area is communication,1 which is the process of establishing a commonness or oneness of thought between a sender and a receiver (Schramm, 1955) or the
1
The area of communication theory has to some extent been subject to controversies regarding origin, theoretical foundation etc. In marketing, Schramm’s model of communication is often used as a starting point and it is perhaps important to consider the following; Schramm’s model is congruent with Claude Shannon’s general theory on signal transmission developed in the electronics industry and more speci?cally at Bell Telephone Company. According to for instance David Richie, Weaver (Shannons “co-author” in 1949) in speculative manners extrapolated Shannon’s theory on signal transmission into the social sciences as a theory of communication. Though, this is by no means Shannon’s wrong doing or as Ritchie states “To criticize Shannon’s model as inapplicable to the complexities of human communication is to criticize a rowboat because it is not a whale” (Ritchie, p. 280 1986). Shannon’s signal transmission theory de?nes the information rate of signal transmission and not the information content of a message. “The semantic aspects of communication are irrelevant to the engineering aspects” (Shannon, 1949) and since the purpose of Shannon’s theory was not communication it should not be used for other purposes. Though, other researchers like for instance Gonzales (1988), Finn & Roberts (1984) see it as revitalizing to bring the theory to other situations that are not strictly analogous. 21
process by which individuals share meaning (Dibb & Simkin, 1991). Duncan & Moriarty (1998) argue that marketing theory and communication theory share common roots and thereby they are enriching each other. Amalgamating marketing and communication consequently results into the area of marketing communication. In a marketing communication context, information is distributed to inform, persuade, motivate and to make potential customers aware of an organization’s offering (Keller, 2001). Marketing communication comprises of a variety of activities that aims at communicating with the company’s customers. Marketing communication tools are typically divided into four or more de?ned areas and these tools or sub functions are often referred to as the promotional mix. The tools in the promotional mix are advertising, personal selling, sales promotion, public relations and can also be extended with for instance direct marketing, packaging, point of purchase display and event marketing (Keller, 2001). Of these mentioned tools, advertising will be scrutinized in greater detail since it is the main topic at hand and of special interest in this theoretical framework.
What is Advertising and What is a Web Advertisement?
In the literature, a number of descriptions and de?nitions of advertising can be found. An early de?nition by Daniel Starch (1923, p. 5) suggests, “The simplest de?nition of advertising, and one that will probably meet the test of critical examination, is that advertising is selling in print. Or to put it more completely, commercial advertising consists in presenting a commodity in print to the people in such a way that they may be induced to buy it”. Starch’s de?nition is easy to comprehend but is perhaps not the most elaborate one. A further problem is that the de?nition is too narrow in that it merely recognizes advertising in print as advertising. This means that it is excluding all other types of contemporary media channels for advertising. Yet, it ought to be kept in mind that Starch’s de?nition of advertising was presented at a time when print was indeed the major means to convey advertising messages. In a recent discussion on the topic of advertising de?nitions, the editor of journal of advertising, Faber (2002, p. 1), argued that de?nitions of advertising usually embrace components such as “paid’’ messages, “from an identi?ed sponsor” in “mass media” with the goal of trying to “persuade or in?uence”. The same year as Faber published his notions, Richards and Curran (2002) screened a number of advertising de?nitions in order to capture the essence and put together a “current de?nition” which is as follows; “Advertising is a paid non-personal communication from an identi?ed sponsor, using mass media to persuade or in?uence an audience”. Richards and Curran’s de?nition of advertising aims at being general and thereby harbors all kinds of advertising, including Web advertising. Though, Web advertisements are somewhat different than for instance print advertisements. It is therefore appropriate to elaborate on what a Web advertisement is and what kind of Web advertisement is being targeted in this study.
22
Lei (2000) described the early banners as mini outdoor billboards that could not convey complex messages. This limitation that Lei pointed at in year 2000 has been eradicated and the modern Web advertising does not suffer from it any longer. On the contrary, Web advertisements can now provide quite complex messages as they are not constrained to being button advertisements or mini-billboards as they once were. Visually, the Web advertisements are quite similar to print advertisements but can in fact exhibit greater complexity, especially when animated. It is therefore misleading to speak about “Web banners”, when referring to advertisements on the Web. The word banner, in its basic form, denotes a rather primitive type of advertisement where usually a brand name or company name is printed on a large rectangular piece of fabric with diminutive room for anything else to be communicated. Figure 2.1 below shows what is often accredited to be the very ?rst banner ad used on the World Wide Web. The advertisement in this ?gure is an archetypal example of the very simple banner-like advertisements that were once frequently used on the Web. This type of advertisement is becoming less common today and is not to be studied here.
One of the First Web Advertisements
Figure 2.1
Figure 2.1 Depicted is the ?rst banner ad used on the World Wide Web by AT&T in the on-line magazine Hot Wired in 1994. The advertisement is reprinted with AT&T’s kind permission (2006).
Concurrent with the development of the Web banners, an entire industry of search engine generated keywords or “AdWords”2 have surfaced on the World Wide Web. These are clickable words in a text or small “AdWords” next to search results on a search engine. Albeit a majority of these “text advertisements” can qualify and be de?ned as advertisements, they are not of focal interest in this study. Instead, the following depiction may serve as a ?nal guideline of what is meant by a “Web advertisement” here: The Web advertisements of today, contrary to the early Web banners, often have some basic properties such as an image or description of a product, a brand name, copy text and a headline put together into an advertisement entity – the Web advertisement. Visually, the Web advertisements are similar to print advertisements and they are both intended to convey advertising messages. However, dissimilar to print advertisements, the Web advertisements are frequently published on Web pages on the World Wide Web. Web advertisements
2
See for instance Google. 23
Contemporary Web Advertisements
Figure 2.2
Figure 2.2. Depicted are two contemporary Web advertisements next to each other. On the left in the ?gure, an image advertisement by British Petroleum is presented. Both advertisements are in fact animated using multiple messages that cannot be seen here. On the right in the ?gure, an advertisement by Avis is presented. The advertisements are reprinted by kind permission of British Petroleum Corporation, Great Britain and Avis, a wholly owned subsidiary of Cendant Corporation USA (2006).
can be static (like print ads) but can also be animated or use a pop-up function. In ?gure 2.2, an example of a contemporary Web advertisement is reprinted. As is apparent, the two advertisements in ?gure 2.2 do not have a great deal in common with the old banners of the nineties depicted in ?gure 2.1. These types of advertisements are also especially dissimilar to “AdWords” and clickable text from search engines such as Google or the like. Instead, these advertisements are not very different from print advertisements that are being used in newspapers and magazines. It is this type of advertisement that is of focal interest in this study. One question that can be raised at this point is where to draw the line between simple animation and full streaming video. The answer is not uncomplicated especially considering the fact that more and more animation is being used in Web advertisements and there are a number of techniques for producing motion or animation on the Web. Only a few years previously, most animated Web advertisements were alternating between two or three visual presentations of the advertisement. At present, the advertisement designer is free to choose, from static to full motion picture, making the distinction between simple animation and full motion picture (advanced animation) irrelevant or arbitrary. To confuse things even further, there are Web advertisements that occasionally use both simple animation and streaming video. Considering that animation means moving picture, it is perhaps more relevant to discuss in terms of more or less animation employed and leave it at that3.
3
The exact animation frequency used in the experiments in this dissertation is described in detail in the method chapter. 24
The Web is growing important
Traditional media like TV, Radio, billboards, newspapers, magazines et cetera are well-established channels for advertising. The Web, conversely, has appeared more recently yet is increasing in importance for advertisers as well as publishers. Figures are pointing at a rapid growth of the Web (IWS) and advertisers are also expecting the Web’s importance as a media channel to increase4 Dréze & Zufryden (1997) stated a decade ago that one of the reasons for the increased importance of the Internet is the “tremendous growth” that it has undergone in recent years. The growth after 1997, when Dréze and Zufryden made their remark, has indeed been rapid with an astonishing 1086 million users in the World in October 2006 (Internet World Stats). Hoffman and Novak (1996), who often are considered as forerunners in the area, argued that research in this area is still in its infancy and that there is a lot of research to do to map the terrain. Ducoffe (1996) and Ducoffe et al (1996) declared early that, “continuing developments in the area of new media technology represent the most important in?uence on the future of the advertising industry over the next 10 to 15 years”. Zhou and Bao (2002) argue that it is not merely the Internet’s tremendous growth that makes it interesting among researchers, but furthermore its “distinctive characteristics”. These distinctive characteristics of the Internet, sets it apart from other types of media channels since it is a type of media with hybrid properties (Dréze & Zufryden, 1997). The distinctive characteristics that are being referred to are for instance the ability to use animation, pop-up features, radio, streaming video, TV, interactivity and the like in one and the same media and all at the same time, if one so wishes. Coupey (1999) expresses the possibilities with the Internet and the Web, stating, “The unique features of the media can provide a focus to extend or to create theory, methodologies, and the discovery of new phenomena.” Altogether, there seem to be reasons to believe, that the World Wide Web is an important type of media to study and that research in advertising connected to the World Wide Web is in a state of renaissance.
Advertising is taking place in a setting
Marketing activities such as advertising are taking place in a social setting. These are in actual fact taking place ubiquitously in society, between companies and consumers. For instance, when an individual is starting the day by reading the morning paper she will be exposed to a number of advertisements. These advertisements are placed in a certain setting. Consumers are constantly exposed to brand names, company names and window displays affecting them in various ways.
4
At an industry seminar with representatives of the top advertising agencies in Sweden all participants unanimously expected the Web and the Internet to increase in importance at the expense of other media channels. No one expected TV or any of the other traditional media to have a similar development. Industry seminar (Stockholm, autumn 2005) 25
The number of potential encounters with advertising and other marketing activities are too numerous to describe. To the long line of traditional encounters, the digital ones are now also added. A typical “digital encounter” is when an individual is sur?ng the Web and being exposed to a Web advertisement that may be part of an even greater advertising campaign. This encounter is taking place in a setting that may or may not in?uence the effect of the advertisement. Kakkar and Lutz (1981) have elaborated on how the situation and environmental aspects may affect consumer behavior. According to their view, the situation can have a great deal of impact on how people receive process and make sense of communication. The signi?cance of situational variables has also been considered important by Ward and Robertson (1973, p. 26) who argue that it is likely that situational variables account for more variance than what actor related variables do. For instance, many buyer behaviors are only exhibited under speci?c conditions, in speci?c situations (Lavidge 1966). According to Belk (1975, p. 157), a situation is a point in time and space and furthermore all those factors observable “that do not follow from a knowledge of personal (intra-individual) and stimulus (choice alternative) attributes and which have a demonstrable and systematic effect on current behavior” (Belk, 1974a). To identify the boundaries of what constitutes a situation, Belk (1975, p. 159) puts forward ?ve groups of situational characteristics representing general features. Physical Surroundings – the most apparent features of a situation. The physical surrounding includes such aspects as geographical and institutional location, décor, sounds, aromas, lighting, visible con?gurations of merchandise or “other material surrounding the stimulus object” (Belk (1975, p. 159)). Social Surroundings – which provide additional depth to the situation such as other persons present, their properties and roles, interpersonal interactions going on and the like. Temporal Perspective – a dimension of situations speci?ed in time units and related to some other entity. Task De?nition – “features of a situation include an intent or requirement to select, shop for, or obtain information about a general or speci?c purchase. In addition, task may re?ect different buyer and user roles anticipated by the individual” (Belk (1975, p. 159)). Antecedent States – is another group of factors that can characterize a situation. These are moods or short termed conditions immediately antecedent to a situation. They are states that the individual brings into the situation and not the result from being in a certain situation.
26
Albeit Belk’s characterization of a situation is not targeting the Web case per se, it is apparent that there are similarities. The previously mentioned individual in the “digital encounter” who is reading the news on a Web page will be exposed to the “physical surrounding” of the Web site with its distinct features. It is naturally also relevant to consider the social surrounding as there may be other individuals present. When visiting a Web site, it is applicable to see the event in terms of a de?ned task. The Web surfer who is visiting a Web page has the intention to do something – to solve a task. This task solving can, in more common words, take the form of reading an article, ?nd the weather information for the following day or to purchase something. To de?ne and regard the Web interaction as a task is an appealing way to describe the user – Web engagement. This notion will be further elaborated in the attention section of this chapter and used as an instrument to describe the advertising setting. In a number of studies (see f.i. Lavidge, 1966; Ward & Robertson, 1973; Belk 1974a, 1975; and Kakkar & Lutz, 1981) the signi?cance of situational factors has been deemed imperative. Extending these notions to the advertising area would mean that advertising surrounding us and reaching our minds through our senses, is taking place in a setting where situational or contextual factors evidently are at play. This means that the design of a Web site, the colors, the framing of various areas and, in particular, the complexity of the Web site may have an impact on how people attend to, receive, process and make sense of communication. One setting may be more effective, from an advertising point of view, than another setting. In ?gure 2.3 below it is illustrated how the “total” environment, in which a Web advertisement may be inserted, is built up by external factors outside the screen; the Web environment, the focal area and other surrounding areas on the screen. If the screen environment is described in detail, the following components can be found; the screen frame and the screen. The visual part of the screen comprises of the task bar, which is part of the operating system that is being used. The Web browser is also visible including the navigation toolbar and other toolbars that can be customized. When a Web page is loaded, the Web page and the information presented on the Web page will be covering most of the screen. The Web page itself can be seen as an environment which can be altered in an in?nite number of ways. It is this environment that is being referred to, when stating “Web environment”, in this study. Using the terminology that has been discussed hitherto it is appropriate to state that the Web environment can be seen as a task environment in which a task (for instance to read the Web news) is being solved.
27
The Web Environment
Advertisement Focal area
Figure 2.3
Surrounding areas
The web screen
Other factors (external, outside of the screen).
The web user is reading this and is affected by other areas of the screen as well.
Web environment
Figure 2.3 The environment in which the advertising is presented.
In ?gure 2.3 the various parts that is building up the setting is put together. It is in this context that Web advertising is taking place. This setting is divergent to other settings as it includes new elements that may or may not in?uence the effects of advertising. It is a computer mediated communication situation with a different kind of format as opposed to for instance the print media. It is a different kind of interaction since the user is not turning pages in a paper but instead clicking on links leading to additional information and additional possibilities to present the advertisements – for instance by using animation, pop-ups and the like. This model will be elaborated on at a later stage in this chapter in order to describe its various components in theoretical terms.
Outcomes of Advertising
Hitherto, it has been discussed how advertising is de?ned, that the World Wide Web is a different kind of environment and that advertising is taking place in a setting. Equally important is to add the outcomes of advertising. Companies, governments, non-pro?t organizations, political parties, and other entities employ advertising because they want speci?c outcomes from it. An advertising campaign is generally undertaken in order to pursue certain goals. The goals that are pursued are often formulated by the organization or an external partner, which is designing the advertising. The reason why an organization does advertising varies but it is often done to achieve some basic effect. It can for instance be to inform (Keller, 2001) the public or a target market about services available, or to change the attitudes (Petty & Cacioppo, 1984) towards some object or activity, for example, to not drink alcohol and drive. Other examples of desirable effects are to create interest (Smith & Swinyard, 1982) towards the product that a company is promoting or to motivate or persuade (Richards & Curran, 2002) the consumers to choose offer A before offer B. For companies that are striving to make pro?t it is important to achieve action or an actual purchase (Lavidge & Steiner 1961) which frequently is the ultimate goal in a chain of events. However, before
28
coming to the very end of the purchase process it is imperative, as a ?rst step in this chain, to capture the attention (Janiszewski and Bickart, 1994) of the audience targeted. From a communication perspective, the organization ought to start by assuring that it has the consumers’ attention. Only then it can be effective in its endeavor to transfer information or to present an offer in a favorable and appealing way. If the advertiser does not capture the attention of the consumer, then the opportunity to communicate with her is lost. With an ever-increasing number of advertisements, competing for existence the consumers’ attention has become a scarce resource (Adler & Firestone, 1997; Davenport & Beck, 2001; Pieters & Wedel, 2004). From an individual advertiser’s perspective, the task is to break through the clutter of advertising messages and other kinds of information that is calling for the consumers’ attention. Krugman, one of the researchers who have focused quite some effort on attention in marketing communication and advertising, has argued that consumers are screening “[...] the advertising and distributing their attention selectively” (Krugman, p. 48, 1988), with closer attention to some advertisements and with less to others. Krugman has also emphasized the role of capturing the audience’s attention and pointed it out as particularly important. Janiszewski and Bickart (1994, p. 329) are also pointing out the initial phase of effects, attention, as important. Horace Schwerin (1967, p. 56-57) states, in line with Krugman, that “The opening sequence of any commercial is of key importance, since advertisers must capture and hold the attention of viewers to retard them from mentally or physically tuning out”. Krugman has in his research also studied how to measure attention effects adequately using measures such as recall and recognition. Unfortunate for organizations using advertising to communicate with their target audiences, there are not only desired effects that occur but every now and then also non-effects or even worse, undesired effects. To this category of unwanted effects belong, for instance, mistakes on the part of the advertiser (messages out of tune) so that the audience does not recognize the organization, its offer or brand. It can also be mistakes on the part of the media-planning agency or others involved in the process of bringing the message to the audience. Among all potential errors, and error sources, is an additional error that has been highlighted as the World Wide Web has grown and developed to become a part of everyday life. It is connected with the advertisers’ ambition or overambition to get the attention of the audience. On the Web, it can be termed the “pop-up effect”, which is when Web advertisements are calling for too much attention and thereby become intrusive or annoying to the user sur?ng on the Web site where the advertising is presented. As mentioned in the previous chapter, Brackett, Benjamin and Carr (2001) found examples where the Web surfers considered advertising on the Internet being irritating, annoying, and even insulting to people’s intelligence.
29
This kind of negative effect is not what organizations generally want to be connected with, since it may instill negative5 attitudes in the audience. Attitudes are central when discussing around both positive and negative outcomes of advertising. An attitude is a latent variable that cannot be directly observed. The attitude refers to a tendency of psychological nature that is expressed by evaluating a certain entity with some degree of favorability or unfavorability (Eagly & Chaiken, 1992). The entity referred to, often called an attitude object, can be virtually anything that can be discriminated by an individual. Hence, an individual can have favorable or unfavorable responses towards an attitude object. The responses are generally characterized as being cognitive, affective or conative. These variables, cognition, affection and conation are measurable. Therefore, the researcher is observing these variables in order to state what attitude an individual has. Cognition refers to thinking, affection to emotions and feelings while conation refers to actual behavior towards an attitude object. Using the attitude terminology, it is adequate to state that a company can use advertising to initiate, alter or reinforce a consumer’s thoughts -cognitionaround a particular product or product attribute. Furthermore, it is at times desirable for advertisers to create certain emotions or feelings -affectiontowards a product or its attributes. The last part is to make the consumers take action -conation- in relation to the attitude object, for instance to purchase the product and to use certain product attributes. Even though it may seem obvious that attention getting can produce negative attitudes, other effects can work in the opposite direction. Part of the attention getting will produce brief exposures that have the potential to create an effect in the opposite direction. For instance, a Web advertisement is perceived as annoying, due to aggressive attention getting, while the brief exposure takes place. On the other hand, the consumer may forget about how annoying the advertisement was when recognizing the product and brand in the store. Since the consumer has been exposed to the brand a number of times, the brand will have a greater “familiarity” than other unfamiliar brands. Mandler, Nakamura and van Zandt (1987) suggest that mere exposures can have the potential to produce a kind of vague feeling of familiarity that later could be interpreted as liking. In a recent study, Monahan, Murphy and Zajonc (2000) arrived at the conclusion that “enhancement of one’s mood state can occur by virtue of repetition of exposure” (Monahan et al., p 465). This would indicate that repetition alone has the potential to in?uence liking. In a speci?c Web advertising case, the effects mentioned above are working in two directions. On one hand, too aggressive attention getting may evoke negative attitudes and on the other hand repeating the message may in?uence liking.
5
There are of course also cases where the advertiser wants to evoke negative attitudes towards an attitude object. This can, for instance, occur when a political party is portraying its opponents in a negative way. 30
In addition, when the consumer is in fact about to do the purchase the aggressive attention getting may be forgotten whereas the familiarity and liking may persist, in line with Monahan, Murphy and Zajonc’s ?ndings. There is nevertheless, the risk that aggressive attention getting in the advertising will affect the Web site where the advertising is presented in such a way that users may avoid the Web site or dislike it. These mentioned effects are important when interpreting the results of the studies to be conducted and especially when transferring the ?ndings into a real world situation. In this research, subjects will be exposed only at one occasion whereas in real world situations repetition is obviously a dimension to consider. At the same time, it is also worth to bring to mind that there are signi?cant methodological differences between Monahan et al’s study and the one presented in this study.
Putting the parts together
To summarize what has been discussed this far and to tie various parts together the following can be stated; Advertising is part of marketing communication and has, in recent years, come to be used on the World Wide Web. Initially, a certain kind of very simple “banner ads” were used but they have been developed into a “contemporary Web advertisement” that is described as: The Web advertisements of today, contrary to the early Web banners, often have some basic properties such as an image or description of a product, a brand name, copy text and a headline put together into an advertisement entity – the Web advertisement. Visually, the Web advertisements are similar to print advertisements and they are both intended to convey advertising messages. However, dissimilar to print advertisements, the Web advertisements are frequently published on Web pages on the World Wide Web. Moreover, all types of advertising are taking place in a setting. Web advertising is taking place in a World Wide Web setting, which has, as has been argued around, somewhat different features than for instance print advertising. The Web advertisement and the setting in which it is presented are part of what can be seen as advertising input to the consumer. The last piece to this puzzle is the outcome of this advertising input and in particular certain aspects of this outcome, namely attention and attitudes related to attention getting. Attention, being a construct primarily originating from and connected to psychology, is frequently used to describe the initial phase of advertising effect. In advertising models, this is often visualized when a consumer is exposed to an advertising message. Learning from experimental psychology, this way of describing attention in advertising can be elaborated and extended. It is thereby possible to increase the understanding of attention processes and how attention operates in an advertising setting. Especially when the advertising is presented in a parallel way as it is for instance on the World Wide Web, which is the case here. The parallel way of presenting information can be framed in the same way as is frequently done in research in experimental psychology. A setting where an individual, the consumer, is exposed to an advertising
31
message can be reinterpreted using an alternative terminology. The setting can be described as a task environment where the individual is engaged in some kind of activity, for instance reading or the like, when an advertisement is presented at the same time. Even though most consumers presumably do not see this situation as a task, it can be framed as a task, where there are two or more objects calling for attention. In focus is, from the Web user’s perspective, the text and pictures that the user wants to read and look at. Outside of this focus is the advertisement that is calling for attention. The advertisement is a peripheral and secondary object struggling to catch the user’s attention. The next section will focus on how attention is modeled in advertising models. The concepts that are mentioned hitherto will be further elaborated on in detail in order to get a clear understanding of how attention to advertising in Web environments is in?uenced by various factors.
Central concepts in advertising
Mainstream advertising research originates from the very end of the nineteenth century. It evolved from personal selling in general and from the AIDA model in particular, which in fact was a model of personal selling. It was adapted by researchers in advertising and became a foundation for further research within the area (Strong, 1925, p.76). According to Coolsen (1947) E. St. Elmo Lewis was one of the pioneers in advertising, publishing his book Financial Advertising already in 1908. Lewis argues in his work that good copy ought to attract attention, awaken interest and create conviction. Preston (1982) writes that Lewis actually formulated what was to become the AIDA model already back in 1898. The AIDA construct (Attention - Interest - Desire - Action) was the ?rst formal advertising model (Preston, 1982). From this model emerged the class of persuasive hierarchy models. This model and other hierarchy of effects models have had quite an in?uence in the advertising ?eld. In a study of how advertising works (Vakratsas & Ambler, 1999) the authors are using a model framing some central concepts in advertising, as can be seen below.
32
How Advertising Works
Advertising Input: Message content, media scheduling and repetition
Figure 2.4
Filters: Motivation, ability (involvement)
Cognition
Consumer: Affect Experience
Consumers behavior: Choice, consumption, loyalty, habit and so forth Figure 2.4 How advertising works (Vakratsas & Ambler, 1999).
In the model in ?gure 2.4, the starting point is the advertising input comprising of the message content i.e. the advertisement. Part of the input is the media scheduling and repetition, which together determines the reach and frequency with which the message content will ?nd its way to the consumer. Intermediate factors that affect the consumer response are ?lters such as motivation and ability (involvement) to process information. Part of the model is also cognition, affect and experience where experience represents memories of prior experiences of the advertised product or other products similar in function. Common for many studies in the advertising area is that they often focus on one or a few of the concepts presented in the model above. Also common is that authors tend to use some kind of hierarchical model even though for instance Vakratsas and Ambler claim to ?nd little support for a sequential effect in their study. Two of the early proponents of the hierarchy-of-effects model, Lavidge and Steiner (1961), were elaborating on the functions of advertising. They meant that the effects of advertising actually are long-term but if something will have a long-term effect “…something must be happening in the short run, something that will ultimately lead to eventual sales results” (Lavidge and Steiner, 1961, p. 59). Lavidge and Steiner also argued that this effect should or could be measured in order to evaluate advertising effectiveness. According to their view, advertising could be thought of as a “force”, which is guiding people through a number of stages. The ?rst stage would be potential customers who are unaware of the product or the product offer. The second stage is consumers aware of the product’s existence but is still far from buying. The next stage consists of consumers who know about the product and the offer. The fourth stage in the sequence comprise of those who like the product in question, i.e. having a favorable attitude toward the product. At the ?fth stage, people prefer the product to other alternatives. At the following stage, consumers are not only preferring the product but also convinced that this is the best option and ?nally the last stage is where all previous stages are translated
33
into a purchase. This model described is rather known to marketers as a way to determine what communication objectives that might be appropriate in different stages. Lavidge and Steiner (1961) elaborated further and meant that these stages were an indication of three main functions of advertising namely the cognitive function – the intellectual, “thinking” part, the affective function – the emotional and feeling factor and ?nally the conative function – the action oriented component. Their view was that these three components were crucial to advertising and that the aim of advertising would be to move consumers through these stages, ultimately leading to a purchase. Since Lavidge and Steiner’s article there has been a tremendous development within the advertising area and various models have been proposed, hierarchical and non-hierarchical, focusing on different aspects of advertising and taking into account some concepts while leaving out others. In the next section, a few models that are relevant to this study will be discussed and highlighted.
Models describing advertising
Not surprisingly, there is not just one model used by researchers that describes advertising. On the contrary, there are many and they tend to focus on different aspects of the advertising phenomena depending upon what the author of a particular model has chosen to focus on. The advertising models have similarities and dissimilarities and a few of these models will be looked upon in detail. Since the focal interest is the concept of attention, extra attention will be given to what the models are contributing in that particular respect.
The Classic Hierarchic Model of Advertising Effects
Ad Characteristics Stimulation Credibility Liking Personal relevance Familiarity Figure 2.5
Attention
Memory
Attitude towards the ad
Attitude towards the brand
Purchase Intention
Figure 2.5 The Classic Hierarchic Model of Advertising Effects (Attention, Memory, Attitude and Conation: A Test of the Advertising Hierarchy, Thorson, Chi and Leavitt, 1992)
The model presented above is a typical hierarchic model of advertising effects. The sequential process starts on the left in the model where the characteristics of the advertisement are presented. Stimulation refers to the advertisement’s ability to stimulate the modalities of the consumer such as vision, hearing or more uncommonly, smell, taste as well as touch. The purpose with an advertisement is to stimulate and that is also why advertisements are sometimes referred to as stimuli. Credibility relates to how credible the message is perceived. This is of course subjective since different individuals will have different opinions about whether the advertisement is credible or not. Furthermore, personal relevance, liking and familiarity are important variables that have an impact on advertising effect.
34
The next step in the model is attention, which is part of the consumer and is described as the amount of attention distributed to the advertisement by the consumer. After that comes memory, which is a measure of how well the consumer remembers the advertisement. From the advertiser’s perspective it is naturally preferable if the target group remembers the advertisement. Next, is attitude towards the ad (A.ad.) and attitude towards the brand (A.br.) leading to the ?nal stage namely, purchase intention (P.I.). Purchase intention is the conative component where the consumer acts upon the advertisement and initiates a purchase of the advertised product. The model presented above gives a quite good overview of what has been considered important concepts in advertising. When discussing advertising effects, it is often a matter of measuring variation in these constructs and their relation to each other. As we can see, the model is sequential in nature and it is thought that there is no or limited interaction or feedback between the different stages. In the literature of advertising theory, this model or derivatives of it, has been widely used (Thorson et al. 1992). Despite its wide use it has been criticized and many studies have found little or no support for its design (Thorson et al. 1992). In a study by Thorson et al (1992) it was found that advertisement characteristics did indeed affect purchase intention but the effects were both direct and partially mediated. This mediation did, however, not ?ow through attention and memory to attitudinal and conative responses. Instead, some characteristics such as credibility and liking operate through (A.ad) attitude towards the advertisement and (Abr) attitude towards the brand on purchase intention. Furthermore, it was found that attention affects neither purchase intention nor attitude towards the advertisement. Consequently, conative responses are not driven by attention. In addition, they found that the measure recall was not a good index of persuasion. Their result provided evidence that attending to an advertisement makes viewers remember it better but it had no in?uence on whether viewers would like the advertisement or brand. These conclusions seem to follow a general kind of logic since mere attention to an advertisement can hardly have an impact on liking6. Gibson (1983) has, in line with this logic, criticized the use of recall measures to evaluate advertisements since recall can only have little or no correlation with consumer choices in the market. Interesting still, is Thorson’s ?nding that none of the ?ve advertisement characteristics drive attention or memory despite the fact that they drive attitude towards the advertisement, attitude towards the brand and purchase intention. The impact of advertisement characteristics on attitude towards the ad and attitude towards the brand is an important research area that often has been debated among researchers (Brown & Stayman, 1992). This, since the purpose of advertising often is meant to in?uence attitudes. Attitude has in advertising contexts generally been described as “a learned predisposition to respond in
6
Just the fact that a viewer watches and attend to a commercial should have little effect on liking since there are numerous 35
the consistently favorable or unfavorable manner to advertising in general” (Lutz, 1985). In the classical model above attitude towards the advertisement and attitude towards the brand is, not surprisingly, part of the sequence and has also been subjected to extensive research (Brown & Stayman, 1992). Attitude towards the ad and attitude towards the brand are two measures that are being studied on the dimension favorable/unfavorable or as a composite of other belief items such as informational value, entertainment and the like (Mehta, 2000, Bauer and Greyser, 1968). Millar (1987) has studied how thought prior to an attitude assessment in?uences the correlation between attitude and behavior. In brief, it can be said that thoughts (prior to an attitude assessment) can both increase and decrease the correlation between attitude and behavior. In a study by Wilson et al (1984, in Millar, 1987) one experiment group had a correlation of 0,17 and the other test group a correlation of 0.54 for attitude correlation depending on whether they had been involved in cognitive effort or not prior to the attitude assessment. In Thorson’s study (Thorson et al, 1992) they concluded that stimulation and credibility drive attitude towards the advertisement, credibility, liking and familiarity drive attitude toward the brand and credibility and personal relevance drive purchase intention. Next, two other advertising models, which are slightly different compared to the previous model will be examined. These two models are derived from the classical model but instead of having one single route from advertisement characteristics to purchase intention they have two routes. They are also focusing on advertising from an emotional/non-emotional perspective.
The Two-Route Hierarchic Model for Emotional Commercials
Attention Ad Characteristics Stimulation Credibility Liking Personal relevance Familiarity Memory Figure 2.6a
Attitude towards the ad
Attitude towards the brand
Purchase Intention
Figure A, The Two-Route Hierarchic Model for Emotional Commercials
36
The Two-Route Hierarchic Model for Non-emotional Commercials Figure 2.6b
Attention Ad Characteristics Stimulation Credibility Liking Personal relevance Familiarity Memory
Attitude towards the ad
Attitude towards the brand
Purchase Intention
Figure B, The Two-Route Hierarchic Model for Non-emotional Commercials Figure 2.6 The Two-Route Hierarchic Model of Advertising Effects: Emotional vs. Non-emotional commercials. (Attention, Memory, Attitude and Conation: A Test of the Advertising Hierarchy, Thorson, Chi and Leavitt, 1992)
In the above ?gure, the two-route hierarchic model of advertising effects is displayed. As can be observed, these two models differ from the classical model in the sense that they have two separate routes. In the ?rst model, ?gure A, it can be observed that there is one route that is in?uenced by the advertisement characteristics and a second route that is not in?uenced by the advertisement characteristics. In the second model, ?gure B, both routes are in?uenced by the advertisement characteristics. These two models have in common the “main route” comprising of the attitude towards the advertisement, attitude towards the brand and purchase intention. The attention – memory sequence in the models is treated differently for emotional commercials compared to non-emotional commercials. The reason for this is that an advertisement that elicits emotional response in the viewer is supposed to create a memory engram that is enhanced as opposed to nonemotional conditions (Squires, 1986). It is more likely that there is a direct link between memory and attitudes for emotional advertisements than for non-emotional advertisements (Thorson, Chi and Leavitt, 1992). In the nonemotional version of the advertising model there is no link between memory and other constructs, except for its input from attention. It is rather interesting when examining these two models (and other similar models) from an attention point of view in the sense that the attention and memory constructs seem to fall outside of the model. In ?gure A, the ?rst model, the attention construct does not correlate with the advertisement characteristics and in the second model, ?gure B, the memory construct does not correlate with the attitude towards the advertisement construct. Hence, no uni?ed model can be reached. Instead, different models apply to different kinds of stimuli i.e. commercials. Here, the question can be raised whether it is possible to separate attention and memory from attitude formation in such a way as in these models above. Without any kind of attention to an object and without short-term, long-term, verbal or pictorial memory it becomes hard for the subject to have any knowledge or attitude towards a speci?c advertisement and its characteristics.
37
In the models above attention and memory have, to some extent, been described as processes not intrinsically related to the “Aad, Abr and P.I.”-sequence. At the same time, data have not supported the classic hierarchy of advertising effect model where attention and memory is part of the main sequence. Considering that attention is a construct that is inherently connected with whether input will reach the central nervous system or not, it is indeed disturbing that the attention and memory constructs have a subordinate route or sequence such as can be seen from the models A and B in ?gure 2.6. At the same time, data in some studies do not support the classic hierarchic advertising model. The reason why attention and memory are modeled aside of the main route is because data have not shown to have any correlation between attention and memory and the other constructs. Before going any further into this potential contradiction, an additional model will be introduced. A fourth model which is relevant in this context and which has been widely used, namely, The Elaboration Likelihood Model of Persuasion will be discussed.
The Elaboration Likelihood Model of Persuasion
High involment processing Cognitive Response Belief and attitude change Figure 2.7
Behavior change
Central route Communication (source, message channel) Attention and comprehension Peripheral route Low involvement processing Belief Change Behavior Change Attitude Change
Figure 2.7 The Elaboration Likelihood Model of Persuasion.
The Elaboration Likelihood Model of persuasion, ELM, has dual routes depending upon whether the subject will be engaged in a low or high involvement processing. The low involvement processing comprises of a peripheral route starting with belief change, behavior change and ?nally attitude change. The high involvement processing takes the central route comprising of cognitive responses, belief and attitude change followed by behavior change. The ELM is a model that is describing attitude formation under different levels of involvement and processing. Authors such as Karson and Korgaonkar (2001) are describing the ELM as one of the leading models for how involvement in?uences communication effects. Despite the wide use of the ELM even the inventors of the model, Petty and Cacioppo, later criticized the model since they had found that research on persuasion and involvement indicated that neither the central nor the peripheral approach alone could account for the multiplicity of attitude change
38
that has been observed (Bitner and Obermiller, 1985). Bitner and Obermiller also criticized the Elaboration Likelihood Model mainly by pointing at the following ?ve perceived problems with the model: 1) Central cues – peripheral cues which are which? 2) How does peripheral processing in?uence affect? 3) Are there differences in the strength of peripherally and centrally processed attitudes? 4) Are central and peripheral processing interactive? 5) Can the central processor make do with peripheral cues? The model has also been targeted by Vakratsas and Ambler (1999) criticizing the ELM model together with other persuasive hierarchy models (the AIDA group) chie?y because: this category of models assumes that the “[…] brain works through a series of stages, as if it were a primitive serial computer. The brain actually works in parallel” (Ambler, 2000, p. 305). And also because: “these models ignore the consumer’s experience of previous product usage” (Ambler, 2000, p. 305). Ambler is a critic of the strong theory (as the main theory) stating that the notion that advertising equals persuasion is an ingrained belief and that most advertising “does not persuade consumers to change their minds but strengthens new behavior once changed” (Ambler 2000, p. 300). Instead he is pointing at the advantages of the weak theory (reinforcement) brought forward by Ehrenberg (1974) which has, according to Ambler, received less attention, even though the weak theory is perhaps more applicable to real world situations (Ambler, 2000). Without going further into the complexities of the Elaboration Likelihood Model, we will instead shift focus and look at the Elaboration Likelihood Model of persuasion and examine what it and the other models have to say about attention and comprehension. In ?gure 2.7 we can observe that the attention and comprehension construct precedes the two routes independently of whether it is a low or high level of involvement. Intuitively, this is a more appealing situation and solution as opposed to what was seen in the two previous models. The treatment of the concept of attention in the Elaboration Likelihood Model is rather similar to how it is treated in the model in ?gure 2.5. What have been witnessed this far in the discussed models, is an attention construct that was downplayed in ?gure 2.6 A since attention was considered as having an indirect path leading to memory and attitude towards the advertisement in the end. In ?gure 2.6 B attention and memory was separated from the main route. As stated earlier, it is rather paradoxical when observing that neither attention nor memory has any impact on the other constructs in the model. Consequently, attention and memory is of limited importance in forming attitudes and a purchase intention, if this would be assumed to be correct. This feature of the model in ?gure 2.6 B does not coincide with, for instance, Mack and Rock’s (1998, p. 14) notion that there is no perception without attention. Mack and Rock are stating that if there is a
39
reduced attention, there will be a reduced perception, which in turn will lead to less attitude change or attitude reinforcement. Neurophysiologic evidence is supporting this last position suggesting that there is a reduced processing of unattended stimuli, that is to say, stimuli that a subject has not been attending to (Eysenck & Keane, 2000). Tulving and Thomson (1973) are stating “Only that can be retrieved that has been stored, and … how it can be retrieved depends on how it was stored.” (Tulving and Thomson, 1973, p. 359). Hence, recall as well as recognition will suffer if the stimulus was unattended. Taken together, it becomes problematic with hierarchical models where the attention construct is detached from the attitude formation part and does not precede the attitude formation7. As a result, it seems commonsensical to state that, ?rst of all, an input in the form of a message has to be attended to, at least at some basic level, in order to be perceived. Moreover, the message or parts of it has to be placed in memory in the form of, for instance, an attitude towards an advertisement and an attitude towards a brand. If there is no kind of memory of this (unattended) stimulus, it becomes dif?cult to comprehend how the consumer will be able to recall or recognize a brand. A model that does not include attention in the initiation of a communicative process does not follow the logic of attention to a great extent and it becomes even more problematic when memory is set aside. When studying the two models in ?gure 2.5 and 2.7 it is apparent that these models do have a serial perspective and an attention construct placed in a serial route. The attention construct is placed before memory, as in the case in the classical hierarchical model, or before the central and peripheral routes as in the case with the Elaboration Likelihood Model. From these two models it can be deduced that the process stops if there is no attention. This is logical. When recapitulating the above discussions and in particular when taking into account Belk’s notion, there are factors in the surrounding, the environment or in the situation that has the potential to affect the communication and the attention to the communication. Since there are no constructs in the classical hierarchical model or the Elaboration Likelihood Model that are considering this, a general model where this construct can be found will be discussed.
It is at this point also relevant to mention that an attitude comprises of the cognitive, affective and cunative components (Tesser and Schaffer, 1990) and is a predisposition that is expressed by evaluating for instance a product with some degree of favor or disfavor. Furthermore attitudes are considered as being rather stable over time.
7
40
Summary Model of Web Ad Effects
Ad Processing Attention Learnning Emotions Acceptance
Figure 2.8
Ad attitude
Action 1. Brand inquiry 2. Brand purchase
Exposure (intended, content & Structure Content 1. Content nodes 2. Content node enhancers Structure 1. Micro structure (a) Links (b) Link Effectors (c) Molecules 2. Macro structure (a) No. of molecules (b) No. of links between molecules (c) Intended order (d) Navigation aid
Web and schema
Moderators 1. Web navigation ability 2. Category need 3. category expertis 4. Situational factors
Communication Effects Identification 1. category needs 2. Brand awarness Evaluation 3. Brand attribute 4. Brand attitude 5. Brand intention 6. Brand purchase facilitation
Figure 2.8 Summary model of web ad effects. (From John R. Rossiter and Steven Bellman, 1999)
The model in ?gure 2.8 is indeed detailed and takes into consideration additional aspects of the advertising phenomena, as opposed to the previously discussed models. The model is in fact an adaptation of a general advertising model that Rossiter and Percy presented in 1997 (1997). This model is dissimilar to the previously studied models in the sense that it is speci?cally designed to model Web advertising. Rossiter and Bellman are stating in their article that Web advertising has different characteristics than other kinds of advertising. Mainly, since the consumer is “actively skipping about searching for interesting items of information, often pausing to be entertained and perhaps missing or deliberately bypassing content that the advertiser thinks is important.”(Rossiter and Bellman, 1999, p. 15) This is an exquisite observation capturing a general picture of Web behavior among Web surfers. What they are describing in essence, is the attentional system in action, which is constantly engaged in selecting, including and excluding information when guiding through the environment – the Web site, where advertisers are competing for attention. This behavior is not unique for the World Wide Web. In fact it is very similar8 to the behavior when “sur?ng” through print material, such as newspapers, magazines and the like. In the model by Rossiter and Bellman above, the situational factors are taken into consideration. The situational factors are moderators affecting the Web ad schema and the advertisement processing where the concept of attention can be identi?ed. Even though the model is mentioning these constructs it is not speci?cally targeting the interrelation between attention and context represented as a task environment.
8
There are of course differences in the interactive possibilities though. 41
Moreover, it is not stressed in any of the other models either. Rossiter and Bellman’s model is the only one, of those examined here9, where situational factors are taken into account. Following the ideas of Belk, situational factors may have an impact on the process of attending to different stimuli and ought to be examined further. Prior to elaborating on the relation between attention and situational factors, it is important to discuss around advertising and the World Wide Web.
The World Wide Web as an advertising vehicle
The development of marketing communication on the Internet has been rapid, driven by a variety of companies exploring the possibilities on the Web. According to Hoffman and Novak (1997b) the Internet is a new “marketscape” that calls for a new marketing paradigm. This, since consumers are more active on the Internet than is the case with other media (Hoffman and Novak, 1996). That a new marketing paradigm is needed as a consequence of the emerging use of the Internet and the World Wide Web is perhaps to stretch matters too far. Nevertheless, the development of the Web poses a challenge to companies in a variety of industries and in particular to organizational functions related to marketing communication and advertising. An opposite stance that some are taking, compared to for instance Hoffman and Novak, is that the Internet and the World Wide Web is not very unique; it is just another communication channel among a number of channels. This stance is, on the other hand, perhaps not quite acknowledging the versatile nature of the Web or the radical impact that it has had in most parts of society. With the Internet, companies have gained a new platform to communicate to and with their target audience. This is possible through various means such as e-mail, chats, instant messaging, discussion groups, news groups, on-line gaming, hypertext links, short message services through gateways to GSM networks, ip-phone communication and the like. One of the most important and current possibilities with the Internet is to use it for advertising on Web sites and Web pages. Advertising on the World Wide Web has increased the last few years giving new opportunities for marketers to reach their customers. Advertising can be conducted in different ways on the Web ranging from static banners and advertisements, click through banners, traditional as well as interactive commercials “broadcasted” with streaming video, pop up advertisements (interstitials), triggered advertisements, or a combination of two or more of these advertising tools. The variation and combination of different kinds of advertisements are greater on the World Wide Web in comparison to any other type of media.
9
Rossiter and Bellman’s model is not the only model in marketing communication and advertising that have situational factors present, see for instance Belk. 42
Advertising on the Web has both differences as well as similarities when comparing to established advertising media such as TV, radio or print. A salient property distinguishing Web advertising from other kinds of advertising is that it has hybrid characteristics combining properties from print, broadcast, outdoor, and direct response media (Dréze and Zufryden, 1997). As a whole, these properties give Web advertising designers room for creativity and development of new combined types of advertising. However, advertising on the Web can be distinguished from other broadcasting media (television & radio) since the advertising message is integrated within the editorial material in the same manner as in print media (Chandon, Chtourou and Fortin, 2003). In television and radio, advertising messages and program content is presented sequentially. That is, program content is presented in a separated block which is interrupted by a block of commercials and subsequently a new block of program content is followed by a block of commercials and so on. Thus, the commercial block creates a “pause” in the program content that can be used for various activities10 by the consumers. The pause represents a possibility to divert the attention from the media and thereby the advertising as well. On the Web, advertisements are competing with the Web page content for the consumers’ attention in a parallel way. The consumers’ attention is thereby continuously directed to an area close to the presented advertising material. In the case of parallel presentation Web advertising is similar to print where advertisements are presented together with program content or the editorial environment. Nonetheless, the dissimilarities between Web and print that are often pointed at are most of all that the Web has a greater interactivity, greater complexity and a high level of ?exibility (Bruner & Kumar, 2000, Karson and Korgaonkar, 2001). There are of course differences in the visual appearance as well.
Web advertisements are placed on the World Wide Web
How to place and where to place a banner or an advertisement on a Web page has been studied by many researchers (see for instance Bruner; Shamdasani, Stanaland and Tan; Dréze and Zufryden; Novak, Hoffman and Yung) with the aim to obtain an understanding of how to gain the best effect out of an advertisement. The complexity of a Website has been a concern since prior research has suggested that increased complexity could reduce the effect of an advertisement (Stevenson, Bruner, and Kumar, 2000). Since the developments in Web design has been racing towards greater complexity with competing sights and sounds found on Web pages (Bruner G., 2000) it is crucial to understand how this affects advertising. One of the main reasons why Web design is heading towards greater complexity is because of the possibility to transfer larger amounts of data per time unit.
10
To fetch some snacks, go to the restroom, watch another program or the like. 43
This development of hardware infrastructure speeding up data networks has given Web designers new opportunities to improve Web design as well as Web advertisements. As a result, advertising on the Web today offers greater possibilities than it did just a few years ago. Increased complexity on Web sites (Bruner G., 2000) is created by using more graphic images, more text (even blinking or the like), more hyperlinks, more sound, more animation and also more full-motion video. The added complexity on Web sites makes them more attractive and enhances the perceived entertainment value and exclusivity. Still, as Web designers add these features to attract users, some studies from psychology and advertising suggest that interest may grow with complexity but not necessarily pleasantness. Already early research (Berlyne, 1960) has shown that pleasantness is optimal at some low to moderate level of complexity. Thus, the new way of designing and presenting Web sites may have negative consequences for the users’ perception and attitude towards the Web site (Bruner G., 2000). Berlyne (1960), was furthermore stating that people, in general, are more motivated to explore complex stimuli when they appear in a setting of simpler stimuli which is in line with the idea that increased complexity may affect the impact of advertising in a negative way. The Web sites and their various environments that is referred to here is to some extent a mirror image of the editorial environment which is the term often used for print media (Appel, 2000). Advertisements in print media are placed into the editorial environment and the editorial material surrounding an advertisement has been considered important as a factor in?uencing the effectiveness of the advertisement. Appel has elaborated around this issue and he is arguing that there is a strong relationship between editorial environment and advertising effectiveness. In Appel’s article he is arguing that when the credibility of editorial material in a magazine was low, it also affected the effectiveness of an advertisement in a negative way. When the editorial environment on the other hand was credible, the advertisement also scored higher on perceived credibility. Appel attributed this effect to composition of audience partly but mainly to the editorial environment. It is plausible that this “editorial environment” effect that Appel is discussing around is present in a Web setting as well as in print. In line with Appel’s ?ndings, there are studies showing that the same source delivering the same message to the same audience can produce different effects depending on the editorial context in which the message was presented (Chaiken and Stangor, 1987; Cooper and Croyle, 1984; Collins, and Miller, 1969). Context has also been shown to be important since it can affect recipients’ judgments of advertised products (Puto, 1987; Smith 1996; Woodside and Singer, 1994) Shamdasani et al (2001) has studied the vehicle source as a variable affecting the effects of an ad. They are stating that “source variables … essentially moderate
44
the impact of a persuasive message by in?uencing the audience’s con?dence in the message endorsed by the source” (Shamdasani et al 2001, p. 8). There are in other words variables that are connected to the source and not the message that affect the impact of the message. Source credibility is important since when credibility is high, consumers counter-argue less with advertising claims and are therefore more in?uenced by the message (Grewal, Gotlieb, and Marmorstein, 1994). When source credibility is low, conversely, consumers discount the arguments made in the message. Thus, messages presented by a highly credible source are more readily accepted and induce greater attitude change in consumers than a source with low credibility (Harmon and Coney, 1982; Sternthal, Dholakia, and Leavitt, 1978). In a study of Web advertising recall and recognition (Mullarkey and Danaher, 2003) the authors found that factors such as duration of page viewing, Web page context factors and viewing mode to be crucial and part of the determinants of advertising effect. The impact of the duration of Web page viewing is consistent with studies conducted of televised advertising. Researchers (Krugman D., Cameron and White, 1995, and Krugman H., 1986) showed that the more time spent viewing and attending to an advertising media, the more advertising content respondents tended to remember. The connection between attention to advertising and viewing duration is tied to the information processing that the consumer is occupied with. A greater opportunity (reinforced by time) to attend to an advertisement and to process information result in a positive effect on recognition. This is also being supported by studies using neuroimaging (Rossiter, Silberstein, Harris and Nield, 2001). The viewing mode in the study by Mullarkey and Danaher (2003) is described in a similar manner as in the study to be conducted here. Mullarkey and Danaher considered it reasonable to divide Web surfers into two categories representing two different modes, i.e. “goal directed” and “sur?ng”. Their terminology is brought in to the advertising area by Janiszewski (1998) who is arguing that there are two broad types of behavior namely, goal directed search and exploratory search. Goal directed visual search behavior has been shown to reduce attention to peripheral stimuli, whereas exploratory search behavior gives peripheral stimuli a chance to be processed. Mullarkey and Danaher argue in line with Janiszewski’s ?ndings stating that when consumer viewing or use is directed toward a particular task there would be a reduction of the impact of advertising. A less goal directed and exploratory sur?ng behavior would on the other hand result in more browsing of media content, including peripheral content such as advertisements. However, a limitation of their study was nonetheless that the respondents never engaged in real Web sur?ng activities since the experiments were designed, and limited, in such a way that the respondents merely watched Websites that were automatically switched every 20, 40 or 60 seconds. Thus, the real hands-on sur?ng never occurred in that study. Since there was no hands-on sur?ng there were no interaction, clicking or sur?ng either which are the hallmarks of Web – user interaction.
45
Web page context factors were also addressed in the Mullarkey and Danaher’s study and in line with other authors (see for instance Bruner G., 2000) the context factor was described as the design and layout of the Web site. The main context effect taken into consideration was how design of the Web page and how competing stimuli “…and potential page cluttering or excessive complexity could affect reception of advertising” (Mullarkey and Danaher, p. 255, 2003). Similar studies of Web context effects have been focusing on Web page background complexity (Bruner and Kumar, 2000, Stevenson, Bruner and Kumar 2000), and also the Web page complexity’s impact on page access (Dreze and Zufryden, 1997). In summary it is adequate to say that Web advertisements are placed in a Web setting. The Web setting is essentially very similar to its print counter part “the editorial environment”. The general development on Web sites the last few years has been towards greater complexity and more competing stimuli resulting in a developed Web setting. The added complexity on Web sites makes them more attractive and enhances the perceived entertainment value and exclusivity. This is an evolution that can on the one hand increase pleasantness but on the other hand reduce overview and access. Simultaneously, being a context variable, complexity is affecting the task environment moderating how attention is distributed towards advertisements. Thus, changes in complexity and environment have occurred and are occurring that may have disadvantageous in?uence on advertising.
The advertisement
Another obvious factor in?uencing the total advertising effect is naturally the advertisement itself. Hitherto, we have not been dedicating much attention to the advertisement itself other than factors surrounding it and affecting the advertising effect indirectly. Next, focus will be on the advertisement and the properties of an advertisement that affect attention, among other effects that can be measured. One of the main principles with advertising is to attract attention (Pieters & Wedel, 2004), in order to get the opportunity to present the company offer to the consumer. However, it is becoming increasingly dif?cult to attract attention because of increased clutter and noise. For advertisers the quest to reach customers is more demanding today when, for instance, half of the material in a typical magazine consists of advertisements. Consequently, the inability to capture consumers’ attention has become more important and prioritized (Adler & Firestone, 1997; Davenport & Beck 2001; Pieters & Wedel, 2004). Some are even claiming that “the power of marketing is eroding … from lack of attention” (Sacharin 2001, p 3. in Pieters & Wedel, 2004). The pursuit to attract attention is a complex one and has many paths. Part of it is the message that is being transferred to the consumer and in particular the properties of the advertisement. That size plays a major role and has a positive correlation with attention is a general notion in advertising (Finn,
46
1988; Abernethy & Laband, 2004) and has recently also been supported by research on Web site advertising (Baltas, 2003). Except for size, there are many other characteristics of an advertisement that have been studied by researchers. In the work of Finn, other characteristics (for print ads) have been shown as in the ?gure below.
Advertisement Characteristics
Size and location Ad size Cover position Facing: ad / editorial Right / left page Layout and pictorial Color Illustration size Photo art Bleed / no bleed Copy Amount Readability Benefits Other Characteristics Headline: Words Phrases Nouns Verbs Adjectives Determiners Type size Product reference Question form Benefit Product as object Figure 2.9
Figure 2.9 Advertisement characteristics. Adapted from Finn (1988).
Of the characteristics listed in ?gure 2.9, not unexpectedly, advertisement size was in fact the property that had shown a signi?cant effect most frequently in the material that Finn reviewed. However, aside of the characteristics stated above later researchers have pointed at a few central characteristics of an advertisement that are important to take into consideration. According to Pieters and Wedel (2004) there are three key elements that can be identi?ed in an advertisement, the brand, the picture and the text. Their position is that these three elements have distinct effects on attention. In Pieters’ and Wedel’s study they found that the pictorial was superior in attracting attention, independently of its size. Other researchers are also reporting that the pictorial is by far the “most important structural element in magazine advertising” (Rossiter & Percy, p. 295, 1997) and studies (with eye movement tracker) are in fact showing that around 90% of the viewers ?xate the main picture in an advertisement before they start focusing on the copy text (Kroeber-Riel, 1984). The enduring effect after viewing the advertisement is the formation of a visual memory of the advertisement that enables subsequent recognition (Finn, 1988). These ?ndings presented here should lead us to convert to the standpoint that advertisements should seldom be designed without adding a pictorial element to it. In that case (without picture) we would suffer the bene?t of base line attention to the pictorial element. Furthermore, the pictorial element is also supposed to lead the reader or viewer to the other elements of the advertisement. The purpose is not only to attract attention but to propel readers to engage in reading the brand and text elements.
47
Not only the pictorial elements but also the text elements are considered as being important and even assumed to be able to capture attention. Text elements are usually divided into two main categories namely, the headline and the copy text. The headline is supposed to be an eye catcher capturing the readers’ attention and being interesting enough to get the reader to continue with the copy text in the advertisement. One of the proponents of the importance of headlines is David Ogilvy. One of the more well-known statements in advertising originates actually from Ogilvy saying that “The wickedest of all sins is to run an advertisement without a headline” (Ogilvy, 1963, p. 130, author’s italics). Some advertisers hold the view that the headline is often perceived as the most vital advertisement component (Pieters & Wedel, 2004). This position among practitioners is however challenged by researchers such as Kroeber-Riel who is pointing out that respondents have a preference for attending to pictorial elements and that they in fact avoided, if possible, “the cognitive effort of reading the verbal elements when skimming ads” (Kroeber-Riel, 1984, p. 593). The text elements are though important in that they can communicate speci?c information about the product offer that is more complex and that cannot be conveyed by the pictorial element. The ?nal element, the brand, comprises of the brand name, trademark and logo (Pieters & Wedel, 2004). The brand name is a potential carrier of image, value and status communicating this information in a subtle way to the audience. There are scholars arguing both for maximizing and minimizing the brand size in an advertisement. Those in favor for a brand that has a prominent position in an advertisement argue that the brand is important to remember and also that it should be easily identi?ed. If the advertisement stands out in comparison to competitors it will attract more attention and generate favorable brand communication effects (Pieters & Wedel, 2004). These properties as a whole will for instance help the consumer recognize and choose the brand when entering a buying situation. Those in favor of a less salient brand placement in an advertisement claim that the brand element signals that the message is an advertisement and since the consumers are not interested in advertising it will be counter-productive. Hence the brand strategist or the advertiser ought to downplay the size and location of the brand in the advertisement. It is worth to notice that it is commonly assumed that consumers are not interested in any advertising whatsoever. As a matter of fact it is assumed that consumers “hate” advertising (Aitchinson, 1999 p. 61. in Pieters & Wedel, 2004). This is a rather extreme position. In a print advertisement context and potentially also in a Web context it is important to take these three key elements (the brand, the picture and the text elements mentioned previously) into consideration when designing an advertisement. Therefore, the three elements discussed have been taken into consideration when developing and designing the experimental advertisement so that the experimental advertisements have these generic elements. However, these elements will not be manipulated since they are not unique to Web
48
advertisements. Instead it is the speci?c Web dimension of advertisements that will be altered. In Web advertising there are additional properties than the brand, the pictorial and the text that may generate increased/decreased attention to the advertisement. A Web advertisement can have properties that reach beyond the list presented in ?gure 2.9, properties that are, to some extent, unique for Web advertising. It is these properties, animation and motion together with abrupt appearance that will be manipulated. In the following it will be elaborated around this extra dimension that Web advertisements have compared to for instance print advertisements.
The Web dimension of advertisements
When studying different World Wide Web advertisements and what various authors have written about Web advertising11 one will soon discover a diversity of types of advertisements and also a general lack of classi?cation. Most authors are referring to and studying the banner advertisement, hitherto few are describing its properties12. Yet, development of the WWW has brought new advertisement tools, which in turn has made it evident that these tools have to be classi?ed. Classi?cations have however been poor or at least not up to date with the rapid development in terms of advertisement design and also development of information technology infrastructure, making it possible with larger and more demanding advertisements on the Web. Despite the rapid development in some parts of the world, researchers are still referring to the banner advertisements as, “…the most common form of advertising currently” (Danaher and Mullarkey, 2003, p. 253), which is a small billboard-like graphic that appears on a Website that is clickable (Hoffman and Novak, 2000), and furthermore, “…is a rectangular shaped image typically located at the top of a Web page” (Chandon, Chtourou and Fortin, 2003, p.217).
11
There is a wide variety of advertisements and advertisement formats that is being used on the Web. The more frequently used advertisements will be discussed with a main focus on those types of advertisements that are the most relevant for this study. 12 A reason why few authors bother de?ning web advertisements might be the fact that the banner advertisement was the ?rst kind of advertisement that was used on a larger scale and they are now taken for granted. However, development of the WWW has brought new advertisement tools making it inappropriate to classify all web advertisements as banner advertisements. Banners were the ?rst commercial form of online advertising and were created by the staff of the magazine ‘Wired’. Banners quickly became adopted as an advertising standard generating revenues for publishing and content sites (www.rightonthespot.com). 49
Web Advertisements
Advertisement in the shape of a small square
Figure 2.10 Avertisements in the shape of traditionional banners Advertisements with a normal rectangular shape
Figure 2.10 Web advertisements in different sizes and shapes.
The notion that a banner advertisement is a rectangular, small billboard-like graphic and also the most common form of advertising currently, does not ?t well with the kind of advertisements that is frequently being used on many north-European Web pages as well as international ones. Pilot study number two pointed in the direction that Web advertisements nowadays have little or no resemblance of their banner counterparts13 from the mid-nineties. The one-colored, static banners that were being used at that time are infrequently used on today’s Web pages. Since modern Web advertisements are becoming increasingly similar to print advertisements as opposed to banners, one could argue that it is about time to rename these modern Web advertisements. In addition, when taking into consideration the fact that banner advertisements have the purpose to evoke click-through at least for high-involvement and functional products (Dahlén, Ekborn, and Mörner, 2000; Dahlén and Bergendahl, 2001; Dahlén, 2002) and that the click-through rates have fallen to below 1 ‰ (or 1 in 1000), it seems as if banner advertisement’s prime time has reached its peak. Still, there are banner advertisements that are being used and that will be used although new sizes and formats are also appearing. Even though many researchers are still holding on to the banner advertisement, the Interactive Advertising Bureau - IAB, is leading the way forward and have presented a new system for categorization of advertisements, which include more updated formats14.
13
The ?rst banners employed on the WWW were often one colored with just a simple text with the brand name on. 14 The IAB is working to “…familiarize publishers, marketers, and agencies with Interactive advertising standards and guidelines formulated and developed by their peers in association with the IAB. Use of these standards and guidelines directly organize the industry so as to foster an environment in which the Interactive medium is implemented with ease, thus allowing the industry to buy more and capture value.” (IAB Standards and Guidelines 2004). 50
The Interactive Advertising Bureau have developed general standards, classifying advertisements into two broad categories, the “In-page” advertisements and the “Out-of” page advertisements. An In-page advertisement is a Web advertisement presented directly on the Web page itself15. An Out-of page advertisement is an advertisement that is not presented directly on the Web page. An Out-of page advertisement could, for example, take the form of a popup advertisement that will appear as soon as one has clicked on a launching link or when entering a Web page or the like. Out-of page advertisements have furthermore been subcategorized into “Over the page advertisements” (pop-ups), “Between page advertisements” (transitional advertisements) and “In-stream advertisements” (Streaming video commercials, see Appendix E). The Out-of page advertisements have in common that they are advertisements presented in a separate window which is separated from the actual Web page that the user is using and visiting. In-page advertisements can have different sizes, shapes and any property that can be presented on a Webpage. It is usually presented in conjunction with program content on the Web site and is thereby competing with the program content for the user’s attention (Chandon, Chtourou and Fortin, 2003). In-page advertisements are either static advertisements or animated advertisements. The use of static Web advertisements are declining and in Chandon’s study only 7.2 % of over 1200 advertisements studied were not animated. Their study, that measured click-through rates, indicated that animation is bene?cial to advertisers since it appears to improve click-through rates. Kisielius and Sternthal (1984) found that vivid information attract user’s attention easier due to its sensory effects. On the Web this means that animation may increase the likeliness to attend to the advertisement (Zhou & Bao, 2002). Another study conducted by Lohtia, Donthu and Hershberger (2003) indicated that the animated advertisements increased the click-through frequency and the authors are strongly recommending the use of animation. In a study by Sundar and Kalyanaraman (2004) arousal and memory were studied as a function of animation speed. Sundar and Kalyanaraman’s ?ndings are pointing in the same direction. That is, animation has an impact on measures such as arousal and recall with an increase when animation is speeded up. Their study deviates from the former studies since it is not based on click-through frequencies. Instead, their study used a physical measure, namely arousal which is measured via electrodes during the session, and also recall which is a measure often employed in advertising research. In these reported studies, recognition has not been used as a main measure, a measure that for instance Krugman is advocating when measuring advertising effects (Krugman, 1986). An out-of-page advertisement is a group of advertisements comprising of “overthe-page” advertisements, “between-page” advertisements and “in-stream” advertisements. In-stream advertisements and between-page advertisements, on the other hand, will not be covered in this study.
15
A banner advertisement would of course belong to the category In-page ads. 51
An over-the-page advertisement is a category of advertisements with the characteristic that the advertisement will appear over the Web page that is being viewed by the user. The main advertisements in this category are the well-known pop-up advertisements (also called interstitials) and also the ?oating advertisements. Floating advertisements are advertisements that appear when one enters a Web page, and the advertisements “?oat” or “hover” over the page for a short period of time, typically 5-10 seconds. Depending upon the attention getting strategy chosen, the ?oating advertisement might be placed either in a peripheral position or in a more central position obscuring the view of the Web page (Pilot study 2). The second type of advertisements belonging to the over-the-page category is the pop-up advertisements. The characteristics of the pop-up advertisements are that they “[…] utilize a Web-browser initiated additional window to deliver an advertisement impression either directly above or below the existing browser experience”16 (IAB report, 2004). The pop-up advertisements have two features that distinguish them from their static counterparts, namely some degree of surprise since they tend to appear from nowhere when the Web user enters a Web page or clicks on a certain hyperlink or some similar event. These features contribute to enhance the attention getting effect. The pop-up advertisement also adds motion or visual disturbance to the Web page area where it pops up. This is equivalent to the effect of animation even though the pop-up itself might not even be animated. According to Beard (2001), the pop-under advertisement has become widely used in the industry. The reason for this is that the pop-under is a salient piece of stimulus since it appears on the screen in an unexpected manner and without the user’s consent. In doing so, it is more effective compared to traditional banner advertisements. The attention getting features of the pop-under advertisements are equally relevant for pop-up advertisements as they appear on the screen in an unexpected manner and without the user’s permission. One of very few systematic studies on pop-up advertisements, by Diao and Sundar (2004) found that pop-up advertisements were superior in achieving ad-recall. At the same time they did not observe this effect for ad-recognition, contrary to their own hypothesis. On the other hand, they used a somewhat different methodology in comparison to this study. They used for instance heart rate in beats per minute to measure the response of the respondents. Furthermore, they relied on data that could be confounded. For example, they did not take into consideration differences in size between advertisements that were compared. Such a procedure is risking the validity of the results.
16
When presented under the existing browser it is usually called a pop-under advertisement.
52
Research on basic forms of animation found that moving or ?ashing objects could be useful in attracting attention from one area of the screen to another as a consequence of its visual distinctiveness (Cropper and Evans, 1968, Smith and Goodwin, 1971). An alarming negative ?nding regarding animation was that it could be “distracting”, “obtrusive”, “disruptive” and “fatiguing” (McCormick 1970, Stewart 1976). Later research has, to some extent, supported these ?ndings stating that the effect of animation at times can be “annoying”, “irritating” and even “evil” (McGalliard 1998, Nielsen 1996, 1997, Spool et al. 1999). Hong, Thong and Tam (2004) have studied the phenomena of animation from an information systems approach and found that ?ash indeed is a salient feature that has the potential to capture attention. A negative aspect (from the perspective of an advertiser or a Web site publisher) is that this attention getting technique can, under certain conditions, lower users attitudes’ towards using the Web site in question (Hong, Thong and Tam, 2004). Another problem less focused upon is the aggregate effect of attention getting techniques on the users of the World Wide Web as a whole. The more attention getting techniques used, the more mental effort that will be used to avoid advertisements, and in particular advertisements that are non-relevant from the user’s perspective. One should not simply assume that there are no countermeasures against attention getting techniques. The users’ attention system will use more resources to screen out advertisements. This could then result in what Cho and Cheon (2004) call “Banner Blindness”, which is a negative development for Web advertising in general. Regarding the pop-up advertisement, which is one of the most attention getting advertisements, there is the possibility to close them by clicking on the small “x” in the upper right corner on the pop-up window. According to one study, a majority of Web users are closing down the pop-up window within 3 seconds (Eyetrack, 2004). Another strategy is to not close the popup down but rather to avoid looking at it and after about 25 seconds to click somewhere else, which will cause the pop-up to disappear (Eyetrack, 2004). The pop-up advertisements are not only becoming a victim to Web users that are clicking them away. There are other initiatives taken by organizations where the purpose is to avoid advertisements (in this case pop-ups). Software producers developing Web browsers have sometimes added functions that are screening out pop-up advertisements. In Windows XP, for instance, the pop-up suppressing function is on, in default mode. This has of course been counter balanced by new versions of pop-ups that are passing these ?lters. It was indicated earlier that practitioners are in favour of animated advertisements compared to static or non-animated advertisements. The non-animated advertisements’ share of all Web placed advertisements have decreased and was about 7,2 % in 2003 when Chandon et al conducted their study. It is believed that the static advertisements are less effective as opposed to animated in-page advertisements and pop-up advertisements. The pop-up
53
advertisement is in fact a special case of an animated advertisement (no matter whether the content in the pop-up window is animated or not) since it “uses” motion in the area where it appears, but is different compared to its in-page counterpart as it can be moved around or clicked away. Advertisers are eager to use attention getting techniques to attract users to their advertisements yet there are risks connected with this as well. It seems that the advertiser has to walk the ?ne line between getting attention and evoking fury, at worst. In summary, Web advertisements have additional features, compared to for instance print, that can be used to attract attention. These attention getting techniques brought up here is: a static advertisement – where no technique is being used, an animated advertisement – where some kind of motion picture is being used and ?nally pop-up – where the advertisement is being abruptly presented somewhere on the Web browser creating disturbance and motion. The purpose with these attention getting techniques is to attract the attention of the audience. In the previous it has been discussed around the attention construct to quite some extent without going very deep into the area. Next, a deeper review of the attention construct will be done in order to gain further insight into this imperative aspect of advertising.
The concept of Attention
The concept of attention is central to advertising. In order for the consumer to perceive an advertising message she will have to distribute at least some attention to it. Otherwise the message will go by unnoticed; an unwished event by most advertisers. Therefore, attention is important for advertisers and their aim is that the audience a ttends to the commercial messages that they are communicating. Thus, advertisers commonly use attention getting techniques such as animation, pop-up, ?ash or other means in order to attract attention. With this said it is appropriate to examine further what attention is and how attention can be related to advertising. What is attention? And how is attention de?ned and described? William James once wrote:
“Everyone knows what attention is. It is the taking possession of the mind, in clear and vivid form, of one out of what seem several simultaneously possible objects or trains of thought.Focalization, concentration, of consciousness are of its essence.”(William James, 1890, pp. 403 – 404 in Eysenck and Keane, 2000).
James description of attention is a rather early one nevertheless like others of more recent date he is pointing at selection as a core feature and also at the bringing into consciousness. Mack and Rock (1998, p. 25) are de?ning attention as a “process that brings a stimulus into consciousness. It is in other words, the process that permits us to notice something.” This view and description of
54
attention is the one used further in this study and is also in line with the notion that has been expressed when discussing the advertising models previously. A stimulus can for instance be an advertisement or any other event that can be brought into consciousness. The fact that Mack and Rock are stating “the process that permits us to notice something” implicates that a reduction of the process of attention will at the same time also reduce what is being noticed and brought into consciousness. The major conclusion of Mack and Rock’s very extensive research in the area of attention is that there is no perception without attention. Perception in turn is the means “by which information acquired via the sense organs is transformed into experiences of objects, events, sounds, tastes etc” (I. Roth 1986 in Eysenck and Keane, 2000). In order for information to enter the internal system of a subject this individual will ?rst have to distribute attention to the information so that it can be perceived and thereby be further processed. From an advertising perspective this would clearly mean that subjects ought to attend to an advertisement, at least at some minimal level, if we are to expect them to be able to recognize an advertisement or product at a later stage. Consequently, we have to capture the audience attention so that they can perceive what we are communicating. However, what means do we use to capture the audience’ attention? Earlier in this chapter it was highlighted that advertising researchers have found that advertisement properties such as picture and headline are said to capture attention. The pictorial in particular seemed to be an important means in attracting attention and Kroeber-Riel’s study indicated that subjects actually tried to avoid the text part in favor for the picture part. It was also discussed that size had been reported to have a positive correlation with the capture of attention. Furthermore, motion, ?icker and ?ash have also been identi?ed as properties that can capture a subject’s attention. The properties of objects that can capture attention have been studied extensively within the theoretical realm of visual search. In the area of visual search quite some research has been conducted in regards to investigate what a basic feature is and how humans are performing visual search tasks. A stimulus that supports both effortless texture segmentation and ef?cient search is usually perceived as being a basic feature. Effortless texture segmentation can, for instance, be described as when a region of green spots in a ?eld of red spots is instantly segmented from the background or when spots at one stereoscopic depth among spots at another stereoscopic depth are immediately identi?ed (Wolfe, 1999). Ef?cient search, on the other hand, is when an increase in set size (set size is the number of distracting objects in a visual search task) does not produce an increase in reaction time. When looking closer at the various stimuli and their properties it can be observed that color is a typical example of a basic feature and is one of the best ways to make a stimulus “pop out” (Bundesen & Pedersen, 1983).
55
Generally, a greater difference between colors will make it easier to discriminate between targets and distracters (Nagy and Sanchez, 1990). Orientation is another basic feature and refers to, for instance, the orientation of multiple lines, where one line deviates in orientation from the other lines (Foster and Ward, 1991a). Curvature and vernier offsets are also considered as basic properties and refer to a curve among straight lines and a broken line among straight lines respectively (Treisman and Gormican, 1988; Levi et al., 1985). Size, is yet another basic feature that has been extensively studied in various ways and shown to possess the properties to be easily identi?ed (Bilsky, Wolfe and Friedman-Hill, 1994). Motion, is an uncontroversial basic feature as it is easy to ?nd a moving stimulus among static ones whereas the opposite, to ?nd a static among moving spots, is more dif?cult (Dick, Ullman and Sagi, 1987). Other basic features are depth and gloss since both produce ef?cient search times (Enns and Rensink, 1991; Nakayama and Silverman, 1986; Wolfe and Franzel, 1988). These basic features can be used to guide the subsequent deployment of attention (Pashler, 1998). The deployment of attention can in turn be stimulus driven (also called bottom-up) or user driven (top-down). Stimulus driven, means basically, that the external target is suf?ciently different from its distracters so that the stimuli “pops-out”. User driven, means, that we deploy the attention to a stimuli that we have decided being worthy of our attention (Pashler, 1998). Information from top-down and bottom up analysis of stimuli is used for attentional prioritization where attention will be directed to the item with highest priority (Wolfe, 1994). Braun has studied the bottom-up and top-down distinction by designing an experiment with a demanding task at ?xation to see what attributes of a peripheral stimulus could be identi?ed. Many of the basic features described above could withstand this test (Braun 1993; Braun 1994; Braun and Sagi, 1990). This is taken as evidence that attention can be seen as active for the central task while it is passive for any peripheral task (Braun and Julesz, 1994). It could also be argued that this means that preattentive processing continues across the visual ?eld while attention is busy somewhere else (Wolfe, 1998). The basic features have also been of interest in regards to attention capture and the properties that attract attention have been under scrutiny in experimental psychology and debated elaborately (see for instance Yantis, 1993 a & b, Yantis and Jonides, 1996, Gibson, 1996, Turatto, Galfano, Gardini and Mascetti 2004, Franconeri, Hollingworth and Simons, 2005). The perspective has been somewhat different and much of the interest has been directed at understanding basic properties and their in?uence on attention. Some of these basic features that have been studied are color, numerosity, motion, ?icker, shape, location and texture segregation (Mack and Rock, 1998). An early notion was that some of these properties were pre-attentive, perceived without attention, but was reinterpreted to mean that “numerosity, location, color and motion were properties that could capture attention” (Mack and Rock, 1998, p. 14). However, later research with more sophisticated methods, under an
56
inattentional paradigm, showed that not even these properties were able to capture attention. Instead, Mack and Rock argued that among a long array of features it seemed as meaningfulness and size were the properties that could capture attention. Other researchers in the area like Yantis (1993a) and Yantis and Jonides (1996), have argued that what is really capturing attention is the appearance of a new visual object in the visual ?eld. What is already in the visual ?eld will be part of the stable visual setting and only new visual objects will have the capacity to capture attention. Yantis and Jonides are however stating that abrupt visual onsets may not be unique in capturing attention. When studying the experimental research done in psychology, it is important to remember that the experiments are conducted with visual searches for stimuli with basic features. Yet, in the real world most people are rarely looking for a “T” among “X’s” or a “red dot” among “orange dots”. Moreover, we also have to keep in mind that these studies that have been referred to often use very short duration times of the stimulus presentation. Typical duration times are well below 1000 milliseconds. Despite this obvious dissimilarity between how attention capture is described and studied in experimental psychology, it is of importance to have an understanding of the basic features that are considered to have the power to capture attention. Remarkable, typical properties that are believed to capture attention, such as motion, ?icker or color were not found (for instance in Mack and Rock’s work) to attract attention at a 200 millisecond exposure. Instead, size together with meaningfulness was found to be decisive. Meaningfulness was for instance measured in an experiment where the respondents’ names were exposed and compared to other names. A particularly surprising ?nding was that subjects did not observe a “word”, presented for 200 milliseconds, almost indistinguishable from their own name; i.e. if a person’s name was Jack and it was changed into Jeck then the subjects were largely blind to the stimulus. This strongly indicates that a stimuli with a high level of meaning has the capacity to break through and capture attention (Mack and Rock, 1998). On the other hand, one should have a somewhat ?exible view when considering these results since a typical advertising exposure can last for much longer than 200 milliseconds. It is not uncommon that an advertising exposure will last for minutes and under these circumstances properties such as motion, ?icker and color may have a different impression on subjects. Presumably, because a longer time period will signi?cantly increase the chance a person distributes some degree of attention to a certain area. That is, the area where the advertisement is presented. In addition, interesting in this context is the fact that Mack, in Mack & Rock’s study of approximately 5000 respondents, never found any gender differences. The belief is that if the critical stimulus is equally salient and meaningful for women as for men, attention capture will be equal for women and men.
57
Following is a discussion into the area of how attention works when screening the world around us. It will be explored how attention works when conducting everyday tasks, such as reading some news on a Website while being exposed to an advertising message.
How the attention system operates
The world around us ?lls our senses with information, but only a small fraction of it is relevant to our goal-driven thought and behavior (Milliken and Tipper 1998, p. 191). As a consequence, we have to process information selectively since we otherwise would be overloaded with inputs. An intelligent behavior demands selectivity and that we direct our attention to certain targets that sometimes are external objects and at other times internal thoughts. This selectivity appears to be an enhancement of processing of relevant information and seem to dominate our mental activity (Milliken and Tipper, 1998). In recent years, consensus seems to have been reached that inhibitory processes play a vital role in selective attention. Broadbent’s (1958) ?lter theory, Walley & Weiden’s concept of lateral inhibition (1973) and Moran & Desimone’s (1985) single cell recordings share the view that a crucial function of the attention system is to prevent central mechanisms from interference and overload due to irrelevant input (Milliken & Tipper, 1998 p. 195). This means that the processing of unattended information is impeded and conversely that the processing of attended information is enhanced. Mack & Rock (1998) have also studied the phenomena of inhibition in detail. As has been discussed earlier, in order to achieve a deeper processing of information to take place, the subject ought to attend to an object. Merely looking at an object does not guarantee the fact that the subject is attending to it. As a matter of fact, they found that inhibition has the potential to be the greatest for stimuli presented at ?xation. They are also stating that since evidence of inattentional blindness17 is greater at ?xation, “attention can actively be inhibited from operating on input from some particular spatial location, and this was con?rmed in a series of experiments”.
17
A related phenomena that Mack & Rock were studying at the time. Subjects failed to detect a supratreshold stimulus presented at fovea and this phenomena was named inattentional blindness. 58
Focused and Peripheral Attention
Center of Attention
Figure 2.11
Less processing of events away from C.A.
Figure 2.11 When attention is directed to the cross within the circle then mental effort and processing will be impeded outside of this area. Less focus will be directed to events that are taking place for instance at the vertical line on the right of the circle.
Not only can attention be inhibited, Milliken and Tipper (1998) argue that attention can be used to generate expectation, to concentrate on ongoing processing demands of a primary task or used to direct behavior selectively toward an event among competing events. Still, trying to perform too many tasks at once in a certain task environment will have a negative impact on how these tasks are being performed. The best performance is achieved when the tasks are done one at the time. Attention researchers are often studying this kind of phenomena by presenting a primary task to the subjects to solve and then measuring the performance on a secondary task that overlaps temporally with the primary task (Milliken and Tipper, 1998). Of interest is to study to what extent the response to the secondary task suffers (Pashler, 1984). Kahneman (1973) states that attention refers to the mental effort or cognitive capacity allocated to a task. In this case a limited pool of resources is used by cognitive processes to perform tasks. The more demanding a task, the more cognitive resources is needed for the task. Therefore, performance of a secondary task will be constrained by the resources needed for the primary task. Another view of the phenomenon is that inef?cient performance in dual-task situations is due to an attentional bottleneck that is inherently serial in nature (Milliken & Tipper, 1998). The poor performance on a secondary task would thereby be explained by the fact that the primary task is blocking the serial stage (the bottleneck) forcing the secondary task to be delayed. This kind of explanation would suggest that individuals are not quite capable of diverting mental capacity to two or more tasks at the same time. Instead, it is in fact a matter of sequential processing.
59
The Limited Pool and Attentional Bottleneck Contrasted
Attentional bottleneck Serial view Activity A Activity B Activity C Activity D
Figure 2.12
Limited pool of resources Parallell view Activity A Activity B Activity C Activity D
time
0,0 msec
200 msec
Figure 2.12 The limited pool and attentional bottleneck concepts contrasted. Note that this is just a schematic view to describe the different perspectives.
Common for the limited pool and attentional bottleneck view is that a certain task environment may constitute of a primary task, setting the stage for any other task in that environment. A secondary task will be affected by the demand structure of the primary task in such a way that a more demanding primary task will reduce performance on the secondary problem and other events calling for attention. Many researchers (see for instance Broadbent, 1958, Deutsch & Deutsch, 1963, Treisman, 1964, Walley & Weiden, 1973, Kahneman, 1973, Moran & Desimone, 1985) have modeled attention elaborately. Among these theories on attention Treisman’s model seem to be more appreciated and adequate as opposed to other models (Johnston & Heinz, 1978). Treisman’s view on attention is that there is a sensory register that allows stimulus input to pass, an attenuator that reduces or attenuates processing of information outside focal attention. As a result, there is a limited amount of information that can reach short-term memory where information is stored temporarily and used for different mental processes (Treisman, 1960, Eysenck and Keane, 2000). The components of Treisman’s attenuation theory are sketched in ?gure 2.13 below.
60
Treisman´s Attenuation Theory
Sensory register Attenuator
Figure 2.13 Short term memory
Limited capacity
I n p u t
Figure 2.13 Treisman’s Attenuation theory. (Eysenck and Keane, 2000)
Treisman’s model is different from Broadbent’s selective ‘?lter’ model. Broadbent’s idea was that non-attended input was mechanically screened out (Eysenck and Keane, 2000), which was shown to be incorrect by Treisman (1960). The point with Treisman’s notion is that there is some kind of a barrier that is allowing attended input to pass through whereas non-attended input is attenuated. In her article from 1960 Treisman refers to, as mentioned above, an ‘attenuating’ mechanism. In her view this mechanism may signi?cantly reduce the incoming input, but not necessarily block information in the same way as in Broadbent’s ?lter (Treisman, 1964, 1960). In a more recent work, Pashler refers to the ?lter or attenuator (if using Broadbent and Treisman’s terms) as a “?ltering mechanism” (1998). This term, “?ltering mechanism”, will be used in the following since it is straightforward. From an advertising model perspective it is adequate to state that the model in the ?gure above is opening up the box of attention and giving an elaborate description of how incoming stimuli is being processed. This kind of detailed explanation and modeling infused into an advertising frame can provide greater insights and understanding about how attention operates in advertising. In addition to the descriptions of attention that has been presented, it is also important to consider the different impact of intentional and unintentional attention to an advertisement. From an advertising point of view the best case is, most often, when the individual is intentionally distributing attention to the advertising message. In that case, the individual will have time to read, see and comprehend the entire message communicated. There are of course different levels of intentional attention. One individual may explicitly direct the attention to the advertisement and study it in detail, whereas another individual may brie?y, and intentionally, glance at it. In the case of intentional attention, the advertisement is in the center of attention and not a peripheral object. This means that there is a much greater chance for processing the advertising message.
61
Unfortunately for advertisers, this type of intentional and high level of attention to a message is rarer and increasingly less common in advertising. Instead, the majority of encounters with advertising are brief exposures (Krugman, 1986). These exposures may be unintentional and the individual may not even be aware of the fact that she has seen or perhaps even deny having seen (Krugman, 2000). Brief or involuntary exposures can take place, for instance, when screening a page with a number of advertisements placed together with program content. In this screening process, the individual is trying to localize where the program content is in order to direct focal attention to this area. When engaged in this screening the individual will, simultaneously, be exposed to advertising content when the gaze is wandering over the surface to be screened. In relation to the phenomena of brief exposures Krugman is stating that;
“Thus, to some small but measurable degree, one must note, perceive, or identify what one will not attend to in order to reject it. I suggest that this “pre-attentive process” leaves some trace of the rejected material, certainly not enough to be recalled, but perhaps enough, with repetition of the process and repeated rejections, to be recognized later on.” (Krugman, 2000, p. 53)
From this quotation there are two points that can be made. First, not attending to or avoiding, through rejection, an advertising message will generate an effect. Secondly, to measure this effect the researcher ought to use an observation method that is sensitive enough to capture this effect, i.e. recognition. In Krugman’s argumentation on this issue he is also pointing at that it only takes a very short look to effectively convey the advertising message (Krugman, 2000). As a consequence, the unintentional attention will generate an advertising effect through brief or partial exposures (Krugman, 1986). Janiszewski is arguing in the same way by stating that pre-attentive mere exposures can encourage an individual to have more favorable attitudes even though she cannot recall the initial exposure (Janiszewski, 1993). According to Janiszewski the pre-attentive mere exposures belong to the category of incidental exposures. This is synonymous to the unintentional exposures that have been discussed. The pre-attentive processing of information can be seen as a parallel processing system that is parsing the environment into units that are meaningful, selecting adequate subgroups of units to receive additional attention (Janiszewski, 1993). Wolfe (1998) is describing preattentive information as something that is there to be used and in particular, that pre-attentive processing can be used for the subsequent deployment of attention. The pre-attentive information could therefore be considered as a basic processing of the environment which takes place prior to attending to a particular area. Transferring this discussion into an advertising context would mean that the entire setting is pre-attentively processed at a basic level. This pre-attentive
62
processing includes advertising present in the setting. The individual thereafter selects an area where to deploy the attention, for example the headline of an interesting article in the program content. When brand names and product offers (in the shape of advertising) are included in the secondary information in an environment, they are processed in a pre-attentive manner (Janiszewski, 1993). Extending Treisman’s, Kahneman’s, Mack and Rock’s, Krugman’s and others’ notions of attention to an advertising context gives a richer theoretical picture of the scene presented in the starting point model in ?gure 2.3. The context or media where an advertisement is presented can be seen as a task environment where the primary task is the program content transmitted to the receiver. The receiver is most often interested in the program content and is intentionally directing her attention to this material. There will also be other events, – tasks –, calling for the attention of a subject. Advertising can clearly be seen as one of these tasks that the subject can either attend to or inhibit. Attending to stimuli means more processing of the information whereas inhibition means limited processing of the same stimuli. The individual can also attend to the advertisement/s, which is becoming increasingly less common. Yet, even when the individual is not explicitly attending to an advertisement, brief, partial or mere exposures can have an impact. These unintentional exposures are also important and are making up the bulk of advertising exposure as Krugman states (1986).
Advertising, Attention and Media
A number of media, for instance TV and Radio, that are routinely being used for information transfer of advertising messages are most often being used in such a manner where the program content is interrupted (a break for advertising) and then the advertising messages are being sent. In this case the message stream is serial in nature18. In other media such as newspapers, magazines or the World Wide Web, advertising messages are being presented in connection to the program content. To the viewer or reader, the program content present is calling for the attention of the subject. At the same time, the advertising material is competing with the program content for the attention of the subject. In the latter case, but not in the former, stimuli are presented in a parallel way. This constitutes a setting very much similar to the one in ?gure 2.3. By describing the setting in this way a task environment has been de?ned from an attentional point of view comprising of the media, its program content and the advertisement, see ?gure 2.14.
18
There are however also situations where there are advertisements or logos present within the program content, for instance when it comes to football games, ice hockey or the like. 63
The Web Task Environment
Center of Attention Media vehicle
Figure 2.14 Secondary task
Site Name, Headlines etc Headline Text fields etc Headline Text fields etc Advertisement
Figure 2.14 Viewing the media as a task environment with a primary and a secondary task.
Part of the complete task environment and also valuable to take into consideration is the user activity mode or task mode of the subject being exposed to the stimuli. As mentioned before, attention can be either externally or internally driven. The task environment described in ?gure 2.14 comprises mostly19 of the external task environment. However, a second ingredient, the internally driven aspects have to be considered as well. A user’s activity mode can be set by the mood, time of the day, etc (see for instance Belk 1975) or an intention (Milliken & Tipper, 1998). An intention to solve a given task in a particular way is thereby also part of the task environment and sets the subject in a certain mode when interacting with the media vehicle. When visiting a Web site (a task environment) a person may be in different modes (just as one reader is reading ‘The Times’ to educate herself, another one is reading it to kill time and a third is searching for a speci?c information to quote) and different modes in turn affects the perception of presented stimuli in different ways. The mode a certain Web surfer is in, is dependent on what intention the user has. The user mode can also be altered or triggered by the functions or tasks that the subject encounters on the Web site or as a consequence of the way information is presented. A less goal directed mode might be to just surf around without any speci?c goal, perhaps solely for fun and pleasure, allowing the subject to wander from one place to the other in the media. It is plausible to assume that a user activity mode of this latter kind would make it more likely that attention, to a greater extent also will be directed to the secondary task. In other words; a user activity mode that is highly focused on the primary task, may direct less attention to the secondary task. A user activity mode, which does not require such a focus on the primary task, may open up for more attention to the secondary task.
19
An external environment can however evoke a certain, internally driven, mode.
64
Levels of Involvement
Media, context and ad
Figure 2.15 Media, context and ad
A person in a dual task environment under high involvement
A person in a dual task environment under low involvement
Figure 2.15 In?uence on advertising effect as a consequence of different user activity modes. The user is engaged in activities connected to the web page.
Janiszewski (1998) has studied differences in goal-directed search tasks and exploratory search tasks. He notes that goal-directed search is a process involving selective attention and that the “demands of the search task, organization of the information in the search environment, and the taskrelevant experiences of the searcher determine the ease/dif?culty of the search activity” (Janiszewski, 1998, p. 291). In addition, he is stating that exploratory search routines are used to “monitor the environment when a person is not actively searching for a piece of information” (Janiszewski, 1998, p. 291). Moreover, it can be a screening process on the look out for targets that can be subjected to goal directed search. In this screening process, the exploratory search have procedures to evaluate to what extent attention will be directed to a certain location in the ?eld of view before going to another location. Janiszewski’s view has characteristics that are similar to the view that Web users are either activated in a mode where they are actively looking or searching for some piece of information or just screening, sur?ng, the Web environment to see whether there is something appealing that they can direct their attention to. Sheehan (2002) has in his typology of Internet users, described users in the same way as here. Sheehan argues that there are Web users that have “a speci?c goal in mind; for example, an online session spent searching for speci?c information…” (Sheehan, 2002, p. 63) and also that there are users that “use the Internet for diversion, escape, and/or relaxation”(Sheehan, 2002, p. 63). The ?rst case quoted above coincides with the search concept whereas the second case quoted coincides with the surf concept. Sheehan is developing the typology and describes even more re?ned categories that will not be further developed in this study. Other researchers have identi?ed similar categories as those discussed above and common to them is that they have searching and sur?ng as broad categories in one way or the other (Korgaonkar and Wolin, 1999, Rodgers and Sheldon, 1999, Stafford and Stafford, 2001). Taken together, this reinforces the argumentation that searching and sur?ng indeed can be seen as broad and
65
separated behaviours exhibited by users when using the World Wide Web. Worth to note is that in general there is interplay between the external task environment and what activity mode the user will be set to. When visiting a search engine it is more probable that one is in a search mode. When visiting a Web site for general entertainment, on the other hand, it is plausible to assume that the visitors to a greater extent are sur?ng around for amusement, enjoyment, entertainment or the like. Multi-mode environments are naturally also possible and common, as well as environments used for other purposes.
Measuring Advertising - Click-through and Pricing
Before ending the chapter we will study how Web advertising is measured and how it is being priced. The various advertising means, mentioned earlier, have different purposes as well as effects. The old Web banners, for instance, that recently were much appreciated among Web advertisers had the purpose to generate click-through and bring the user to a target Web site. These kind of Web banners have witnessed click-through rates drop the last few years reaching frequencies below 0.1% (Khermouch and Lowry 2001). Despite this, click-through rates have been and still are widely used as the main measurement of the effect of Web advertising. While the focus on click-through banner advertisements may not disappear, it is likely that advertisements, including animation and video, will be used to a greater extent in a near future to come (Weaver, 2000). This is in fact already taking place. In one of the pilot studies conducted, ?ndings suggest that static banner advertisements are being replaced with more vivid banners on Web sites. Furthermore, since fewer users are actually clicking on the banners, the promotional content has to be presented straight on the host page as an advertisement and not on the page where the hyperlink leads to. Consequently, the decreasing click through rates and other developments,20 have driven the industry to develop advertisements that are more similar to what is being used in print media. There is also empirical evidence suggesting that click-through measures are likely to undervalue the Web as an advertising medium (Briggs and Hollis, 1997). Furthermore, the Interactive Advertising Bureau (Interactive Advertising Bureau, 1997) found that the use of simple involuntary exposure to a banner advertisement without click-through generates an increase in advertisement awareness. There are therefore reasons to believe that there is indeed an effect of advertising beyond the click. At the same time, clickthrough frequency has been, and still is, widely used as a method for pricing in Web advertising. The click-through measure has been a starting point for the development of the pay per click method (PPC). Except for the pay per click method there are also other pricing methods such as pay per view (PPV) or pay per time
20
And also due to technological development such as greater bandwidth, new methods of delivering digital content and the like. 66
unit (PPTU). Pay per time unit is in essence an agreement between the Web publisher21 and the advertiser to use a certain space during a certain time. This might for instance connote that the advertiser buys the upper right corner of a Web site for a period of one month or any other time period. Pay per time unit can also be a package of different locations on the Web site during a prede?ned time period, where some of the locations are prominent and others being less salient. Pay per view or sometimes called “CPM” (Cost per thousand, Hoffman & Novak, 2000) is a pricing method based on impressions or exposures (Mangàni, 2003). In the same way as with conventional broadcast and print advertising the pay per view method is taking into account “[…] the amount of advertising delivered” (Hoffman & Novak, p. 180, 2000). The “amount of advertising delivered” is sometimes used for broadcast as well meaning that the advertiser buys airtime at a certain time when a certain number of individuals are supposed to be reached. The “delivered amount” is however a more accurate measure in the Web case compared to broadcast. An example can reveal the difference. If an advertiser buys airtime the publisher will ask a certain price that depends on for instance the duration of the commercial, the time of the day and whether the spot is to be aired in connection with a certain program or the like. After taking these aspects into consideration the publisher will calculate and present a ?nal price.22 This price will on the other hand not (usually) take into consideration whether it has only been 13.8 million individuals that watched the show and the subsequent commercial instead of the expected 15.8 millions that normally watches the show. The pay per view pricing method on the other hand is taking each and every visitor into account since the pay per view measure, like the pay per click, is with a high degree of precision counting every exposure. There are not only “pure” pricing methods like those described here but also hybrids that combine different aspects to suit the needs of clients, publishers or both. Using the pay per click method as a means for pricing does not appear to be the most rational choice when the click-through frequency measure has an additional communication effect as measured by, for instance, recognition or recall, that is not accounted for by the pay per click method. This has the inherent consequence that those buying advertising space on Websites are receiving a service at constant discount, a discount that we do not know the magnitude of. From a Web publishers’ perspective this is a state that ought to be recti?ed. It is, however, perchance explicable that there may be a widespread malpractice in pricing matters considering that research in the area has not been able to keep even pace with the development. Hence, practitioners have been left to ?nding their own way.
21 22
From the perspective of those selling advertising space or the like.
There may also be other factors taken into consideration such as ?rst time discount, bulk discount, repetition and the like. 67
In this framework it is worth to mention that for example the search engine industry is using the pay per click pricing method to a larger degree. The industry is also increasingly successful in their endeavor to generate revenue. In the case of the search engine industry the pay per click method is tied to the number of clicks on keywords or Adwords that are presented as search results. In connection to the issue on what pricing methods are relevant to use, it is appropriate to also include how advertising is being measured and how it has been measured historically. Click-through has never before (before the World Wide Web came in to being) been a means to measure advertising effect. To measure clicks was a possibility that came along with the inherent properties of the information technology used for the World Wide Web. The click measure was ?rst and foremost a means to register activity on a Web site. When using links connecting a banner to a target Web site it became possible to measure how many users that clicked on the banner and were redirected to the target site. Soon it became a widespread practice to count the number of clicks as a measure of advertising effect. However, the perhaps most prominent researcher in the area of how to measure advertising is advocating other means than the click. Krugman has elaborately studied the use of recognition and recall as measures of advertising effectiveness. He is suggesting that recall and recognition are actually measuring two different things. Recall, which is when an individual is being asked if she can recall seeing an advertisement in a certain media and tell what brand name it was, is part of measuring verbal memory. Memory for words is though relatively low with a high level of forgetting (Krugman, 1985). Recognition on the other hand is measuring pictorial memory. When measuring recognition, an individual who has been exposed to an advertisement (the target ad) is presented with a number of irrelevant or phony advertisements of which one is the target ad. The main purpose is to see if the subject recognizes the target advertisement. This, is an “easier task” as opposed to when testing recall since right-hemisphere memory for pictures and images is exceptionally high with very little forgetting (Krugman, 1986, Krugman 1985). Even though both recall and recognition can be used for measuring advertising, Krugman is most of all arguing in favor of recognition as the main measure. Krugman is stating:
68
“Advertising researchers have argued about the relative merits of using recognition or recall as measures of advertising effect. Because the criterion of recognition is much more easily achieved than that of recall, it has been criticized as being less sensitive. Underlying this “technical” controversy, however, is the fact that the use of recognition justi?es modest advertising expenditures. I would reposition the recognition versus recall problem with this proposed addition to the theory of involvement – i.e., the nature of effective impact of communication or advertising on low-involvement topics, objects, or products consists of the building or strengthening of picture-image memory potential. Such potential is properly measured by recognition, not by recall. The use of recall obscures or hides already existing impact.” (Krugman, 2000, p. 53)
When using the click-through to measure advertising effect neither recall nor recognition is measured; only those who have clicked are accounted for. This may severely underrate the effect of an advertisement. Consequently, when a publisher uses click-through as a basis for pricing, the entire advertising message delivery will not be taken into account. A more sensitive measure ought to be used, i.e. recognition. It is therefore the pay per view method that is argued for at the expense of pay per click method. The pay per view method is closer related to actual exposures and to measure recognition is an appropriate way to measure exposures. Since the industry is frequently using the click-through measure, it is pressing to make known that there may be more adequate means when advertising effect is to be measured. It is always good to remember that “The practice of evaluating Web advertising on the basis of click-through is like evaluating TVads for automobiles on the basis of how many people visit the showroom the next day” (Briggs and Hollis, 1997, p. 33).
This chapter
This chapter set off by identifying the origin of advertising with the broader concept of marketing communication and marketing in general. Afterwards, it was elaborated around the essence of advertising and how advertising has developed into the World Wide Web. A number of issues that has bearing on the research questions that were formulated in the introduction chapter have been covered, placing attention to advertising on the World Wide Web into a theoretical context. Theories from disciplines other than advertising have been brought in to the theoretical framework to complement and develop the knowledge at hand. In the next chapter the different parts of the framework will be integrated in order to formulate relevant hypotheses taking as a starting point the research questions.
69
Chapter Three
Formulating Hypotheses
With the research problems as a starting point and the theoretical chapter as a guiding tool, parts of the theory that are relevant will be integrated and used for the formulation of the hypotheses for this study.
70
First section - Attention effects of context
In the theoretical chapter it has been elaborated on how the setting or the context in which the advertising resides can in?uence the impact of advertising. It is important to study this contextual dimension to understand how it affects advertising and whether it is powerful or can be neglected. Extending Kahneman, Janiszewski, Milliken, Tipper and Treisman’s theories of attention to a advertising context may provide insight to how attention is affected by context and activity mode. In the theory section, the Web environment was described as a task environment where the primary task is the program content transmitted to the receiver. The secondary task in this context is then the advertisement. Furthermore, a secondary task will be affected by the demand structure of the primary task in such a way that a more demanding primary task will reduce performance on the secondary task and all other events calling for attention. Therefore; H1: A situation where respondents are exhibiting a less goal oriented sur?ng behavior will elicit more attention to an advertisement than what a more goal oriented sur?ng behavior will. To be rejected… [H0H1 A situation where respondents are exhibiting a less goal oriented sur?ng behavior will elicit equal or less attention to an advertisement than what a more goal oriented sur?ng behavior will.] If H1 is correct, we can conclude that individuals set to search mode, pay less attention to an advertisement than what a respondent in a surf mode does. Another contextual dimension that has been elaborated on is the physical appearance of the environment where an advertisement is presented. This physical environment that was described as a task environment may also in?uence advertising. One aspect of this task environment is its inherent complexity which can be manifested in a number of ways. One such way is the complexity of the search system on a Web site and the search depth that it has. Given what has been learned in the theory it is plausible to state that; as complexity increases, the demand on the primary task will increase and thereby reducing the attention to the secondary task. Hence: H2: A Web environment that is less complex, with regard to its search depth, will elicit more attention to an advertisement than what a more complex Web environment does. To be rejected… [H0H2: A Web environment that is less complex, with regard to its search depth, will elicit equal or less attention to an advertisement than what a more complex Web environment does]
71
If hypothesis H1 and H2 will be supported then there is reason to consider an alteration of the advertising models displayed in the theoretical chapter. If H1 and H2 are correct then the models studied can be elaborated. More speci?cally, the attention construct should be described more in detail taking into account what is happening between the advertisement, context, attention getting etc and the attention construct. What is lacking in the models is a more elaborate description of the attention construct. If Thorson’s models are taken as an example to compare with, it can be said that what is lacking is an environment or setting construct that takes into account the environment’s effect on an individual’s attention. There is also a lack of a construct that can impede the incoming stimuli. These issues that have been pointed at in the previous paragraph will be further addressed under the section “Modeling attention to advertising”.
Second section - Advertisements and their respective attention effect
In the previous chapter it was argued around different attention getting techniques and that one and the same advertisement using different attention getting may receive different amounts of attention from individuals. Chandon’s study indicated that animation gave a greater response measured as click-through frequency compared to non-animated advertisements. He also recognized that many advertisers had stopped using non-animated advertisements. Lohtia, Donthu and Hershberger received results that pointed in the same direction and they strongly recommended the use of animation as a means to attract attention to the advertisement. Sundar and Kalyanaraman’s ?ndings were also in favor of animation. Common for the studies above is that none of them used recognition or even recall when measuring results, instead they used click-through frequency as a measure. In the attention section a number of features were shown to have the ability to attract attention. Mack and Rock had studied how color, numerosity, motion, ?icker, shape, location and texture segregation were performing when attracting attention. After a great number of experiments Mack and Rock concluded that properties such as meaningfulness and size indeed could attract attention. Yantis and Yantis & Jonides were arguing that in particular new objects presented in the visual ?eld had the capability to attract attention. These studies referred to here were studying the phenomena of attracting attention on a very basic level where these basic features were studied. In advertising these basic features are put together on a greater scale and the exposure times are much longer. It is therefore important to understand how attention getting works on a “macro” level. An animated advertisement is actually using some of the basic features that our experimental psychologists have studied in attention experiments. An animated advertisement will by nature have the possibility to present new objects in the visual ?eld since the animation itself is not stationary. It also has motion or ?icker since animation is adding motion to an object that otherwise would be static.
72
Taken together it seems plausible that animation will evoke greater attention than what a non-animated advertisement will. Hence: H3: An animated advertisement will elicit more attention than a static advertisement. To be rejected… [H0H3: An animated advertisement will elicit equal or less attention than a static advertisement] As was stated already in the theoretical framework, a pop-up advertisement is merely a special case of an animated advertisement from a visual perspective. When the pop-up advertisement appears in a certain location it creates “motion” where it appears. This could be seen as an “abrupt onset” as Yantis described it theoretically, since the pop-up appears in an empty area (sometimes also in other areas than empty areas). Cropper & Evans and Smith & Goodwin found in their study support for animation as a means to attract attention from one area of the screen to another area because of the “visual distinctiveness”. Relevant in the case of pop-up advertisements, since it can be perceived as a special case of an animated advertisement, is of course all the arguments presented in favour of animation in hypothesis 3. With this said: H4: A pop-up (traditional pop-up, over-the-page) advertisement will elicit more attention than a static advertisement. To be rejected… [H0H4: A pop-up (traditional over-the-page pop-up) advertisement will elicit equal or less attention than a static advertisement.] It is also expected that that a pop-up advertisement will elicit more attention than an advertisement that is just animated. The reason for this is that the popup advertisement is not only animated but also that it appears in an empty area creating extra “motion” in the area where it appears compared to its animated counterpart that is just animated and staying in the same area the whole time. It is expected to be easier to inhibit the animated advertisement than what it is inhibiting a pop-up advertisement that is unexpectedly popping-up in the viewers’ visual ?eld. Therefore: H5: A pop-up advertisement will elicit more attention than an animated advertisement. To be rejected… [H0H5: A pop-up advertisement will elicit equal or less attention than an animated advertisement]
73
The Second Section Extended
During pre-analysis of hypothesis 5 that was conducted during the data collection, a few interesting ideas surfaced that ultimately lead to extend the study somewhat. This part was mostly driven by a few theorems that originated from Yantis, Jonides, Mack and Rock. When analyzing hypothesis 5 it seemed obvious that the pop-up advertisement would be more effective than the animated advertisement, and so it was. And the result would clearly follow Yantis and Jonides idea that a new element in the visual ?eld would attract more attention. However, before jumping to premature conclusions there was though, one additional aspect that could be taken into consideration, namely the frame. The frame of a pop-up advertisement is an additional element to consider which was not considered in isolation initially, since previous studies have not taken this into account. The frame on the pop-up advertisements that were used in the treatments added 50 pixels on the height and 25 pixels on the breadth. This means that the pop-up advertisement in fact comes with a frame that makes it 21 075 pixels or 29.2% larger than the animated advertisement. At the same time we have to remember that it is not the advertisement itself that is larger, they are of the same size, however the difference is just in the frame. The difference in effect that was observed in both the pre-analysis and the ?nal analysis of hypothesis 5 may (or may not) originate from the frame around the advertisement in the pop-up and not because of the “abrupt presentation” of the advertisement. It could of course be that there is an effect from both of these factors – the abrupt presentation and the frame. These ideas were considered of such an importance that a 6th hypothesis was formulated based on the reasoning in the two paragraphs above. In order to study the effects of the frame separately, a new set of experiments had to be designed and conducted. This hypothesis demanded the study to be extended with 292 more respondents and the hypothesis is an extension complementing the original hypotheses. It complements the 5th hypothesis in such a way that we may obtain an answer to the “frame effect versus the abrupt presentation effect controversy”. What we will in fact do here is to put in an advertisement that is animated, but “popping up” only as an in-page advertisement and not as an over-the-page advertisement like traditional pop-ups. By doing, so we will have a pop-up advertisement that does not have a frame. Thereby we can compare the effect between a pop-up with a frame with a pop-up without a frame. By using this procedure we will perhaps be able to resolve the controversy. This can be formulated into hypothesis number 6a) and 6b). H6a: A traditional pop-up advertisement will elicit more attention to the advertisement than a FL1 pop-up (in-page) advertisement.
74
To be rejected… [H0H6a: A traditional pop-up advertisement will elicit equal or less attention to than a FL pop-up advertisement.] And furthermore: H6b: A static advertisement will elicit more attention to the advertisement than a FL pop-up (in-page) advertisement. To be rejected… [H0H6b: A static advertisement will elicit equal or less attention to the advertisement than a FL pop-up advertisement.] A result that would be in line with hypothesis 6a) and 6b) can give some information about the “abrupt presentation effect” and the “frame effect” so that we perhaps can resolve the controversy. However, there will still be obstacles in our way if we want to attribute effects to the right construct. The point is that not even an answer to the “abrupt presentation – frame” construct pair will be suf?cient for a comprehensive understanding. The problem is the dual property of the construct under examination here. Let us elaborate further in order to decipher what this means. First, the abrupt presentation effect has to do with the popping up of the advertisement and will be measured in the hypotheses above. The frame effect has to do with the frame of the pop-up advertisement and can be divided into subordinate constructs that is building up the frame. One can of course measure and analyze the entire frame effect and get an answer of that effect and this will be measured in the hypotheses above. However, one can also analyze the subordinate constructs. From a stimulus point of view an advertisement that is popping up with a frame has at least two properties that are relevant here. We can talk about the frame as having a size component and a “frame-meaning” component. The inherent properties of a pop-up advertisement (when it comes with a frame) will inevitably increase the size of the advertisement. Just the fact that this popup package is larger in size may have an effect when measuring the response from this stimulus. This has to be considered. Furthermore, there may be a second effect originating from the frame that does not have to do with the size of it. This second “frame effect” might be connected to meaningfulness or formulated differently; the frame may represent a distinct meaning compared to an advertisement without a frame. The rationale for this is that a frame that is popping up represents an intrusion in the visual ?eld of
1
The special pop-up advertisement will hereafter be called FL (FrameLess) to make it easier to separate it from the traditional pop-up advertisement. 75
a Web user and this intrusion may have a certain meaning to the user and that could then be the reason why they attend to the advertisement. The meaning that it represents is most likely not a positive one, but a negative. So instead of applying Yantis idea of new objects in the visual ?eld to the described situation it might be even more fruitful to lean towards Mack and Rock’s notion of meaning, or perhaps both. Following Mack and Rock’s notion of meaning and late selection then a popup advertisement without a frame should carry “less meaning” from a user’s perspective than what a pop-up advertisement with a frame carries. This would clearly mean that the frameless pop-up advertisement will elicit less attention to it than the pop-up advertisement with a frame. Nonetheless, as we may recall, Mack and Rock were discussing around what actually attracts attention and they were pointing at “meaning” as one factor. They also discussed around a late selection theory to explain why subjects for instance saw their names when spelled “Jack” but not when spelled “Jeck”. Translating this to the advertising situation; can it be that they see the advertisement when they see the frame since it is a carrier of negative “meaning”? Without carrying on any further we can already here say that we will obtain the effect of both size and frame-meaning at the same time2 when we are measuring the frame and then we can not with certainty attribute the effect to the right construct. Though, the fact that we will not obtain an answer to the frame-meaning issue is however not a reason to not pursue this endeavor, on the contrary. As we have discussed earlier we may reach a greater understanding of the boundaries of the pop-up effect and to resolve the “frame effect”3 versus the “abrupt presentation effect” issue. The assumption of hypothesis 6b is that if 6b is correct then the abrupt presentation for the FL pop-up advertisement plays a minor role when comparing it to the static advertisement. In this experiment the two tested advertisements have indeed equal size. The very reason why the static advertisement is expected to elicit more attention is because it has longer exposure duration. If the abrupt presentation, on the other hand, is strong this should more than well compensate for the shorter exposure time for the FL pop-up advertisement and the attention to the FL pop-up advertisement should be greater than that for the static advertisement. Finally, we should be open for the fact that there may be other not yet known effects that may have an in?uence on the effects that we are
2
What is clearly needed to potentially resolve this issue to the fullest and to be able to attribute the effect to the right ”location” is a pop-up advertisement with a frame and where the size of the frame should not exceed the size of the advertisement X. Then the size effect would at least be equal but the advertisements would be unequal. A higher effect for the pop-up advertisement with the frame would then indicate that it is indeed the frame and not the size that is behind the increased effect. This could however not be tested in this study due to fund and time constraints. 3 With the sub constructs not separated. 76
observing. Nevertheless, here is an opportunity to reach a bit further in our understanding of this phenomenon, especially since no research has been done in this particular area.
Modelling attention to Web advertising
Vakratsas’ and Amblers’ model (see ?gure 2.4) is the most similar to the model and ideas to be presented here. However, there are some differences that are related to the in?uence of the environment/context dimension and also the attention getting techniques. Furthermore, from the theoretical description of attention in the previous we know that the attention system is an intrinsic part of cognition and the point here is to open up the cognition part of Vakratsas’ and Amblers’ box somewhat. Drawing upon Treisman’s attenuation theory and Vakratsas’ schematic model it can be visualized what is taking place when taking into account the studied factors’ (the ones in hypothesis H1 and H2 as well as H3 - H6b) interplay with attention. Based on this the model is to be built and tested. Starting at the left (in ?gure 3.1) there is a contextual dimension, the Web environment/mode (hypotheses H1-H2), in which the advertisement is placed in. Given that it is a question of a Web advertisement the advertising message has a potential “propellant”, the extra attention getting technique, which can enhance attention (hypotheses H3-H6b). The environment/mode and the attention getting technique together with the advertising message constitute the input. As we have seen in the theoretical chapter there are a number of terms that are being used for the construct that Broadbent initially called a “?lter”. Treisman has used the term ?lter too in her work but also pointed out that her “?lter” is an attenuator which is attenuating the input. She has also used the term “barrier”. Still other researchers have used the term bottle neck and of course also the limited pool that can be seen as a barrier. No matter what terminology that is being used one could say that there is a ?ltering mechanism, borrowed from Pashler, which is located between the input and attention. The ?ltering mechanism is placed as a barrier in the center of the model. This construct is actually a part of the attention system so the separation is merely to visualize how the system works. The ?ltering mechanism can allow advertising messages to pass through to a greater or lesser extent depending upon how the contextual dimensions affect the mechanism. On the right there is unintentional and intentional attention leading to memory. The two different kinds of attention are covering both the situation when an advertisement has been attended to in a peripheral unintentional manner and when it has been more intentionally attended to. If isolating and focusing closely on the critical constructs that have been discussed, it could take the following form;
77
The Proposed Model
Filtering Mechanism Unintentional Attention Intentional Attetion Unintentional Attention Affecting the permeability of the filtering mechanism
Figure 3.1
Contextual dimension influencing
Attention Getting The advertising Technique message Contextual dimension influencing
Memory
Figure 3.1 A model of attention in Web advertising.
A straightforward description may summarize the relationships between the various parts of the model. The advertising message, on the left in the ?gure above, is the source of stimuli input and may be propelled with an attention getting technique. The advertisement is placed into a context that may impact the advertisement’s ability to penetrate the ?ltering mechanism in order to be attended to, intentionally or unintentionally. The advertising message will at the end be delivered to memory. From the author’s point of view the proposed model in ?gure 3.1 is a plausible translation of relations between concepts, given that the hypotheses are supported. That is, that support is generated for hypotheses H1, H2 and the general notion4 expressed in H3-H6b.
Third Section – measurement and pricing
This third section of the hypotheses is targeting how advertising on the Web is being measured and can be tested directly with hypotheses. The second area is not being tested directly with hypotheses but the results from the ?rst area will have implications for the second area since they are intrinsically tied together. In one of the pilot studies that were conducted, ?ndings suggested that banner advertisements, with the sole purpose to redirect users to their own Web sites, have to quite some extent been replaced by advertisements containing a complete message on the advertisement itself. This result is supported by other studies conducted, which has been reported in the theoretical chapter. Furthermore, since fewer users are in fact clicking on the banners, the promotional content has to be presented straight on the host page as an advertisement and not on the page where the hyperlink leads to. In recent years this is actually what has happened and recent studies are indicating that Web advertisements are increasingly designed for this purpose as well.
4
That there are attention getting techniques in Web advertising, as described, that enhances attention to an advertising message 78
However, despite this development, advertising effect is still being measured as click-through frequency. Krugman, one of the prominent researchers of advertising effects, have advocated the use of recall and most of all recognition in order to better account for the effects of advertising. This is in contrast to a widespread practice in the Web advertising industry. In order to investigate whether it is a correct practice or not to use click-through as a measure of advertising effect, the following hypothesis is formulated. H7: A Web advertisement will generate more respondents being aware of the advertisement as measured by the recognition measure than it will generate clicks on it. To be rejected… [H0H7: A Web advertisement will generate equal or less respondents being aware of the advertisement as measured by the recognition measure than it will generate clicks on it.] If H7 is correct it would imply that there is an advertising effect taking place that is not accounted for if only the numbers of clicks are measured. Adequate measures ought to be used in order to really know the effect of advertising on the Web. This would be in line with for instance Krugman’s view on how to measure advertising effect and contrary to a widespread practice among contemporary practitioners only measuring click through frequencies. It would also have implications on models for pricing in the area of Web advertising. This would indicate that the basis for the pay per click- pricing method (PPC) is inadequate and should be altered or dropped in favor of pay per view (PPV) models.
Other effects regarding Web advertising
In the theoretical chapter it was pointed out that it is not entirely positive when advertisers are over ambitious in their striving to capture attention. In some cases there are of course individuals that gladly will look and spend time and attention on viewing an advertisement, perhaps even clicking on it. This is though not always the case. Research by for instance McGalliard, Nielsen, and Spool et al. show that some subjects do not appreciate advertising to such a great extent and that attention getting techniques can be perceived as “annoying”, “irritating” and even “evil”. Based on this it is reasonable to expect that the more attention getting techniques employed the more annoyance evoked. In this experiment three different attention getting techniques will be tested against each other. H8: The risk for an advertisement to be perceived as annoying will be greater when more attention getting techniques are employed in the advertisement. This would mean that a static advertisement will be the least annoying, an animated advertisement will evoke intermediate annoyance and a pop-up advertisement will be the most annoying.
79
To be rejected… [H0H8:All other alternatives.] A rejection of H0 will tell that there is a difference between the various treatments. This difference would indicate that attention getting comes at a “price”, the price of the advertisement being perceived as annoying by the audience. With further analysis, it may also tell what kind of attention getting that is perceived as more or less annoying by the respondents.
Summary
The hypotheses in this chapter are summarized in the following: H1: A situation where respondents are exhibiting a less goal oriented sur?ng behavior will elicit more attention to an advertisement than what a more goal oriented sur?ng behavior will. H2: A Web environment that is less complex, with regard to its search depth, will elicit more attention to an advertisement than what a more complex Web environment does. H3: An animated advertisement will elicit more attention than a static advertisement. H4: A pop-up (traditional pop-up, over-the-page) advertisement will elicit more attention than a static advertisement. H5: A pop-up advertisement will elicit more attention than an animated advertisement. H6a: A traditional pop-up advertisement will elicit more attention to the advertisement than a FL pop-up (in-page) advertisement. H6b: A static advertisement will elicit more attention to the advertisement than a FL pop-up (in-page) advertisement. H7: A Web advertisement will generate more respondents being aware of the advertisement as measured by the recognition measure than it will generate clicks on it. H8: The risk for an advertisement to be perceived as annoying will be greater when more attention getting techniques are employed in the advertisement. This would mean that a static advertisement will be the least annoying, an animated advertisement will evoke intermediate annoyance and a pop-up advertisement will be the most annoying.
80
Chapter Four
Research Method
In order to answer the research problems an adequate method is needed that can develop the information required. In the following chapter the research method is outlined and elaborated in detail, showing the path of the studies conducted.
81
Approach
This study is in its essence a study of social activities. Social activities can be studied in a variety of ways employing different methods to achieve understanding and to capture and create knowledge or knowledge statements. Independently of the research strategy chosen, it can be seen as a map or a tool for the researcher on her way through the research process to ful?ll the purpose of the study (Bengtsson, 1994, p. 41). The researcher is though not a blank sheet when initiating the research. Instead, she will have all prior knowledge and views as a luggage that may tilt the map in some way. While some researchers are more prone to view the world in a more objective way prioritizing certain kinds of knowledge, others on the other hand may be more interested in other more subjective aspects of the social world. My view is that there is a whole spectrum of possible views of the world and that they all can contribute with different kinds of understanding. Czarniawska (1999 p.9) states that different views contribute in their own respective ways, and it is reassuring that, many versions of the world are possible. Johansson-Lindfors (1993 p.41), when reasoning around Morgan’s views, is stating that objective as well as subjective views on reality can be relevant when studying different research problems in social science. Bonnedahl (1999, p.39) is arguing in line with this stating that knowledge about reality to a greater or lesser extent can be seen as either objective or subjective. Following Bonnedahl’s eloquent phrasing, this study is not aiming at providing an objective truth but rather descriptions that can be inter-subjectively understood. The research problem and the purpose of the research is also of importance in this context since they, to quite some extent, will be guiding the research process and also reveal what kind of knowledge the researcher is aiming at. The research problem is also to at least some extent narrowing down the potential approaches and methods that can be used (Yin, 1994 p.4). However, a given research problem can be approached by using either a qualitative or a quantitative approach, or a combination of the two. Furthermore, one might use a survey, case studies, interviews, experiments or some other strategy in order to answer the research problem and to extract information. These strategies have various advantages as well as disadvantages that would make a certain strategy very suitable in one case while perhaps completely inappropriate in another setting. Whether to choose one strategy, at the expense of another strategy, is a delicate decision that the researcher has to make. For instance, the qualitative approach is perceived to provide rich descriptions of the social world (Denzin & Lincoln, 2002 p. 10). According to Denzin (2002, p. 3) the qualitative research is a “situated activity that locates the observer in the world” and furthermore that “qualitative researchers study things in their natural settings”. The quantitative approach, on the other hand, makes it manageable to study a larger selection of individuals and thereby provides, to a larger degree, knowledge statements that can be generalized to a larger population.
82
The research strategy chosen for this study has been based upon a few considerations that were taken into account. First, a quantitative approach is appropriate as one of the aims is to be able to identify differences between treatments in a selection and with the ambition to generalize to a greater population. A qualitative approach is not quite suitable for that task, a notion supported by Johansson Lindfors (1993, p. 61-65). Furthermore, theory suggested that some of the phenomena to be studied are assumed not to occur frequently in a population. When an infrequently occurring phenomenon is to be studied and compared with other occurrences then a quantitative approach with a fairly large sample is preferable. Another reason is that, in one part of the study the objective is to ?nd (or not ?nd) variation and rates of occurrence between groups and within groups, which makes the qualitative approach less suitable. Consequently, the research strategy chosen is quantitative in nature and includes a survey to extract information from a large number of subjects. Considering that the primary aim of the study is to investigate differences in response between subgroups subjected to different treatments then some kind of test has to be conducted. In light of this ambition the experimental approach is a natural choice. In the words of Cook and Campbell;
“The word experiment denotes a test, as when one experiments with getting up two hours earlier to see if this makes one’s working day more productive.” (Cook and Campbell, 1979, p. 2)
This study is however not a matter of testing productivity but to test a number of speci?c advertising messages under some speci?c conditions. The purpose of the various experiments is to answer and highlight different aspects of the overarching research problem. A number of hypotheses have been formulated, originating from the research problem. These hypotheses are derived from theory and represent statements, which “speci?es how two or more measurable variables are related” (Churchill and Iacobucci, 2002, p. 93). Formulated differently one can say that the hypotheses are a possible answer to the research question (Aaker, Kumar and Day, 2001, p.49). Multiple experiments were designed and set up to arrive at a comprehensive understanding of the many facets of the research problem. How the experiments were conducted and how they were designed will be elaborately described in the following section.
The essence of an Experiment
Compared to other methods experiments demand rigorous planning in advance in order to avoid pitfalls and obstacles that may lead the researcher astray. An example will best explain the essence of an experiment. For instance, Ivan Pavlov, a Nobel Laureate in 1904, conducted his well known experiments in such a way that he rung a bell and subsequently gave food to his research subjects, the dogs, for a period of time. After a short “learning”
83
period he could just ring the bell and afterward measure an increase in the saliva produced in the dogs’ digestive glands. In that case Pavlov could with a degree of certainty say that the measured effect (increase of saliva) originated from the ring in the bell and no other occurring events. The reason for this was because Pavlov controlled the environment so that there were no other events that could explain the effects that were observed. Hence, Pavlov could rule out other potential “z-variables” (other variables that could have explained the event) as having any impact on the observed effect. Pavlov was “so concerned with isolating the cause of the dog’s salivation that he constructed a ‘Tower of Silence,’ with eight soundproof rooms separated by spiral staircases” (Becker, 2003, p. 88). Pavlov had thereby shown, without the interference of other variables, the cause (ringing bell) and effect (salivation) of conditioned re?exes and at the same time paved the way for a more objective way of studying behavior. In another rather well known experiment1 a horse was claimed to have the capacity to perform some simple calculations. When asked by his master, Wilhelm von Osten, what is the sum of 3 plus 2, the horse tapped his hoof ?ve times. This and other tests of the intelligence of the horse (which went by the name “Clever Hans”) were taken as evidence of an understanding of human language as well as simple calculus. Albeit many researchers looked into the case it took quite some time before it was found that it was in fact another variable that had an effect on Clever Hans’ counting skills. Oskar Phungst, a psychology researcher, found that Hans was able to respond to subtle physical cues (more or less equivalent to body language) so he stopped “counting” when he reached the predestined number, not because Clever Hans could count, but because he could read the body language of the person who was asking the question (Fernald, 1996). When Clever Hans reached the right number the body language of the one asking the question was slightly altered and then Hans stopped. This unexpected variable “z” gave an explanation to the whole phenomena and at the same time disproved that the horse had an unprecedented intelligence2. In short, what is preferred is the situation of Pavlov and to avoid the situation with Clever Hans. It is for that reason imperative in experimentation to control the environment in order to be able to reach non-confounded conclusions. This and many other problems are the focus for the researcher’s effort to ensure reliable and valid results. When taking the discussion above and formulating it in a more formal way, then an experiment is, in its essence, a test of what the outcomes are under
1 2
Actually the events started out as a show for the public but eventually led to tests conducted by researchers.
Even though Clever Hans was not intelligent in calculus, he was very skilled and intelligent in reading subtle physical cues. 84
certain criteria. It is distinctly stated “certain criteria” for the reason that the researcher often controls the environment and wants to be assured that it was actually and solely due to the treatment X that the change of Y occurred and not for any other reasons. The researcher has a strong wish to be able to exclude other explanations Z which could interfere with the conclusion that Y happened because of X and nothing else than X. The general idea with an experiment is to observe the variation in one variable as a consequence of the manipulation of another variable, while other variables are being kept stable. If this endeavor is accomplished then the researcher may reach a state where she can say, with some degree of certainty, that X caused Y. To conduct research in a natural setting is often what researchers aim for. It is though not always achievable. There are many reasons for this, for instance scarce resources such as time and money but also practical problems for instance how to study a phenomena and to have control over disturbing or distracting z-variables in a natural setting. Another way to conduct the study, which is often more manageable, is to conduct research in a laboratory setting where it is more uncomplicated to control variables. The strength with the laboratory rests on the possibilities it provides to keep the environment stable in the way the researcher wants it to be. It thereby becomes easier to be certain that the observed effect can be attributed to the right cause. A typical experiment is often staged so that the researcher is ?rst doing a pretest where the treatment group is being measured on the variable to be examined. Afterwards this group will receive a treatment followed by a posttest to see what effect the treatment had. Usually the researcher will have a control group as well where respondents are not receiving any treatment or receiving an irrelevant treatment, i.e. placebo. This will put the researcher in a position to compare the treatment group with the non-treatment group. In the posttest the researcher can notice whether there is a difference between the groups and if so, how big the difference is. When properly conducted, the researcher can then conclude that the effect observed in the posttest is caused by the treatment. In this study, this strategy described was, however, not used. Instead, a slightly different kind of experimental design was employed; a design which is very frequently used in advertising research as well as and in other areas. The design included multiple treatments since the overall idea was to do multiple comparisons. This experimental strategy is sometimes called “planned variations” (Cook and Campbell, 1979, p.349) or more commonly “factorial design”. Compared to the “traditional” experiment described in the above paragraph the multiple treatments experiment does not include a no-cause baseline “making it dif?cult to test hypotheses about “absolute” cause as opposed to hypotheses about differential impact” (Cook and Campbell). This is in line with the purpose of the study, which is to investigate differences between treatments, rather than to state the amount of the absolute level of a certain treatment in comparison with the placebo group. A non-treatment group would therefore not have contributed to the interpretation of the
85
results. Albeit the lacking of an explicit or separate control group, it should be kept in mind that each of the treatments “acts as a control for the others” (Aaker, Kumar and Day, p. 345, 2001) making a traditional control group redundant. This procedure is adequate since, as noted earlier, the purpose of the experiments is to compare differences between multiple treatments. However, a non-treatment group could have provided some information about the level or how many of the respondents that falsely claimed to have seen an advertisement in a treatment when they actually did not see one. This problem was however overcome by using an alternative procedure. The respondents were not simply asked whether or not they had seen advertisement “X” but instead they had to point out which one out of ?ve different alternatives they had seen. Rothschild and Singh (1983) are stating that response bias may be eliminated by using multiple alternatives when the stimulus item in the recognition test is presented together with one or more distracter items. By doing so, the respondents could not falsely claim to have seen an advertisement they had not seen. If the respondents guessed which advertisement had been presented their responses would have been evenly distributed between the ?ve alternatives making it possible to calculate the effect of the guessing and to subtract this from the results. However, this never became a problem indicating that the ?ve alternatives was an adequate means for deterring the respondents from guessing. Another noticeable difference, compared to traditional experiments, is that there were no pretests in order to measure the level of the dependent variable or any other variable. Since the treatments were ?ctitious innovations then we know with certainty that all respondents in all treatments were equivalent to “zero”3 before they received a treatment. Consequently, any pretests would have been super?uous.
3
They were “zero” in that respect that no one had ever seen or attended to the treatments before. The attention to the advertisements in the various treatments was the variable to be measured. A pre-test would not provide additional information and would also be logically inconsistent to implement. 86
Overall design
The research design comprise of a pre study phase, which is followed by the main study. Due to the nature of the research problem and also because the area of Web advertising is a fairly new one, pilot studies were undertaken in order to get the required pre-knowledge to do the main study. The studies, which vary in their width and time span, were conducted in the beginning of the project. The studies can be described as here below:
General Description of the Research Design
Pilot study 1, How are Websites being used, Web advertising and Web categories
Figure 4.1
Pilot study 2, Test of Web experiments and on-line questionnaire etc.
Main study, Three experiments of advertising effects
Figure 4.1 General description of the research design.
The pilot studies were targeting different areas of the subject. The purpose of pilot study I was to obtain an overview of different kinds of Websites, advertising and advertising means on the Web. The intention was to identify frequently used Web functions and design elements. This provided basic data for an elementary Website categorization on how Websites are being used by Web users. Pilot study II was a test of the developed experiments, the various treatments and the on-line questionnaire. It was a live test on a group of 30 respondents with the purpose of revealing problems with questions, concepts, inconsistencies or any other ?aws with the research instrument. Findings from pilot study I was used to develop and to design the experiment and the questionnaire whereas pilot study II had the purpose to re?ne and ?ne tune the main study. The main study comprised of three separate experiments. The experiments were however conducted more or less simultaneously in order to save time. The experiments were followed by a survey to measure the effects of the different treatments. These both parts, the experiment and the survey, were integrated into one single HTML design that was presented sequentially to the respondents.
87
The three experiments in the main study can be portrayed as in ?gure 4.3.
The Research Design
Attention getting A site & A instruction Ad 1 Ad 2 Ad 3 Ad 4 Goal orientation Ad 3 + B-site
Figure 4.2
Task environment
Ad 3 + B-site
Figure 4.2 The research design with its treatments in three dimensions.
Figure 4.2 is visualizing and providing a schematic overview of the dimensions studied in the experiments. The horizontal dimension in the ?gure is targeting the attention getting effects between various advertisements, i.e. ad 1 through 4. Initially this experiment comprised of ad 1 through 3 but was extended with ad 4 (treatment 6), the “frameless ad”. This experiment used Web site “A” and instruction “A”. The vertical dimension in ?gure 4.2 is aiming at the impact of the setting or task environment in which the advertisement is presented. In this experiment the advertisement variable, Ad3, was kept stable and instead the Web environment was altered. In the last dimension, which is diagonal in the ?gure, goal orientation or mode was the main issue. In this experiment the advertisement variable, Ad3, was kept stable as well. At the same time the task environment was kept stable. The experiment was to test different levels of goal orientation manifested through two separate instructions given as treatments.
The design of the experiments
The three experiments in the main study were conducted in order to answer the research questions posed. The reason why experiments were employed was to investigate if any differences between advertisements and Website environments could be found. The expectation of differences is natural when doing experiments. Graziano and Raulin (1993) are stating that the expectation of variation is logical when using experiments – if there is no variation there are no differences to test. Four different Web advertisements and two complete Web sites were set up. The Websites were designed4 in line with how a typical Web site in that particular
4
The Websites were designed by the author together with Web designer Magnus Lindgren and two computer engineers who were solving some speci?c problems. 88
category is designed. The design of the two Websites was guided by information collected in pilot study I. In the same way, the advertisements were designed5 in line with the appearance of a typical Web advertisement. The design of the advertisements was guided by information collected in pilot study I. The experimental Web site was a complete news Web site with the same breadth and scope as an average national Swedish news Web site. The Web site had more than 130 individual Web pages that could be accessed by the respondents. Mullarkey and Danaher (2003) did a study that was similar in some respects as the study at hand. However, in their study the respondents could not surf as one normally does when visiting a Web site. Instead the respondents were restrained to merely watching a Web site that was changing pages automatically with different time intervals. In the experiments conducted here the respondents were able to actually interact and surf in any way they wanted. Thereby this experimental design is an improvement that makes the situation for the respondents more realistic and comparable to a real world situation. The experiment Web site design included one Web version (hereafter called the “A-site”) that could be seen as representative for Swedish news Web sites and a second Web site (hereafter called the “B-site”), actually a derivation of the ?rst one, which was less complex with respect to the search depth. Both Web sites had exactly the same content, news, headlines, name and visual appearance. The only difference was that the B-site had a shallower search depth and was, consequently, less complex. If we take a look at the outside world we can see that the A-site is similar, when it comes to the search depth structure, to for instance the Swedish Web sites “DN.se”, “HD.se” and “SVD.se”, to mention a few. The B-site is more similar to the search depth structure of “Aftonbladet.se” and “Expressen.se”. The search depth system employed in the experiments was thus not arbitrary but an imitation of what is commonly used in real. The Web sites were designed so that they would look similar to real Websites and this was achieved by using the same kind of links, subjects, concepts and general structure as can be expected in a traditional news Web site. The two Web sites had the same color theme, the same background color, the same frame colors and text fonts. Furthermore the same system for links and buttons were used. The Websites were speci?cally designed in such a way that there would be as small differences between them as possible except for the search system and search depth. The search depth was an operational variable to be tested and used as a measure of the complexity construct.
5
The advertisements were designed by the advertising agency 89
In the advertisement experiment, targeting attention getting effects, four different advertisements were used. The advertisements were exhibiting brands, products and offers that were entirely ?ctitious so that there would not be any problems with bias (for instance, that some respondents would recall or recognize the advertisement beforehand while other respondents would not recall or recognize the advertisement). All advertisements had in common that they had the same brand name, picture, headline and copy text. This general advertisement was then used in different ways. Here is a short description of the advertisement: the brand name was “Dentax”, there was an image of a female mouth covering the mid section of the advertisement, the headline stated “För en hälsosammare mun”6, the copy text stated “Nu ?nns den första tandkrämen som motverkar tandsten och samtidigt ger vitare tänder”7. The advertisements can be found in appendix A. The ?rst stimuli, Ad1, was a static version of the general advertisement placed to appear on the right upper part of the screen. The size of Ad1 was 283 times 250 pixels. The second stimuli, Ad2, was a dynamic animated banner advertisement based on the general advertisement. It was placed at the same location as Ad1. The size of Ad2 was 283 times 250 pixels. The difference between Ad1 and Ad2 comprised solely of the animated effect added to Ad2. The animated effect was accomplished by using a java script producing a “slideshow” where 5 images were presented for 3000 milliseconds each. The “slideshow” produced an animation that altered the mouth that was displayed on the advertisement. This created an effect so that it appeared as if the mouth was moving and forming into the shape of a “kiss”. The third stimuli, Ad3, was a dynamic animated banner advertisement based on the general advertisement. The animation was equivalent to Ad2. Ad3 also had pop-up properties distinguishing it from Ad1 and Ad2. Ad3 was programmed to appear in the same area as Ad1 and Ad2. The size of Ad3 was the same size as Ad1 and Ad2, 283 times 250 pixels. Though, since Ad3 was a traditional pop-up advertisement it came with a standard Windows pop-up frame that added 50 pixels on the height and 25 pixels on the breadth. This meant that the pop-up advertisement in fact came with a frame that made it 21 075 pixels large or 29.2% larger than Ad1 and Ad2. This size difference was an unwanted but necessary difference in order to study advertisements of the same size. To control for the size effect just mentioned there was also a fourth advertisement, Ad4, designed. This advertisement, Ad4, was similar to Ad3 since it was a pop-up advertisement. However, it did not have the frame that Ad3 had. The treatment with Ad4 was primarily designed to investigate the effect of abrupt presentation and the effect of the frame presentation.
6 7
In english: For a healthier mouth In english: The ?rst toothpaste that ?ghts tartar and gives you whiter teeth at the same time.
90
The advertisements Ad1 and Ad2 were presented on the Web sites throughout the whole experiment session of 6 minutes. Ad3 was a pop-up advertisement and was launched after 15 seconds, 60 seconds and after 180 seconds. In the second round of experiments the average exposure time was recorded for the respondents subjected to Ad3 and was less than 90 seconds. The exposure time here is stated as “less than” 90 seconds and the reason for this is that the automatically set exposure time for Ad4 was 90 seconds. So as long as the time for Ad3 was less than these 90 seconds we can be sure that the average exposure time for Ad3 has not been in favor of Ad3 compared to Ad4. The hypothesis for the comparison between Ad3 and Ad4 expects Ad3 to be more effective and a longer exposure time for Ad3 would jeopardize the conclusions. An equivalent or shorter exposure time would on the other hand ensure us that exposure time is not a factor here. As a matter of fact, in a majority of cases the exposure time was considerably less than 60 seconds for Ad3 making the exposure time a less important factor in this case. This exposure duration ?ts well with for instance the study reported in the theory chapter where a majority of respondents clicked away pop-ups within 3 seconds after appearance or clicked on some other link or area within 25 seconds making the pop-up disappear. Advertisement Ad4 was also a pop-up advertisement and was automatically launched after 15 seconds, 60 seconds and after 180 seconds, the same time intervals as Ad3. Ad4 was only used in the second round and gave the respondents a total exposure time of 90 seconds. The ?rst experiment aimed at testing the following treatments: Treatment 1 comprised of Ad1 presented in Web site “A”. (static) Treatment 2 comprised of Ad2 presented in Web site “A”. (animated) Treatment 3 comprised of Ad3 presented in Web site “A”. (pop-up) Treatment 6 comprised of Ad4 presented in Web site “A”. (ad without a frame) The internal drive experiment was designed to study how internal drive or goal oriented behavior impacts the effect of attention to advertising. In order to test this hypothesis an experiment was created where the entire design actually was based on a difference in instructions. The respondents were given slightly altered instructions in order to induce an internal drive and to make them more goal oriented. This was accomplished with a few very simple questions that could be answered with a simple yes or no. The instructions that were given can be found in appendix B. The standard instruction that was given is called the “A-instruction” and the alternative, internal drive inducing, instruction is called the “B-instruction”. The group that got the B-instructions was treatment 4.
91
This experiment aimed at testing the following treatments: Treatment 3 comprised of Ad3 presented in Web site “A” with the “A-instructions”. Treatment 4 comprised of Ad3 presented in Web site “A”. (B-instructions, goal oriented behavior) The Web environment experiment was designed to study how the physical task environment or the Web environment impacts the effect of attention to advertising. In order to test this hypothesis the experiment was designed so that two different Web environments, with respect to the search depth structure, were given as treatments to the respondents. The two task environments differed in their complexity and this was the variable to be studied. One environment was less complex, Web site “B”, than the other environment which was more complex, Web site “A”. Hence, the advertisement stimulus was placed in two different contexts. The Web environment experiment aimed at testing the following treatments: Treatment 3 comprised of Ad3 presented in Web site “A” with the “A-instructions”. Treatment 5 comprised of Ad3 presented in Web site “B” (less complex environment) with the “A-instructions”. In the ?rst round of experiments treatment 1 through 5 were run and in the second and extended round of experiments treatment 1, 3, 5 and 6 were conducted.
The main measures used
In order to measure the effects of the various treatments a number of means to assess the effects of the stimuli were employed. The main measures that were used were recognition and recall. The adequacy of using these measures has been elaborated on in the theoretical chapter but in addition to those descriptions the following statement can be added:
“The difference between recall and recognition is that in a recall test a subject is given a set of information and is later given some minimal cue and asked to retrieve and reconstruct the original information, whereas in a recognition test the subject is confronted with the original material and asked whether it has been seen or heard before. Hence, for recall the individual must describe the stimulus which is not present; for recognition the stimulus must merely be identi?ed as having been previously seen or heard” (Singh & Rothschild, 1983, p. 235).
92
In some sense it is “easier” to recognize than to recall and recognition is more important than previously thought (Bettman, 1979). Krugman has also commented on the issue:
“[…] the inability to recall something does not mean it is forgotten or that it has been erased from memory. The acid test of complete forgetting is if you can no longer recognize the object” (Krugman, 1972, p. 14)
The recognition measure is thereby a more sensitive measure that can detect effects where recall cannot. In connection to the experiments both recognition and recall was used to measure the effects but recognition was used as the main measure in the analysis of the data. Recall was measured by asking respondents whether they could recall if they had seen any advertisement on the Web page that they had visited. If they answered yes they were presented with four additional questions where they could write the brand name, headline, copytext and/or describe what the advertisement had portrayed. In order to measure recognition, those who could not recall (and also those who could recall) the advertisement were thereafter presented with images of the target advertisement and four distractor advertisements. The respondents were then asked to indicate if they could recognize any of the ?ve advertisements presented. An additional measure that was used was the click-through frequency. The clickthrough frequency was measured when a respondent indicated in the survey that she had actually tried to click. The reason why there was no possibility for actual click-through to a target Web site was that these respondents would have gotten a different stimulus treatment than the other respondents, had this possibility existed. However, it has been mentioned previously that the clickthrough measure has never been a legitimate measure of advertising effect, which is also one of the reasons why it is being included and studied. In addition to these main measures described above attitude measures were used. These questions were asked using a 7 point likert scale where the respondents could indicate whether they had a positive or negative position towards the statement or question. These questions were carefully developed and ?ne tuned based on input from the tests of the questionnaire.
Selection
The procedure of selecting the respondents for the main study is mainly guided by an interest of targeting individuals with skills relevant for the study. Since the experiments require familiarity and skills in how to use the Web, the Internet and also knowledge of how to use computers, a younger selection was preferred. Even though computer literacy is fairly high in Sweden, younger individuals have greater skills than those who are older. Furthermore we also
93
know that students (compared to non-students) are forced to use computers on a day to day basis as part of the various educational programs. It is therefore more likely that students, as opposed to other groups, will satisfy the basic requirements for the experiments. Support for these notions can be found in other researchers’ studies of the World Wide Web and the Internet. For instance, Gallagher, Parsons and Foster (2001) conducted a series of studies in laboratory environment as well as home environment and used student samples as well as random samples. They also replicated their studies. Their conclusions were that there sometimes are circumstances when a student sample is particularly suitable to use. They state that “…university student samples may provide the best available indicators of where the general population will be when the Web becomes as familiar as traditional media, such as print and television.” With these above arguments at hand it was decided to do the study on a selection of students and in particular students at the Umeå University. By selecting students, by de?nition a convenience sample, for the study there were also additional advantages achieved. A major problem that could be anticipated was a reluctance to participate, i.e. non-response. A random selection of individuals from the Umeå area could have resulted in a nonresponse level well above 90%8. It is not unreasonable to believe that it would have been perceived as quite an obstacle from the respondents’ perspective to travel on average 5 kilometers and sometimes up to 30 kilometers just to participate in an experiment. The concern for a low response level was an additional consideration and concern when it was decided to use students for this study. By targeting students the problem with travel and ?nding the location was assumed to be reduced since students know the locations and frequently visit the area where the experiments were being conducted. Umeå University has around 28 000 students and is located in a community that has the highest Internet penetration in Sweden (Liljenäs and Öberg, 2001) and most likely also in the world. Already in 2001 there were more than 80% of the households that could access the Internet and were using the Internet on a regular basis (Liljenäs and Öberg, 2001), which is a fairly high percentage when comparing internationally. Of these 80% more than 50% of the households are connected to the Internet thru a broad band connection at a speed greater than 256 Kilo bits per second. Typically the transfer speed is 1.0 or 10 mega bit per second and sometimes even as high as 100 megabits per second. Taken together the use of Internet, with high capacity, in Umeå is thereby more widespread than what it is in any other place in Sweden. To study students in Umeå community was thereby a very suitable opportunity.
8
40 phone interviews, with individuals randomly selected from Eniro’s Umeå section, showed that less than 10% were willing to participate. The ?gure would most likely have been even lower if the request to participate would have been sent by letter, which was the initial intention. 94
In order to attract respondents to the experiments posters were put up on bulletin boards and the like at the University. On the posters students were asked to sign up for an experiment in “informatics and social science”. In addition to this a number of classes were approached and information was given to the students that they had a chance to participate in an experiment in “informatics and social science” and that they could sign up for participation. The classes that were approached were from all faculties at Umeå University the faculty of science and technology, faculty of social sciences, faculty of arts, faculty of medicine and faculty of teacher education. The selection comprised initially of 410 respondents but was extended with another 292 respondents. In summary, a total of 702 respondents (about 3% of the students at the University) were selected on the basis that they wanted to participate out of free will. The respondents were informed that they would receive 100 SEK as compensation9.
The experimental setting – the lab
The experiments were conducted in the A wing of the Naturvetarhuset at Umeå University (The building for science and technology)10. At my disposal for the experiments I had an entire corridor with seven identical of?ces and one board room. In ?ve of these of?ces equipment for the experiments were installed. The equipment in every of?ce comprised of 1 Dell stationary computer, 1 keyboard, 1 mouse, 1 mouse pad, 1 IBM 17” CRT screen and the necessary cables. Windows 2000 was installed on the computers as well as the internet browser Internet Explorer 6.0. The interior of the of?ces were similar, a desk with a calendar, a plant in the window, a bookshelf and an of?ce chair. The doors to the of?ces were kept open so that it was possible to observe the respondents. Though, they were not aware of that they were being observed. The corridor where the experiments were conducted was de?ned as restricted area and unauthorized individuals were dismissed. The reason for this was to avoid any unwanted interference with the experiments11. In connection to the corridor with the lab rooms there was a waiting room where respondents could sit down and wait for their turn.
9
In the evaluation of the experiment and the questionnaire the respondents received 50, 70 or 100 SEK. The respondents did however not know that it was an evaluation. 10 The choice of location is perhaps even reducing bias since respondents will not build up expectations about social science and advertising in Naturvetarhuset’s environment 11 At only one occassion the experiments were interrupted. Some chemistry students refused to follow instructions and consequently the experiments were terminated and cancelled for that entire day. They were however resumed the following day. That particular class had been chosen to participate in the experiments but was replaced with another class in order to avoid any negative in?uence on the experiments. 95
Instructions and procedures
First the respondents were welcomed and asked to take a seat in the waiting room. When it was time to start the experiment they were called in from the waiting room and asked to stand in the door way to the of?ce and experiment that they had been randomly assigned to. Thereafter the respondents received instructions that were read to them to them. The instructions were given to the respondents by either the author or by an assistant to the author. The instructions were read straight from a paper without deviation and with the same tone of voice. Since the respondents were Swedes the instructions were given in Swedish. However, the instructions have been translated into English and are reproduced in the appendix. The standard instruction to the respondents can be found in appendix B and is called the “A”-instructions. The goal oriented instruction given to the respondents can be found in appendix B and is called “B”-instructions. After the respondents had gotten the instructions they were asked to take a seat next to the computer. They were asked if they were sitting comfortably. They were also not to do anything before they got the command from the instructor. At the command of the instructor they commenced the experiment. After the experiment they were asked to turn around on their chair so that they would not look at the Website anymore. Thereafter the respondents were given the on-line questionnaire so that they could ?ll it out. When the respondents had ?lled out the questionnaire they submitted their answers to the database and were asked to come out of the lab. Some of the respondents were then chosen to participate in the “Exit poll”. Those who were not chosen for the exit poll went straight to the debrie?ng. All respondents were debriefed after the experiments and the exit poll; see appendix B “Debrie?ng”. The ?nal activity was to give the respondents 100 sek for their participation and they also got to ?ll out some personal information to participate in a lottery and to receive information about the results of the study at a later occasion.
96
The respondents’ journey through the experiment can be described as in the ?gure below.
Procedural Steps for the Experiments
Welcome Instructions Have a seat Start the experiment now! On-line questionnaire 6 minutes Figure 4.3
Exit Poll Debriefing Payment Thank you and have a nice day!
Figure 4.3 Schematic view of the procedural steps for the experiments
The survey
After conducting the experiment on the computer an on-line questionnaire immediately appeared on the screen for the respondents to ?ll out. The on-line questionnaire comprised of a total of 60 questions12. The questionnaire was designed in such a way that 5 to 10 questions appeared at the same time on the screen13. The respondents were to ?ll out these questions and could continue to the next 5 to 10 questions after clicking on a “continue” button. At the end of the questionnaire they pressed on a “submit” button to ?nish the on-line form. The questionnaire took on average 13 and a half minute to complete. The questionnaire is enclosed in appendix H.
The “exit poll”
After the respondents had conducted the survey 100 of them were randomly chosen to participate in an “exit poll”. The reason for this was to follow up some issues of interest for the study. Considering that the advertisement in the experiments was about tooth paste it was of interest to ?nd out to what extent the respondents could be considered as being in the target group. Only 2% answered that they had not bought tooth paste the last 12 months. These 2% stated that a parent was providing this product for them. A second question was whether they perceived the entire Web sur?ng experience to be realistic or not on the scale unrealistic-neutral-realistic. 89 percent stated
12 13
In the questionnaire the questions are numbered up to 63. Despite this fact there are only 60 questions. See Appendix H for an example. 97
realistic, 7 percent stated neutral and 4 percent stated unrealistic. This ?gure corresponds fairly well with the results from the survey. Notice that the questions are slightly different. A third question was asked to respondents that had not recognized the advertisement. These respondents were asked if they had had an intention to click on any advertisement on the Web site. Out of 42 respondents 0 respondents or 0% answered that they had had such an intention. A fourth question was asked regarding how dif?cult or easy they thought it was to “do what they had been instructed to do on the Web site”. The respondents that got the A-instruction responded almost exclusively “very easy” whereas the respondents who got the B-instructions most often responded “easy”. A ?fth question that was asked was what kind of expectations that they had on the experiment and what they thought that it was all about before it started. 4 % thought that it had something to do with business administration and none (0%) expected something in marketing communication (remember the information given when students were invited to participate and also the location where the experiments were conducted). A sixth and ?nal question was asked to hear out the respondents whether or not it had worked out well to conduct the experiment and to ?ll out the questionnaire. All respondents stated that it had gone smooth and without problems.
Validity and Reliability
Validity comprises of a collection of subdivided concepts with the overall purpose to describe or ascribe quality to a study. Validity, in its different guises, are providing a set of instruments to evaluate whether or not the ?ndings from a study can be stated to be valid and reliable in a number of respects. Internal validity Internal validity addresses to what extent we can infer that a relationship between two variables is causal or not (Cook & Campbell 1979 and Calder, Phillips & Tybout 1982). It should be further noticed that when we say “relationship” we are dealing with the relationship between the research operations “irrespective of what they theoretically represent (Cook & Campbell, p. 38, 1979). Thus, internal validity has to do with the relationship between X and Y and not the abstract constructs that X and Y may represent. Nevertheless, when the researcher is estimating the internal validity of a relationship the “investigator has to systematically think through how each of the internal validity threats may have in?uenced the data.” (Cook & Campbell, p. 55, 1979) The process is very much one of ruling out different threats to validity. Among the many threats to internal validity we can immediately rule out a number of them as a consequence of how the experiment was designed and
98
conducted. History effects, Maturation effects, Testing effects, Instrumentation effects, Regression effects, Mortality effects, Ambiguity about the direction of causal in?uence effects, Selection effects, Selection maturation effects and the like are not posing any problems since the respondents were randomized in this study. That is to say that they were randomly assigned to the treatment they received. This position is supported by Cook & Campbell (1979). Randomization makes causal inference easier. Randomization is however not ruling out all potentially harmful effects, for instance, imitation of treatments, compensatory equalization, compensatory rivalry and demoralization in groups receiving less desirable treatments. Demoralization can be ruled out because no one was in the position to know anything about potentially less desirable treatments and even if they would have known anything it is doubtful whether any of the treatments could be perceived as less desirable. Compensatory effects can also be ruled out since there were no reasons for respondents or a group of respondents to compensate in anyway for a treatment that they were subjected to. Actually it is hard to see how compensation in its normal form would take place in this case. The problem of one treatment imitating another treatment can also be ruled out in the same way as the compensatory effects. However, one risk that potentially could have impacted and negatively in?uenced the study is if the respondents violated the instructions and shared information between respondents. The potential effect of this information sharing could however be diminished since no one knew before hand in which treatment they belonged. Moreover, they were not informed of that even after the experiment either14. A speci?c point with the debrie?ng was to inculcate in the respondents the importance of not sharing information about the experiments with others (those who had not yet done the experiment) by stating the following in the last three sentences of the debrie?ng: “Furthermore, please, do not talk about or tell about the content in this experiment to anyone during the next 4 weeks since more than 600 students at Umeå University will participate. It is preferable if those who will participate do not know in advance about the content of the experiment since this will in fact ruin the experiment. The cost for this study is more than 1 million Swedish crowns so it is good if this study can be conducted in a correct and undisturbed way”. (Instructions in the debrie?ng, translated into English) Construct Validity Construct validity is what researchers are concerned with when they are trying to evaluate if the results are confounded or not. Confounding refers to the possibility that “the operations which are meant to represent a particular cause or effect construct can be construed in terms of more than one construct” (Cook & Campbell, p. 59). The example we saw before with
14
The respondents were briefed in a general way what they had been subjected to. 99
Clever Hans highlighted the issue – are we measuring the right thing? What one investigator interprets as a causal relationship between A and B might be interpreted by another researcher as a relationship between X and B or perhaps even X and Y. Measures to increase construct validity and to avoid confounding includes thorough planning and pre-experimental testing so that the constructs are clear and de?ned. Throughout the study this has been achieved by explicitly using constructs and their meanings that are commonly used in the theories used for this type of research. There are no deviations from the research main stream in that respect. The dependent variable, attention, is inherently connected with memory when it is measured with for instance recognition. In fact, both attention and memory are measured at the same time when using recognition as a measure. However, in this case the focus was more on the former than on the latter since measurements were conducted immediately after exposure, making the memory construct less of an in?uencing factor. In advertising research where the focus is on memory the time between exposure and posttest is counted in weeks or months rather than seconds and minutes as it was here. In all attention research, whether it is in advertising, psychology or some other discipline, the researcher will always face the inherent problem with the attention – memory constructs when using recognition as a measure. What can be said is that the problem has been observed and was reduced to the point where no additional countermeasures could be undertaken, and still use recognition as a main measure. The reason why the experimental design was integrated into one solution was because of this issue. It was thereby possible to reduce the time between the exposure and the following recognition test to an absolute minimum. When measuring the context’s in?uence on attention, two versions of contexts are used, i.e. the internal drive and the Web environment. Both the internal drive and the Web environment are facets of the same context dimension but they are expected to impact cognitive resources in a similar way. That is, they may both use more or less cognitive resources depending upon how demanding the situation is. The results in the analysis are indeed showing that these two measures of the context dimension can use either more or less cognitive resources as expected. Using two or more dimensions that are targeting the same construct increases the likelihood of achieving adequate results, especially if the constructs are co-varying (Churchill, 1979). From the analysis we can see that the measures co-varied and that they also behaved as expected, criterions that are positive indicators of construct validity (Churchill). In the case of studying attention getting techniques four advertisements are used for the study. These advertisements are supposed to re?ect contemporary attention getting techniques on the World Wide Web. That these advertisements are good representations of and thereby also measuring various levels of
100
attention getting is merely based on the mimicking of the function and visual appearance of common advertisements used in the 100 Web sites that were studied in the ?rst pilot study. Using these types of advertisements is in line with how attention getting has been researched in other studies. External Validity External validity is focusing on whether a causal relationship can be generalized to and across different persons, settings and times (Calder, Phillips and Tybout, 1982 and Cook & Campbell, 1979). Considering that a student sample is used it is appropriate to discuss how this may impact the ?ndings. Gallagher, Parsons and Foster (2001 A, 2001 B) conducted a series of studies in laboratory environment as well as home environment and used student samples as well as random samples. Their conclusions were that there sometimes are circumstances when a student sample is particularly suitable to use. They state that “…university student samples may provide the best available indicators of where the general population will be when the Web becomes as familiar as traditional media, such as print and television.” and furthermore “In these cases student samples are superior to samples of the general population, even when researchers wish to generalize the results to the larger population.” (Gallagher, Parsons and Foster, 2001 A, p. 81). Due to the fact that university students are experienced users make them more suitable for testing than those who are not familiar with handling computers, Web browsing etc. An additional argument in favor of the appropriateness of using university students is that, in Sweden where the study took place, about 50% (SvD, 2003) of the young population is continuing with academic studies. This means that “students” is not a small minority group that may deviate signi?cantly in their behavior. On the contrary, half of the population is “represented” when using student samples. Calder, Phillips and Tybout (1982) go one step further than Gallagher et al. when arguing in favor of using student samples. Their point is that if a homogeneous group is used for the study then there will be less variation in background variables within that group. They in fact state that “Random sampling … may actually interfere with achieving a severe theory test. This is because it is likely to increase error variance” (Calder, Phillips and Tybout, 1983, p. 241). Using a more homogeneous sample will have the effect that whatever phenomena that is under study may be more protruding or at least not obscured due to increased variation within the sample. They are furthermore stating that controlling, rather than manipulating and examining the background factors should be encouraged. Some proponents of the importance of external validity have argued that research that is weak in external validity is not considered being an adequate test of theory (Lynch, 1982). Cook and Campbell are though taking a different stance and in line with Calder et al. stating that:
101
“The priority among validity types varies with the kind of research being conducted. For persons interested in theory testing it is almost as important to show that the variables involved in the research are constructs A and B (construct validity) as it is to show that the relationship is causal and goes from one variable to the other (internal validity). Few theories specify crucial target settings, populations, or times to or across which generalization is desired. Consequently, external validity is of relatively little importance. In practice, it is often sacri?ced for the greater statistical power that comes through having isolated settings, standardized procedures, and homogeneous respondent populations. For investigators with theoretical interests our estimate is that the types of validity, in order of importance, are probably internal, construct, statistical conclusion, and external validity.” (Cook and Campbell, 1979, p. 83):
Considering these convincing lines of argumentation a student sample was employed for the studies conducted in this research. Transferability Even though external validity was considered the least important to Cook and Campbell this did not discourage the author to put in considerable effort to reassure the transferability aspect. It was considered important that the experiments would be good representations of hands-on experiences in the out side world (outside of the lab). This meant that the experiments were constructed to be as realistic as possible. As a matter of fact, many of the respondents asked informally if it was a new news Web site that was used for the experiments and where they could ?nd and use it at a later point in time. This kind of inquiry could perhaps be interpreted as a quality mark for the experiments’ realistic design. Czarniawska (1999, p 7) argues that the transferability of an experiment to any other kind of social reality is a matter of whether the reality of the experiment have traits in common with these other realities. Hence, the issue is not whether the experiment is considered “unreal” since the experiment as a matter of fact constitutes a reality in itself, and thereby it is real. Instead it is a question of how many real situations that this tampered reality is a good representation of. That is though a question that can best be answered in relation to the properties of similar or dissimilar realities. In order to get an understanding for the phenomena in question it was decided to include a measure of realism. The respondents were asked to give their subjective view of how realistic they subjectively perceived the experiment to be, as a representation of a real world Web site experience. Their responses to this measure was that 94.5% of the respondents thought that the Web sites used in the experiments had a high or very high resemblance of real Web sites. Another 4.8% thought that the resemblance was neutral or slightly dissimilar whereas merely 0.6% or 4 respondents thought that the experiment Web sites
102
had little resemblance with real Web sites. Not one single respondent, of the 702 that was included, ticked the alternative that the experiment Web sites had very little resemblance with their out side world counterparts. These ?gures are assuring a high level of transferability as discussed in the above. Reliability Reliability refers to how stable results are from one test to another. If results would differ between tests conducted at different times then the measures are unreliable or suffer from low reliability. Low reliability in?ates standard errors and reduces statistical power, such as ?nding differences between means of different treatment groups (Cook & Campbell, p. 43). Often it is dif?cult, when conducting research, to point at how reliable a certain study is or how stable it has been over a period of time. This issue can sometimes not be addressed since the appropriate measures are absent. Though, in the present study it was decided to run a second round of experiments, which gave an opportunity for a test – retest measurement. Instead of merely enlarging the study with the new parts, it was decided to include the old treatments as well. This made it possible to see whether the results were stable. In the ?rst round treatments 1 through 5 were conducted and in the second round treatments 1, 3, 5 and 6 were conducted. Thereby the new results from treatment 1, 3 and 5 could be compared with the older results. The ?ndings from the analysis of reliability revealed that the measurements were indeed very stable since the differences within treatments were low. The difference in treatment 1 was 2.4% between test and retest, the difference for treatment 3 was 2.9% and the difference for treatment 5 was as low as 0.3% when measuring the dependent variable. These ?gures give an indication of high reliability and ensure stability over time. This also made it possible to merge the data into one data set in the analysis. The time between the test and the retest was 1 week.
Analysis
A major part of the work that was conducted in this study was the statistical analysis that took place after the data was collected. Data was collected through the on-line questionnaire and stored in a database. Data was then compiled into data ?les to be analyzed with the software package SPSS version 13.0. The software package is a sophisticated tool for statistical analysis in general and multivariate techniques in speci?c. Three different means for statistical analysis were used; Chi-square, Logistic Regression and ANOVA. ANOVA and Logistic regression will be described in the following.
Logistic Regression Analysis
The classical linear regression model can be used as a starting point in deriving the logistic regression model. As we know the linear regression analysis is
103
aiming at predicting the dependent variable with one or more independent variables as predictors. In traditional linear regression the dependent variable is continuous in nature and the independent variables can be both continuous and dichotomous in nature. The ordinary linear regression model is typically written in the following way: y= ?+?*X+e where y is the dependent variable to be predicted, alpha is the intercept, beta is the regression coef?cient and X is the independent variable. The term e is a random disturbance term with mean zero and variance ?2. This model can easily be extended to a multiple regression with the following formulae: y=?+?1X1+ ?2X2+... ?kXk+e The multiple regression is adding additional independent variables and their respective betas. However, both the simple as well as the multiple regression are adequate when using continuous or metric variables, but they are less appropriate when the dependent variable only can assume either of two values like categorical variables. In a logistic regression the dependent variable does not have to be continuous. Due to the underlying formulae the logistic regression is appropriate in particular when the dependent variable is dichotomous. For instance whether an individual owns stocks or not is a dichotomous variable. This can for instance be coded as 1 for individuals who own stocks and 0 for individuals who do not own stocks. Now, the dichotomous property is making it somewhat problematic to use either of the linear regression models above. The inherent problem with the dichotomous dependent variable, is that it can only vary between 1 and 0 when the classical linear regression model assumes a dependent variable that can vary between -? and +?. To solve this problem a logarithmic link is used. This link is called the logit function and it is the most popular linking function (Sharma, 1996). The logit function is a nice solution where log of odds are used since log of odds can vary between -? and +? and are symmetric, contrary to odds that can only range from 0 to +? or probabilities that range from 0 to 1. By using log of odds this link is added thereby forming a logistic regression. The logistic regression expression takes the following form: P(y=1)=1/e –(?0 + ?1X1 +
?2X2 + ... ?kXk)
Another way to formulate the model is: log [P(1)/P(0)] = ?0 +?1X1 +?2X2 + ...?kXk
104
When running a logistic regression in SPSS a considerable amount of output can be produced but for the purposes in this study the main interest is the calculated odds for the events that have been studied. Since odds and also log of odds are the measures being used in logistic regression these measures will be used in the analysis and presentation of results. A small recap of odds and their relation to probabilities is perhaps in place since odds are not as frequently used in Sweden as for instance in the U.S. Probabilities can range from 0 to 1. When stating that there is a “?fty-?fty” probability for the event Q to occur, then this means that the probability that the event Q will occur is 0.5 or 50% and the probability of non-occurrence is 0.5 or 50%. If this is transformed into odds it is adequate to say that the odds are 1:1 for the event Q to occur. This implies that the odds that the event Q will not occur are 1:1. A different probability, for example 0.66 or 66% for the occurrence of Q, will produce a different odds. Then the odds for the occurrence of Q are 2:1. A probability of 0.75 or 75% gives the odds 3:1 and so forth. A straightforward way to calculate the relationship between probabilities and odds can now be formulated: Oi=Pi/(1-Pi) When reading the diagrams, which essentially is output material produced by SPSS, in the analysis it is advisable to keep in mind that the logistic regressions are not primarily used for prediction. Instead, what are being aimed at are the differences in odds between treatment groups and whether this difference is signi?cant or not.
Analysis of Variance - ANOVA
A second tool for analysis that was used in this study was analysis of variance. It is often desirable to measure differences between treatment groups. The researcher may for instance be interested to see if there is an effect of a certain treatment. One way to go about is to measure the difference in means between the treatment group and the control group. This can for instance be done with an independent samples t-test. A problem that will surface pretty soon is though if there are many groups to be compared. The independent samples ttest is suitable for situations where the researcher is interested in whether two means differ. If there are more than two groups the researcher would have to conduct multiple pair-wise tests since comparing three or more groups at the same time is not possible with the t-test (Howell, 1997, p. 299). At the same time, conducting multiple t-tests may result in an increased risk of incorrectly rejecting a null hypothesis of no difference among groups. The problem with multiple t-tests is that as the number of groups increases, so will the likelihood of ?nding a difference between any pair of groups, simply by chance, i.e. when no real difference exists. Hence, the risk for making this error (a type I error or ?) increases as the number of comparisons made increases (Chu, 1999). One possible solution would be to use a Bonferroni correction, which means that the signi?cance level is divided with the number of tests conducted (Chu).
105
An additional and more appropriate way to go about is to use analysis of variance, ANOVA. The Anova analysis is widely used and allows to measure whether three, four or k means differ (Howell, 1997). This kind of analysis makes it redundant to conduct multiple pair wise comparisons and is thereby reducing the risks for errors related to the increased numbers of comparisons. When using the Anova to compare multiple means we are actually investigating whether there are no differences between the treatments or if there indeed are differences. If there is a difference in one or more means then we can state, with some degree of certainty, that this is a treatment effect. That is, an effect related to the various treatments distributed to the various groups. This is the exact way that this analysis technique is being used in the analysis section.
Operationalization of Hypotheses
In order to appropriately answer the research questions posed, the hypotheses generated in the previous chapter have to be examined in some way. The hypotheses do not say much about how they are going to be tested at an operational level. They are merely statements that are pointing in a certain direction. These statements have to be tied to speci?c measures and in particular to parameters that can be tested statistically. The hypotheses are operationalized by transforming the statements into parameters that can be tested. In this case it is the exponential betas of the logistic regression that are being tested, Exp B. Recall the expression stated for logistic regression: 1/e –(?0 + ?1X1 + ?2X2 + ... ?kXk). In the following the hypotheses H1 and H8 are operationalized and formulated in the way that they will be tested. Hypotheses H2 through H7 are operationalized following the same pattern as H1. The display of H2 – H7 is therefore redundant. In the hypotheses chapter hypothesis H1 together with the null hypothesis for H1 was formulated in the following way: H1: A situation where respondents are exhibiting a less goal oriented sur?ng behavior will elicit more attention to an advertisement than what a more goal oriented sur?ng behavior will. To be rejected; [H0H1 A situation where respondents are exhibiting a less goal oriented sur?ng behavior will elicit equal or less attention to an advertisement than what a more goal oriented sur?ng behavior will.] In the statistical model to be tested in connection to hypothesis 1, attention is the dependent variable, the dichotomous y-variable. The independent xvariable is the treatments which are treatment 3 and 4 for hypothesis H1. In the statement in H1 it is expressed that in a “situation where respondents are
106
exhibiting a less goal oriented sur?ng behavior”, which they are expected to show in treatment 3, “will elicit more attention to an advertisement than what a more goal oriented sur?ng behavior will”, which they are expected to show in treatment 4. The Exp B - exponential beta - is expected to be greater than 1 if there is a larger effect in treatment 3 compared to treatment 4. The input data in the model is such that; the y-variable is either 0 (for non attention) or 1 (for attention) and the x-variable is 0 (for treatment 3) and 1 (for treatment 4). The model used in the statistical test is then; Logit(P1/P0)=a+BX. When hypothesis H1 is operationalized it is expressed in the following way: H1: Exp B > 1 H0H1 Exp B = 1 The reason why H0H1 is not expressed as Exp B ? 1 is because when conducting the statistical test then H1 is in fact compared with a point in H0 where no difference between treatments is assumed, i.e. Exp B for the two treatments are equal15. The test of hypothesis 8 is a test of multiple treatments means using analysis of variance. The variable to be tested is “annoyance” that is expected to increase as attention getting increases. It is this direction of the relationship that is expressed in the hypothesis. H8: The risk for an advertisement to be perceived as annoying will be greater when more attention getting techniques are employed in the advertisement. This would mean that a static advertisement will be the least annoying, an animated advertisement will evoke intermediate annoyance and a pop-up advertisement will be the most annoying. It would be preferable if hypothesis H8 could be operationalized in the following way: [H8: µs
Commercial advertisers often seek to generate increased consumption of their products or services through "branding," which involves associating a product name or image with certain qualities in the minds of consumers.
Attention to Advertising
ATTENTION TO ADVERTISING
AKADEMISK AVHANDLING
vid Samhällsvetenskapliga fakulteten Handelshögskolan i Umeå UMEÅ UNIVERSITET
Carl Patrik Nilsson
Framlägges för vinnande av ?loso?e doktorsexamen till offentlig granskning vid Handelshögskolan vid Umeå Universitet, Hörsal G, Humanisthuset, Fredagen den 10 november, 2006, kl. 13:15.
Attention to Advertising
Carl Patrik Nilsson, Umeå School of Business, Umeå University SE-901 87 Umeå,Sweden
Abstract
In advertising, capturing consumers’ attention is imperative. Yet, how attention to advertising works in a World Wide Web context has not been researched extensively. This area is the main focus of this research. The World Wide Web is an interactive medium which has characteristics that deviate from traditional media channels. Considering the unique properties of the World Wide Web it was deemed important to study how attention to advertising is affected by various contexts as well as attention getting techniques. The study was conducted in order to develop an understanding of how advertisements in different Web task environments are being attended to as well as how attention to advertisements varies between different attention getting techniques that are being used in Web advertising. Furthermore, a model which describes the relationship between context, attention getting techniques and attention to advertising was developed and tested. In connection to this it was tested whether click-through or click-rate is indeed an adequate method when measuring advertising effect and thereby to what extent the pay per click pricing strategy is appropriate. In order to address the research issues a theoretical framework was assembled. Nine hypotheses were formulated deriving from the framework. To test the hypotheses and the model, an experimental research design was employed. Three experiments were designed to study the hypotheses formulated. 702 individual experiments were conducted. The conclusions from the study reveal that Web task environments indeed have an impact on attention to advertising. Increased complexity or a more dif?cult task environment demands more cognitive resources which in turn produces lower attention levels to the advertising stimulus. Reduced complexity, on the other hand, releases cognitive resources that can be spent elsewhere. Thus, a higher attention level to the advertising stimulus was recorded. This also con?rmed the properties of the proposed model; namely that the permeability of a ?ltering mechanism is dependent on for instance level of goal orientation and complexity of the physical task environment i.e. the complexity of the Web site. The study revealed that attention getting techniques such as pop-up advertisements increase the attention to advertising. However, it was found that pop-up advertisements are effective, not mainly because of their abrupt presentation, but because of the distinct properties of the frame. It was found that the frame has a negative meaning for Web users and when the frame comes into the visual ?eld it will attract their attention. At the same time attention will also be distributed towards the advertisement itself. This is recorded as an increase in attention towards the advertising message. The results show that the click-through measure is not an appropriate method when measuring advertising effect. The click-through measure may severely underestimate the advertising effect. Consequently, using click-through frequencies as a basis for pricing is inappropriate, unless the effects at play have been well understood and taken into consideration. Key words: Attention to advertising, attention getting techniques, Web advertising, Web environment, click-through Studies in Business Administration, B series, No 60. ISSN 0346-8291, ISBN 91-7264-189-4 Umeå School of Business - Sweden 2006
Attention to Advertising Studies in Business Administration, B series, No 60, 2006 ISSN 0346-8291 ISBN 91-7264-189-4 © Patrik Nilsson 2006 Stockholm 2006-10-10 Font: Sabon & Avenir Cover Design: Björn Hollström, Honeymoon Sthlm AB Graphic Design: Robin Bergström, Honeymoon Sthlm AB Printed in Stockholm, Sweden by Tryckbolaget All rights reserved. Except for the quotation of short passages for the purposes of criticism and review. No part of this publication may be reproduced, stored in retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior permission of the publisher. This material is protected under the Swedish copyright act (1960:729) and international equivalents.
ACKNOWLEDGEMENTS
This doctoral dissertation is the end product of a long voyage. It has been a privilege to cooperate with a number of organizations, colleagues, friends and loved ones. I would like to express my gratitude for their contribution and support. I wish to thank my main tutor, Professor George Tesar. I highly appreciate your commitment and help and that you set aside time to give feedback, whenever needed. I will, for instance, always remember our last early morning meeting at Arlanda Airport where you made time to discuss the thesis between ?ights. It was an inspiring meeting. I also wish to thank my tutor, Professor Anders Söderholm who has motivated and believed in my study. This motivation was particularly important towards the end of the project when there were still loose ends to tie together. My gratitude also goes to my third tutor, associate Professor Agneta Marell, who also served as opponent at the internal seminar. Together with two additional opponents, Vladimir Vanyushyn and Timo Mäntylä, insightful comments and feedback were given that have helped me improve my work. Timo was also very helpful at earlier stages of the project and contributed with important ideas and comments. Thanks to all of you. I am also indebted to my friend Thomas Laitila who has been available for meetings, discussed various topics and given fast feedback. Thanks for your very kind help and that I was able to disturb you every now and then, especially when I was working on the ?nal analysis. Furthermore I would like to thank Owe R. Hedström who inspired me to come back to Umeå and conduct research. I have never had a reason to regret my decision. Owe, thanks for your personal support and for being such a great colleague and friend. Even though I have designed all the experiments myself it is handy to get the help of specialists every now and then, such as system managers, art directors, designers and programmers. Therefore; Björn Hollström, Michael Lee Bjerkö, Johan Jönsson, Fredrik Lindgren, Mattias Johnson, Johan Tufberg and Magnus Lindgren, thank you very much for helping out with various parts of the experiments. Thanks also to all the respondents. Without your participation not much would have been accomplished.
In the following is a list of those who have, in various ways, contributed to the ?nalization of this project. Please accept my sincere gratitude. Roland Carlsson, Umeå Kommuns Näringslivsservice Barbara Cornelius, Professor, Umeå School of Business - USBE Owe R. Hedström, Director of Studies, Umeå School of Business - USBE Anders Hellquist, System Manager, Umeå School of Business - USBE Ola Henfridsson, Associate Professor, Viktoria Institute Björn Hollström, CEO, Honeymoon Sthlm AB Linus Holm, Ph.D student, dept of psychology, Umeå University Jonny Holmström, Professor, dept of informatics, Umeå University Mattias Johnson, System Manager, Umeå School of Business - USBE Johan Jönsson, Mathematician and ICT specialist Thomas Laitila, Professor, Örebro University and SCB Michael Lee Bjerkö, Design Manager, Honeymoon Sthlm AB Per Levén, Assistant Professor, UCIT and dept of informatics, Umeå University Henrik Linderoth, Assistant Professor, Umeå School of Business - USBE Fredrik Lindgren, IT Specialist, Volvo Corporation Magnus Lindgren, Programmer - IT Specialist, Teknikhuset Agneta Marell, Associate Professor, Umeå School of Business - USBE Timo Mäntylä, Professor, dept of psychology, Umeå Univeristy G.M. Naidu, Professor of Marketing/Professor Emeritus, University of Wisconsin - Whitewater Karl-Gösta Nilsson, former U.N. envoy and former Lecturer in marketing Katarina Pousette, Executive assistant, Umeå School of Business - USBE Claudia A. Rademaker, Project Manager, Berghs School of Communication Anders Söderholm, Professor, Umeå School of Business - USBE George Tesar, Professor, Umeå School of Business - USBE Johan Tufberg, Webmaster, Umeå School of Business - USBE Vladimir Vanyushyn, Ph.D student, Umeå School of Business - USBE Rickard Wahlund, Professor, Stockholm School of Economics Thanks to my family and in particular to my father, Karl-Gösta Nilsson, who has always inspired and encouraged me to pursue my goals. Last but not least, my gratitude goes to my wife Claudia and our daughter Raquel. Your support has been invaluable. Claudia, thanks for reading innumerable versions of the manuscript and for coming with input at all times. Stockholm, October 2006 Carl Patrik Nilsson
In memory of Frank Rademaker
1941 - 2006
Contents
Chapter One Attention to Advertising .......................................................................1
Introduction .................................................................................................2 Why study Web advertising ..........................................................................3 A background and knowledge platform .......................................................4 Placing Web advertising in a broader context ...............................................5 The Internet, the World Wide Web and contexts ..........................................6 Advertising and attention getting on the WWW .........................................11 Modeling attention to Web advertising ......................................................14 Measuring advertising on the Web and pricing methods ............................15 Outline of the dissertation ..........................................................................19
Chapter Two Theoretical Framework.......................................................................20
The origin of advertising ............................................................................21 What is advertising and what is a Web advertisement?. ..............................22 The Web is growing important ...................................................................24 Advertising is taking place in a setting........................................................25 Outcomes of advertising.............................................................................28 Putting the parts together ...........................................................................31 Central concepts in advertising...................................................................32 Models describing advertising ....................................................................34 The World Wide Web as an advertising vehicle ..........................................42 Web advertisements are placed on the World Wide Web ............................43 The advertisement. .....................................................................................46 The Web dimension of advertisements .......................................................49 The concept of attention. ...........................................................................54 How the attention system operates.............................................................58 Advertising, attention and media................................................................63 Measuring cdvertising - click-through and pricing......................................66 This chapter ...............................................................................................69
Chapter Three Formulating Hypotheses .....................................................................70
First section - attention effects of context ...................................................71 Second section - advertisements and their respective attention effect ..........72 Second section extended .............................................................................74 Modelling attention to Web advertising .....................................................77 Third section – measurement and pricing ...................................................78 Other effects regarding Web advertising .....................................................79 Summary ....................................................................................................80
Chapter Four Research Method ..................................................................................81
Approach ...................................................................................................82 The essence of an experiment .....................................................................83 Overall design ............................................................................................87 The design of the experiments ....................................................................88 The main measures used.............................................................................92 Selection .....................................................................................................93 The experimental setting – the lab. .............................................................95 Instructions and procedures .......................................................................96 The survey. .................................................................................................97 The “exit poll”. ..........................................................................................97 Validity and reliability ................................................................................98 Analysis....................................................................................................103 Logistic regression analysis.......................................................................103 Analysis of variance - ANOVA .................................................................105 Operationalization of hypotheses .............................................................106
Chapter Five Results and Analysis ..........................................................................108
First section ..............................................................................................109 Second section ..........................................................................................111 Impact of external context factors on attention ........................................113 Third section. ...........................................................................................114 Extending hypothesis 5 ............................................................................120 Modeling attention to Web advertising ....................................................125 Fourth section. .........................................................................................126 Collateral effects of Web advertising ........................................................129 Control ....................................................................................................132
Chapter Six Conclusions and Further Research ................................................137
Attention to advertising on the Web. ........................................................138 Attention to advertising differs between attention getting techniques. ......140 A model of attention to advertising on the Web .......................................142
Collateral effects of attention getting........................................................144 Measuring advertising on the Web ...........................................................146 Pricing methods on the Web .....................................................................148 Future research.........................................................................................150
Svensk Sammanfattning ....................................................................152 References.............................................................................................154 Appendices ...........................................................................................171 Figures
Figure 1.1 .....................................................................................................5 Figure 1.2 .....................................................................................................8 Figure 1.3 ...................................................................................................19 Figure 2.1 ...................................................................................................23 Figure 2.2 ...................................................................................................24 Figure 2.3 ...................................................................................................28 Figure 2.4 ...................................................................................................33 Figure 2.5 ...................................................................................................34 Figure 2.6a .................................................................................................36 Figure 2.6b .................................................................................................36 Figure 2.7 ...................................................................................................38 Figure 2.8 ...................................................................................................41 Figure 2.9 ...................................................................................................47 Figure 2.10 .................................................................................................50 Figure 2.11 .................................................................................................59 Figure 2.12 .................................................................................................60 Figure 2.13 .................................................................................................61 Figure 2.14 .................................................................................................64 Figure 2.15 .................................................................................................65 Figure 3.1 ...................................................................................................78 Figure 4.1 ...................................................................................................87 Figure 4.2 ...................................................................................................88 Figure 4.3 ...................................................................................................97 Figure 5.1 .................................................................................................126 Figure 6.1 .................................................................................................139 Figure 6.2 .................................................................................................143
Tables
Table 5.1 ..................................................................................................109 Table 5.2 ..................................................................................................110 Table 5.3 ..................................................................................................112 Table 5.4 ..................................................................................................114
Table 5.5 ..................................................................................................115 Table 5.6 ..................................................................................................117 Table 5.7 ..................................................................................................118 Table 5.8 ..................................................................................................120 Table 5.9 ..................................................................................................122 Table 5.10 ................................................................................................123 Table 5.11 ................................................................................................127 Table 5.12a ..............................................................................................128 Table 5.12b ..............................................................................................128 Table 5.13 ................................................................................................130 Table 5.14 ................................................................................................131 Table 5.15 ................................................................................................133 Table 5.16 ................................................................................................134 Table 5.17 ................................................................................................135 Table 5.18 ................................................................................................135
Chapter One
Attention to Advertising
This ?rst introductory chapter will describe the area to be researched. Concepts and theoretical areas key to the study will be elaborated. Subsequently, the purpose of this study will be pointed out in greater detail. At the end of this chapter, an outline of this dissertation will be presented.
1
Introduction
A number of media channels exist that can be used for conveying advertising messages. To the well-established channels such as TV, radio and print, the World Wide Web has been added. As much as there are differences between, for instance, print and TV there are reasons to believe that the World Wide Web has unique or special features that differentiates it from other media channels. The World Wide Web has characteristics that deviate from traditional media channels, making the World Wide Web a prime target for investigation. One signi?cant disparity, which is often pointed out, is that the World Wide Web represents an interactive media with a greater richness of information as opposed to what can be found in traditional media (Evans & Wurster, 1997). Researchers have found that consumers are more active on the Internet and the World Wide Web than is the case with other media channels (Hoffman and Novak, 1996). During a Web session, the Web user is active and interacting with the material presented on the Website, a behavior that may impact advertising on the Web. A main property distinguishing Web advertising from other kinds of advertising is that it has hybrid characteristics combining properties from print, broadcast, out-door, and direct response media (Dréze and Zufryden, 1997). In addition, Karson and Korgaonkar (2001) are pointing out no less than eight unique characteristics of which four1 will be mentioned here. As mentioned earlier, one of these characteristics is Interactivity, a second one is Complexity - Web sites exhibit a greater complexity than both print and broadcast media (Bruner & Kumar, 2000), a third is Flexibility – this type of media is very ?exible, a property making room for creative advertising. Web advertising can range from static to full motion audio-video, which cannot be accomplished in print advertising. A fourth characteristic is Usage - unlike broadcast media, the Web is being used in a number of ways such as a channel of distribution of digital products, creation of Web communities (Karson and Korgaonkar, 2001), for Web logging, Web shopping and Web chatting to mention a few. These four characteristics differentiate the Web environment context from the context dimension in print and other media channels. When summarizing the above, it is reasonable to state that the World Wide Web is distinctly different from other types of media channels with regard to the environment in which the advertisement is being placed. Properties such as greater interactivity, complexity, usage and ?exibility create this difference. Furthermore, the advertisements on the World Wide Web can be propelled to stimulate and attract attention in an additional and different way than what is the case in, for instance, print media. Eliciting or attracting attention can
1
The other four that are not mentioned here are Variability, Accessibility, Salience and Stand alone. These properties are not included here since they do not have the same signi?cance for the uniqueness of the World Wide Web today as they may have had in the past. 2
be accomplished through various kinds of strategies and techniques such as animation, ?ash, size changes or pop-up. Taken together there is an array of unique features whose impact on advertising on the World Wide Web is yet to be investigated and it is this knowledge gap that is the very target of the research at hand. Consequently, the over arching research problem is how attention to advertising is affected by various contexts as well as attention getting techniques.
Why study Web advertising
Studying how the context impacts advertising has been conducted previously (see for instance; Bruner and Kumar, 2000, Moore, Stammerjohan and Coulter, 2005, Dahlén, 2005, Janssens and De Pelsmacker, 2005) but never, to the best of the author’s knowledge, where structural differences between Web environments have been compared. Appel (2000) reviewed eight studies of the editorial environment in print in the United States of which none gave strong support to the environment hypothesis. According to Appel a number of investigators have studied the issue but none has “demonstrated the validity of the concept … to the point where it is generally accepted” (Appel, 2000, p.94). Even though Appel’s statement is somewhat discouraging, it was at the same time challenging and inspired to take on the task in a Web context to seek a convincing answer to the issue. Since the World Wide Web differs from traditional media channels with respect to its interactivity, ?exibility, complexity and usage it was considered plausible to expect greater differences between environments on the Web. Attention getting strategies and techniques has been studied earlier for print and broadcast media. At the time when this research project was initiated, in 2002, there were no known studies of this subject in Web advertising. As late as 2004 Diao and Sundar (2004) wrote that there was little or no empirical research that had been conducted in the area. Recently, studies have been done where attention getting in Web advertising has been investigated. In the Eyetrack III study (2004), attention getting techniques was studied. In this study, different backgrounds have been used when various advertisements were tested making it hard to speci?cally attribute the effect to either the difference in context or the difference in attention getting technique. Similarly, when comparing for example the pop-up effect with animation, advertisements of different sizes were used making it impossible to know whether the registered effect originated from a difference in size or a difference in technique. Consequently, there is still a need to scienti?cally investigate the scope of the special attention getting techniques that Web advertising offers. In an editorial by Kover (2000) he expressed concerns of how to measure advertising on the World Wide Web. His concern may have been related to the Web industry’s practice to use click-through as a main measure of advertising effectiveness. Even though this practice is in con?ict with how advertising is being measured in other media channels than the Web, the click-through
3
measure is the most widely used method (Chandon, Chtourou and Fortin, 2003). Given this discrepancy and the signi?cance of employing correct measurement methods it was deemed important to study this issue in this research. In connection to the click-through measure, a pricing strategy based on this measure has evolved namely, the pay per click pricing strategy. The adequacy of this strategy can be challenged, especially when taking the following logic into consideration; if the click-through measure is inappropriate when measuring advertising on the Web then it may be inappropriate as a basis for pricing as well. Therefore, this research issue was incorporated into the study.
A background and knowledge platform
To be able to comprehend the relationship between various phenomena in Web advertising, it is advisable to obtain a brief overview of the Web advertising area. Hence, a knowledge platform is presented in the following. Advertising on the World Wide Web has developed rapidly over the last decade. Initially, advertisements were designed as simple one colored banners for Web users to click on. The reason for why banners were uncomplicated was because the limited bandwidth of the Internet did not allow the use of sophisticated graphics. More elaborate graphics affect the size of an advertisement which in turn affects the amount of information that has to be transferred to an individual user who is sur?ng a Web page. With the modems that were frequently used just a few years ago it took excessively long time to download a Web page if the graphics were very sophisticated. As a result, most graphics on Web pages, including advertisements, were of a basic character. However, as the infrastructure of the Internet has been upgraded with a greater bandwidth, advertisers have re?ned the graphical design (Zufryden and Dréze, 1997) leading up to today’s full ?edged Web advertisements. The old Web banners typically carried a very short or incomplete message that often ended with the two words “click here” or “click now” that was blinking to call upon the attention of the Web surfer. The purpose with these banners was to generate click-through to a target Web site where the actual good or service was promoted in greater detail. The Web advertisements of today, contrary to the early banners, often have some basic properties such as an image or description of a product, a brand name, copy text and a headline put together into an advertisement entity – the Web advertisement. Modern Web advertisements are quite different compared to their older counterparts. They have a different visual appearance and also a different purpose since they are communicating a complete message rather than aiming at generating click-through to a target Website. These re?ned Web advertisements in particular, that are quite similar to print advertisements, are to be researched in this dissertation.
4
Placing Web Advertising in a broader context
Marketing communication is an important part of marketing with the purpose to communicate with companies’ target markets2. Marketing communication can be described as representing “the voice of a brand and the means by which companies can establish a dialogue with consumers concerning their product offerings” (Keller, 2001, p. 823). The reasons for why companies communicate to their markets vary although often the purpose is to bring to customers’ attention information regarding new goods and services, to change attitudes toward a product or to remind consumers about products (Keller, 2001). The companies’ communication activities can furthermore be part of an information exchange between company and customer with the intention to maintain and develop customer relationships (Reid, Luxton & Mavondo, 2005). Regardless of the purpose, communication is an integral part of the marketing function and is executed by staff and functions within the organization and/or together with external organizations specialized in the area of communication. Marketing communication, frequently called promotion and being one of the four P’s3, is a broad concept that covers the sub areas of advertising, personal selling, public relations, sales promotion, direct marketing and areas connected to these (Duncan & Moriarty, 1998). The various areas are serving different purposes and aiming at achieving different goals. For instance, sales promotion is short term incentives to stimulate purchase whereas advertising can be used to build a brand image or to create a presence on a new market. Common for these sub areas is that they all include an element of communication that is directed towards the market. Communicating with the market and the consumers is important since it over-bridges the gap between companies and customers.
Marketing
Marketing Figure 1.1
Other sub areas in marketing
Marketing communication
Advertising
Personal selling
Sales promotion
Public Relations
Web Advertising Figure 1.1. Web advertising as a sub function of Marketing Communication and Advertising. Adapted from Duncan & Moriarty (1998, p. 9).
2
The communication aspect of marketing is explicitly formulated in the AMA de?nition of marketing, “Marketing is an organizational function and a set of processes for creating, communicating and delivering value to customers and for managing customer relationships in ways that bene?t the organization and its stakeholders.” (The AMA de?nition as reported by Keefe, 2004, p. 17 – 18) 3 Jerome McCarthy was the one who introduced the 4p concept into marketing in 1960 (Harvey, Lusch and Cavarkapa, 1996). 5
At a basic level, communication will make it possible to inform and make potential customers aware of a company’s business, its offering or to persuade customers to enter into an exchange relationship. One way to communicate with customers is through advertising in digital media channels such as the World Wide Web. This type of advertising, Web advertising, is an additional way to conduct marketing communication. In ?gure 1.1 above, it is clari?ed how Web advertising is related to marketing communication and marketing. Web advertising has not always existed. It is only in recent years that novel means to conduct marketing communication, such as Web advertising, have evolved. Technology development, and in particular the development of Internet technology and the Internet, has provided an infrastructure for marketers which can be used when communicating with customers. Despite the Internet’s relative youth, organizations have been fast with experimenting and adopting Internet related marketing communication. It has proved to be an ef?cient way to reach out to the market, or to, literally, reach into the customers’ homes or of?ces. For practitioners as well as researchers the development has meant an expansion of the current advertising ?eld.
The Internet, the World Wide Web and Contexts
The Internet4 is a computer based network that can be used for a variety of activities. Its inherent versatile nature makes it a formidable tool for communication and information transfer5 while it also sparked an interest for using it in a number of ways among companies6 and consumers. This interest propelled the growth of the Internet throughout the nineties and the early years of the twenty ?rst century. The diffusion rate of the Internet has surpassed any other media achieving global prominence in barely ?ve years (Sterne, 1997). Berthon, Pitt and Watson are arguing in the same way and they are highlighting the importance of the Web and also the swiftness of the
4 5
For a brief description of the Internet see Appendix A
The Internet can be used for many different kinds of communication and information transfer using applications such as: email, internet relay chat, ftp, telnet, blog, IP-phone communication, news groups, Web and many others. The Internet provides an arena for all kinds of communication. 6 As an explanatory note consider the following; From a company perspective it is recognized that the Internet can be used for various business activities such as logistics, buying, retailing, production, advertising, and the like. Companies have explored the Internet and its possibilities trying to exploit its properties to their bene?t. In this pursue managers saw great possibilities such as low costs for attracting customers, severely reduced costs for serving another customer, small companies competing on “equal” terms as larger companies, economies of scale, low production costs, low costs for distribution, a large scale network and infrastructure ready to use for any company and so on. All these potential bene?ts together with the business climate at the time (the late nineties) called for companies to take action. Companies just had to go “virtual” in order to defend positions vis-à-vis competitors and to be a part of what was about to come or simply, as Rossiter (1999, p. 1) formulates it, “…because it seems everyone else is doing it…” To the detriment of many companies and despite their efforts they often did not achieve the anticipated success. Recent research has presented explanations to this from different angels and most of all pointed at an unrealistic expectation on what IT and the Internet can do for organizations. The gained knowledge from the early days of the commercial Internet revealed that there are not only possibilities with the Internet but also obstacles and pitfalls that has to be understood if a company is to use it as an ef?cient marketing tool. 6
diffusion of the Web, “The Web is not a transient phenomenon. It warrants serious attention by advertising academics and practitioners. Statistics support this, although one astute observer recommends strongly that all estimates be made in pencil only, as the growth is so rapid. No communication medium or electronic technology, not even fax or personal computers, has ever grown as quickly.” (1996, p. 44). As a comparison, it took Radio some thirty years and Television some ?fteen years to become a wide spread media (Sterne, 1997). A great diffusion of the media also contributes to making it more ef?cient, from an infrastructural point of view, to the bene?t of the users (Rogers, 1995). Today the World Wide Web is part of everyday life and for many people the Web has become one of the major media for consumption (Morris, Woo & Cho, 2003). The Web is not only being used by consumers, it is increasingly being adopted by marketers as well (Prasad, Ramamurthy and G.M. Naidu, 2001). That the Web has grown important for both consumers and companies is undisputed. When discussing around the Internet and the World Wide Web it is important to differentiate between the two. What millions of consumers and companies world wide are using is in fact applications or services that are available on the Internet, such as email, newsgroups or the World Wide Web. This means two things; people are indeed using the Internet but only its infrastructure, secondly and also more importantly, it is the communication bene?t of the applications and services that has attracted the attention of companies and consumers and of most importance to this dissertation is the service that usually is called the World Wide Web, or just the Web. It is in fact the success of these services that has driven the growth of the Internet and in particular email and the WWW (see Appendix A). With the widespread use of the Web in society the phenomena has come to attract the interest of researchers. Even though the ?rst advertisement7 appeared on the Web (in the on-line magazine “Hotwired”) only a decade ago there have been a number of studies conducted that have explored various aspects of the Web. Research effort has been directed at comparing marketing activities on the World Wide Web with activities in the real world (see for instance Gallagher, Foster & Parsons, 2001) suggesting both that there are differences as well as similarities. Initial concerns about the fact that traditional principles of mass media advertising would not apply on the Web was disproved. Gallagher et al found that Web advertising was at least as effective as its print counterpart. They also found that some promotional material that was not classi?ed as advertising was less effective when presented on the Web. Gallagher, Foster and Parson’s research was undertaken in order to get an understanding for Internet’s comparative properties.
7
October 1994 is often marked as the ?rst time when an advertisement appeared on the Web. It was in the on-line magazine “Hotwired” and the advertisers were Zima, AT&T, Volvo and a few others (Brandweek). 7
Research has also been directed to studying the Internet and the Web on its own merits, (see for instance Hoffman & Novak, 1996, Dréze & Zufryden, 1997, Rossiter & Bellman 1999, Gallagher, Foster & Parsons 2001, Zhou & Bao, 2002, Novak & Hoffman, 2003 and Novak, Hoffman & Venkatesh 2004,) which is an important step in determining how marketing communication, and in particular advertising, works on the World Wide Web. The study to be conducted in this research belongs to this category and is thereby not comparing the Web with other types of media. When studying the Internet and the World Wide Web from a marketing communication perspective it is obvious that the Web is one out of many media channels. It can be used to transfer information and messages to consumers. Other media channels that are frequently being used for marketing purposes are TV, Radio, news papers and the like. In this context the World Wide Web becomes just an additional way to distribute marketing communication to target markets. In ?gure 1.2 frequently used media are mapped with the Internet and the Web inserted into the media terrain.
Media Channels
Media Channels Figure 1.2
Media Channels
Tv Newspaper Billboards Radio Magazines Internet
WAP Entertainment Websites
e-mail News paper Websites
IM
News groups Search engine Websites
WWW e-mail Websites
Figure 1.2. Map showing media channels and the new digital media terrain.
The model above shows that the Internet is one out of many possible media channels. The Web, in turn, is merely one out of a number of Internet services. These Internet services are different communication channels that can be used to target consumers and to convey messages. Advertising can for instance be done through e-mail or instant messaging services and of course also on the Web. The Web is however not merely one media but can be divided into different Web site categories depending on, for instance, what activities the user can perform on a particular Web site. There are well known Websites that can be put into the various categories in the model in ?gure 1.2 such as Hotmail, Yahoo, New York Post, Le Monde, Google, Altavista, Disney or Lunar Storm (a Swedish Web site)8.
8
Notice that the New York Post and Le Monde merely have extended their business mission from print to Web whereas the other examples are entirely new enterprises in the Web domain. Disney is perhaps not entirely new since they have for a long time been in the area of entertainment. Though, the kind of entertainment they offer on the Web differs quite a bit from what they offer in their entertainment parks for instance. 8
In line with the above model, Hotmail and Yahoo belong to email Websites; NYP and Le Monde are newspaper Web sites et cetera. Of particular interest here is whether one and the same advertisement would have the same impact when presented in different Websites or Web contexts. Web context is here de?ned in a similar way as, for instance, De Pelsmacker, Geuens and Anckaert (2002), that is, the context comprises of the characteristics of the medium in which an ad is inserted. The impact of dissimilar media contexts on advertising effectiveness have been studied in for example TV and print. The context, in which a commercial or an advertisement has been placed in, has been found to be important for the effectiveness of advertising (De Pelsmacker, Geuens & Anckaert, 2002). Studies have explored the interplay between the advertisement and the media or media related factors, attempting to investigate which media factors that have effect on recognition and recall. Nowland et al and Politz (1962; 1962) have shown that different media vehicles can generate different communication effects even on the same target group using the same advertisements. The media vehicle is thereby central when deciding on which vehicle to use since different vehicles have different impact on the audience. Other studies (for instance, Harmon and Coney, 1982; Sternthal, Dholakia, and Leavitt, 1978; Sternthal, Phillips, and Dholakia, 1978, Hermes 1996) show that also content affects the impact of advertisements as well as credibility of the source. Advertising on the Web can be described in relation to how the advertisements are presented compared to other media channels. For example, TV advertisements - commercials - are presented sequentially so that program content is interrupted by a commercial break when the commercial is aired. This is also applicable for Radio where program content and commercials, with very few exceptions are presented serially. This practice clearly deviates from the normal case for Web advertising, since advertisements on the Web are usually presented in parallel with the Web site content. In print, advertisements are also presented in parallel with the program content. The reader will thereby be exposed to both the program content and the advertising simultaneously. In this respect, Web advertising and print advertising are similar. The differences between the Web and other media channels were pointed out in the previous but can be summarized as; the Web represents an interactive media with a greater richness of information than what can be found in traditional media (Evans & Wurster, 1997). Furthermore, the Web exhibits a greater interactivity, ?exibility and complexity than both print and broadcast media (Bruner & Kumar, 2000, Karson and Korgaonkar, 2001). The Web also has a visually distinct appearance, which differentiates the Web from other media channels. One of the visual features is that pages to additional information are presented as links that are clickable. As a result, the entire browsing system is functioning and organized in a different way as opposed to print or other media channels.
9
In addition, Websites are not homogeneous. On the contrary, there is a great variation in how they are designed and for what purpose they are being used. Websites can be categorized into groups that share similar functions or visual appearances such as search engines, portals, Web shops, news Web sites and the like. These Website categories represent different contexts in which advertising can be presented. Hence, visiting different Websites means visiting different contexts and the impact of context on advertisements may differ among Web contexts. On the Web sites described in the previous paragraph the Web user can perform many different activities. Various activities will call for more or less commitment from the Web user. In some cases the user can just surf around for entertainment and wander from Web page to Web page and be amused by various activities she is performing. In other cases the Web user is searching on Web sites and Web pages in pursue of certain kind of information or to perform certain activities. These two broad cases sketched are examples of different levels of goal orientation or internal drive guiding the behavior exhibited by the Web user. Janiszewski (1998) describes two broad types of behavior when searching for visual information; goal-directed search and exploratory search. Goal directed visual search is the type of behavior when the consumer is motivated and gathers information in an ef?cient way. Exploratory visual search is the type of behavior when the consumer lacks the motivation or experience to search ef?ciently (ibid). Janiszewski is taking as an example of exploratory visual search when a consumer is paging through a catalogue in an unplanned manner. This type of behavior is very similar to browsing the Web as is described above. However, the level of goal orientation and internal drive can also be seen as an additional contextual dimension that may have the potential to in?uence how advertising is attended to and processed. Since the Web offers a high level of interactivity, and other properties, this contextual dimension may be of great importance. Two aspects of the context dimension have been described in the above. The various Web contexts where advertising is embedded can be considered as mainly an external dimension whereas the Web users’ level of goal orientation may be considered as an internal dimension or more connected to the individual. Nonetheless, these two aspects of context may in?uence and necessitate different levels of attention from the consumer and thereby affect the impact of advertising. With the above information at hand, the ?rst part of the research problem is formulated as: Are there differences in effects regarding attention to advertising between different contexts and if so, how can these differences be explained?
10
The purpose of the ?rst study is to compare how the level of attention distributed to an advertisement varies between different Websites and different levels of goal-orientation. A number of experiments are to be set up to study the context dimension of attention to advertising. A more detailed description of the method and the speci?c design of the experiments are to be found in the method chapter.
Advertising and Attention Getting on the WWW
In a media world with increasing din and media clutter, advertisers are struggling to break through the noise and get the customers to attend to their messages. Advertisers are using different tactics to make their advertising to become attention getting or attention grabbing (Campbell, 1995). The quest for attention is part of the competition and re?ned techniques may result in a competitive edge. However, according to Pieters, Rosbergen and Hartog the main focus of consumer research has been on “information processing and on the effects of advertising on attitude change” (1996, p. 242). They are furthermore stating that “…little is known about processes of attention, in particular of visual attention to advertising” (ibid). This statement supports this study and that research in this area would be appreciated by the research community. In line with Pieters et al, the importance of the attention construct has been highlighted and emphasized by other researchers. For instance, Rossiter and Percy (2001, p. 167) are revealing their stance by stating “Before anything else can occur, you must ?rst pay attention to the advertising.” This indicates that attention is a prerequisite that is taking place prior to all other constructs, that advertising researchers aim to study9. The fact that not much research has been conducted in this area, as Pieters et al is stating reinforces the notion that there indeed are reasons to devote more research efforts to this interesting topic. Horace Schwerin (1967, p. 56-57) states in a somewhat similar fashion as Rossiter and Percy that, “The opening sequence of any commercial is of key importance, since advertisers must capture and hold the attention of viewers”. Janiszewski and Bickart (1994, p. 329) are equally concerned about the importance of attention when they argue that “Despite the tremendous amount of money spent on buying consumer attention, little to no research is done on consumer attention”. Additional reasons why attention is important is for instance that, “Catching the consumer’s attention can keep her or him from mentally tuning out, switching focus to an alternative activity, or zapping to another channel.” (Campbell, 1995, p. 226). Furthermore, increased processing attention has been found to lead to increased information processing and more positive
9
For example, memory decay, attitude formation, purchase intention or the like. 11
attitudes (Cialdini, Petty & Cacioppo, 1981, Petty, Cacioppo & Schumann, 1983). Increased attention and processing also make attitudes more persistent and more resistant to negative information (Haugtvedt & Strathman, 1990). Summarizing the above it is not far fetched to conclude that attention is imperative to advertising, no matter what media is being used. The strategies that are used to attract attention differ somewhat between media channels as a natural consequence of the inherent properties of a given media. In TV and Radio, attention getting tactics have been used for quite some time and there are many ways to attract attention. One small example of how to attract attention in the TV media is to delay brand identi?cation (and sometimes also product category) until the end of the commercial (Campbell, 2001). In doing so the customer will not know what product is being advertised until after processing the entire commercial. The purpose with this presentation technique is to create suspense (Fazio, Herr & Powell, 1992). Attention getting techniques in a World Wide Web setting is a relatively new business and the starting point is the Web advertisement and its properties. The modern Web advertisement of today has developed from the rather unsophisticated “banner ads” that were used in the early era of the World Wide Web. While many may have forgotten them it is perhaps in place to refresh the memory. The early Web advertisements can perhaps best be described as a one colored rectangular square blinking with a brand name (or the like) and a blinking text stating “click here”, “buy now” or some similar phrase. The old Web banners were an oddity in the advertisement family (print ads, TV and Radio commercials, outdoor advertising etc.). When comparing to advertisements in other media channels the banner ad was the most similar to billboards or banners used in for instance sport arenas or the like. Its shape and form were similar but the banner ad was placed in a Web environment instead of in a row of banners in an arena. This notion is supported by Lei (2000, p. 469) who describes banners in the following manner “At best, they should be thought of as mini outdoor billboards, with similar creative limitations.” By this Lei meant that the old banners could not convey complex messages. Instead, these banners were designed with the purpose to generate click-through to another target site where more detailed information could be gained. In comparison, the modern Web advertisements have a visual appearance that is more similar to print advertisements that can be found in traditional newspapers and magazines. They are square-like as opposed to the very rectangular banner advertisements. Furthermore, the Web advertisements have an elaborated and “complete” message compared to the headline/punch line oriented banner advertisements. Even though most Web advertisements are still clickable they are not primarily aiming at generating click-through to a target site. Instead, they are presenting the product or the like with the purpose to promote the relevant product by communicating an advertising message that is presented on the Web advertisement. It is worth to notice that an advertising message may serve various purposes. One advertisement may be
12
designed to increase awareness or enhance the company’s image in some sense, another advertisement may have the purpose to explain how a certain product is to be used and a third may be aiming at triggering purchase. Regardless of what purpose the advertiser has, it is no longer necessary to generate clickthrough. It is of course, by no means a disadvantage when the consumer clicks on the advertisement and learns further about the company and/or the product on a certain target Web site. When observing the attention getting strategy and/or the various techniques that can be used for advertisements it is obvious that the Web differs from print. Commonly, in print ads the advertisement designer can use colors, pictures, texts and size to impact the attention to the advertisement. On the Web, the advertisement designer can use the same means (colors, pictures, texts, size etc) but also the speci?c and inherent features that the Web can add to an advertisement, i.e. animation, pop-up effects, ?ash, size increasedecrease and other potentially attention getting techniques. This is apparently a signi?cant disparity that differentiates the Web advertisements from the print advertisements. The print advertisements are, by default, static since they cannot be animated, pop-up or the like. The effect that can be added to a Web advertisement in a Web context is an additional dimension of an advertisement which can increase the attention to the advertisement. How this dimension impacts advertising has not been systematically studied. With this information the second part of the research problem can be formulated: Are there differences in effects between different “attention getting techniques” and if so, how can these differences be explained? The purpose with the second study is to conduct a series of experiments in order to investigate how Web advertisements differ in their ability to generate attention in a Web environment. When studying various types of advertisements and how powerful they can be at attracting attention, it ought to be kept in consideration the fact that certain types of advertisements are not always appreciated by Web surfers. Some of the more aggressive attention getting techniques have been reported to stir up negative emotions among Web surfers. This is an unwanted collateral damage that occurs when the advertiser is eager to get the attention of its audience. How this is affecting the advertising has been researched but results stay ambiguous. Recent studies related to advertising on the Web have given contradicting results regarding people’s attitudes towards Web advertising. Ducoffe (1996) presented ?ndings showing that Web surfers did not ?nd advertising disturbing or interfering with the Web surfer’s activities.
13
From an advertising point of view, this was good news. Other researchers such as Brackett, Benjamin and Carr (2001) have, on the other hand, found examples where the Web surfers considered advertising on the Internet being irritating, annoying, and even insulting to people’s intelligence. The results are alarming, raising questions about what is being done erroneous. It is important to notice that Bracket et al’s ?ndings were presented ?ve years later than Ducoffe’s. It can be argued that Ducoffe’s study was conducted at a time when Web surfers were curious about every aspect of the Web while Brackett et al’s study was conducted when consumers had already gotten acquainted to the Web and when the novelty and initial curiosity had faded. However, attitudes towards advertisements, and the context where it is presented, are important factors since it is anything but positive when the advertisement is considered an insult by the target audience to which the company is trying to communicate with. The ambition is therefore to attempt to shed light on this issue. A second purpose with the second study is therefore to investigate if and how individuals’ attitudes towards attention getting techniques differ between different attention getting techniques, and what the consequences may be.
Modeling Attention to Web Advertising
The core of advertising research can be traced back to the very end of the nineteenth century. It evolved from personal selling in general and from the AIDA model in particular, which in fact was a model of personal selling. It was adapted by researchers in advertising and became a foundation for further research within the area (Strong, 1925, p.76). The AIDA construct (Attention - Interest - Desire - Action) was the ?rst formal advertising model (ibid) and others have followed such as the well known Elaboration Likelihood Model of persuasion (Petty, Cacioppo & Schumann, 1983). Many of the advertising models are including the attention construct but how attention to advertising works in a greater detail (for a comprehensive review of advertising models, see Vakratsas & Ambler, 1999) is seldom elaborated on. There is however a neighboring discipline, experimental psychology, where attention has been studied extensively. A number of researchers (see for instance Broadbent, 1958, Deutsch & Deutsch, 1963, Treisman, 1964, Walley & Weiden, 1973, Kahneman, 1973, Moran & Desimone, 1985) have modeled attention elaborately. Common for these researchers’ models is that they provide a richer understanding for how attention works than what advertising models do. This is perhaps not unexpected but given that attention is often treated, in advertising models, as a box amid other boxes it is perchance in place to intricate on how attention is modeled in advertising (see for instance the ELM or the AIDA model and others). The bene?t of such an approach would be to extend the knowledge of one area into another.
14
Amongst the theories on attention that have been put forward, Treisman’s model has received more positive reception than others’ (Johnston & Heinz, 1978). Treisman has been one of the proponents of an early selection theory (Broadbent’s theory also belongs to the early selection category but is more in?exible). On the other hand, Treisman’s view on attention is that there is a sensory register allowing stimulus input to pass an attenuator that reduces or attenuates processing of information outside focal attention (Treisman, 1960, Eysenck and Keane, 2000). As a consequence, limited amounts of information can reach short-term memory where information is stored temporarily. Crucial in the selective attention process is how input is transferred from the sensory register to the short-term memory. Broadbent suggested a ‘?lter’ that mechanically screened out non-attended stimuli (Keane, 2000), which was shown to be incorrect by Treisman (1960). Treisman, on the other hand, referred to an ‘attenuating’ mechanism. In her view this mechanism may signi?cantly reduce the incoming input, but does not necessarily block information in a similar conduct as in Broadbent’s ?lter (Treisman, 1964, 1960). Elaborating around the concepts and the function of the attention system, rather than treating it as one box, and to build a model infused into an advertising setting could provide greater insights and understanding concerning how attention operates in advertising. It provides a deeper understanding of how contextual factors or attention getting techniques are impacting the attention to a Web advertisement. Thus, a supplementary purpose is to test an adapted model, originating from attention models in experimental psychology, in an advertising setting in order to provide explanations of how attention to advertising works. What is to be focused on, in speci?c, is the relationship between the Web environment/ context, attention getting and the attention system that may allow advertising input to pass through or to be attenuated.
Measuring Advertising on the Web and Pricing Methods
Hitherto, the focal areas in this dissertation have been outlined and described. The research problem has been presented and the core purposes along with it. Still, when studying attention in a Web advertising context there are a number of supplementary areas that can be of interest to investigate. In addition to the main studies, an opportunity was given to add two minor areas namely, “how to measure advertising on the Web” and “pricing strategies for Web advertising”. The ?rst area (how to measure…) is a study of an old problem in a new setting (the Web setting). Nonetheless, it is important to empirically con?rm and replicate how advertising is to be measured on the Web in view of the fact that it is a different media channel as opposed to the traditional ones, which has been explained previously. The results from the ?rst area, however, may have signi?cant implications for the second area “pricing strategies for Web advertising”. If a different measurement method of advertising effect (recognition & recall instead of click-through) is more appropriate than the
15
method that is currently being used, then perhaps Web advertising pricing strategies also need to be revised. In other words; if click-through is not accounting for the entire advertising effect, then the Web advertising pricing strategy needs to be changed since the foundation of the current Web advertising pricing strategy is based on the click-through measure. In connection to these two studies it is in place to emphasize that it is Web advertisements that are being targeted and not hypertext links, keywords or other kinds of Web marketing. It is advisable to keep in mind that keywords generated from search engines and the like often are not quali?ed as advertisements when using a modern de?nition of advertising. Next, the two above mentioned additional research issues will be elaborated on further. The ?rst one to be described is; Recognition & Recall versus Clickthrough as a means to measure Web advertising effect. During the ?fties some inspirational research took place, which Darrell Lucas (1960) described as, “probably the ?nest example of experimental design ever reported on advertising research”, when referring to the PARM study by the Alfred Politz organization (Krugman, 1985). The PARM study (A study of Printed Advertising Rating Methods) was a study to investigate issues regarding measurement of advertising, in particular differences between recognition and recall of advertisements. This study showed that a signi?cant difference between the recall and recognition measure existed and that recognition was more sensitive than recall. At the same time, the study gave support to the small samples that were used by leading practitioners at the time such as Starch and Gallup & Robinson (Krugman, 1986). Ever since the heydays of the PARM study, the advertising industry as well as academia has been working on conducts to improve the techniques to measure advertising effectiveness. Nonetheless, in the case of how Web advertising is measured, the inherent properties of the Web could automatically provide a response measure. This measure, i.e. the click-through measure, is measuring the number of individuals that click on an advertisement. When this method to measure responses was identi?ed, the click-through measure soon became widely used as a measure of advertising effect without much re?ection over its appropriateness. In this context, it is valuable to remember the fact that there is a number of means to measure advertising, for instance recognition, recall, attitude and opinion measurements, comprehension measures, believability measures, persuasion measures, buying predisposition, ad ratings and behavioral measures (Leckenby and Plummer, 1983). These various measures are targeting different effects that advertising can achieve. For instance, recognition and recall are often used to measure the initial attention phase in the advertising sequence. Krugman states that the special virtue of the recognition measure is to elicit information of brief low-attention exposures to advertising, whereas
16
recall is for advertising that has been more closely attended to (1986). Since recognition is more sensitive than recall the recognition measure will comprise individuals who are able to recall the advertising as well. Those who recognize the advertising will also recall it (Krugman, 1986). Recognition as a measure of advertising effect refers to a respondent’s subjective estimate that she or he looked at or attended to an advertisement that has been previously shown. Recall re?ects an advertisements’ ability to register the sponsor’s name (Wells, 2000). These measures that were included in the PARM study were relevant rating methods for printed advertising. Despite the well established position that these methods had, the click-through measure became widely used as an advertising measure when the World Wide Web appeared. In fact, some proponents of the click-through measure were discrediting the previous methods and before long the most frequently used measure was click-through, a position which is still holding (Chandon, Chtourou and Fortin, 2003). In studies of Web advertising, the focus has often been on the effects of banner advertisements with click-through frequencies being used as a means to measures effects. The click-through frequencies have however dropped to below 1% making it an unreliable measure when using small samples (Khermouch and Lowry 2001). There are reasons to assume that an advertisement has effect albeit an absence of a “click on the advertisement” (Interactive Advertising Buereau, 1997). This makes it logical to shift from click-through studies to study advertisements on the World Wide Web using traditional methods (for instance recognition and recall) as measurement. In doing so, it will be made possible to extend the measurement procedures that have been applied for decades in traditional advertising and to transfer them into the Web advertising domain. Using recall and in particular recognition, seem to be a natural option that is also in line with what, for instance Krugman advocates (1986, 1988). Krugman, is one of the researchers that has most credibly pointed at the necessity to use recognition and recall in, for instance, TV media. Researchers have in fact been looking into the Web measurement issue but still there are disbeliefs and most of all, the industry is using the click-through metric frequently and referring to it as a means to measure advertising effect. Briggs and Hollis (1997, p. 33) are stating that “The practice of evaluating Web advertising on the basis of click-through is like evaluating TV-ads for automobiles on the basis of how many people visit the showroom the next day”. Based on this, the purpose with this ?rst minor study is to investigate whether there indeed is an effect of Web advertising as measured by recognition and recall and if so, how big this effect is in relation to the click-through measure. This may enlighten somewhat regarding the appropriateness of using the clickthrough measure as a means when measuring advertising effects. The second additional research issue to be addressed is; Pay per click versus Pay per view as a basis for pricing in Web advertising.
17
Mangàni (2003) argues that pricing methods on the Web has evolved along a dissimilar path compared to broadcast or newspapers. Mangàni is attributing this slightly different evolution to the technological possibility to track and analyze the behavior of Web surfers. On the Web there are mainly two ways to charge for advertisements, pay-per-view or pay-per-click. Pay-per-view (also called CPM10) is a model where the advertiser pays for the number of impressions whereas pay-per-click is a method to charge for the number of Web surfers that actually clicked on the Web advertisement (Mangani). The click-through measure is a straightforward and quick way to measure click frequency on a Web advertisement since the statistics are easily accessed and produced in the same moment as the customer clicks on the advertisement (Berthon et al., 1996). Consequently, publishers and advertisers started to use this measure as an indication of the advertising effect in the early days of the World Wide Web. Without much contemplation or any studies of implications, the click-through measure also became a foundation for pricing. The logic for this is based on the following train of thought; if click-through is measuring advertising effect then it is reasonable that advertisers pay for what they obtain, i.e. clicks. From the advertisers’ perspective this is an appealing idea since this would mean that they are only paying for explicit action in the form of a click. The entire idea rests upon the assumption that there is no effect beyond the click and therefore the advertiser should only pay for the number of individuals that has “clicked through”. The answer from the ?rst area will provide an answer to whether there is an effect beyond the click. If there is no effect beyond the click then it is reasonable to have a pricing strategy which is based on clicks. If there is indeed an effect (in addition to the click) then there are motives to state that an advertiser who is merely paying per click will get an additional effect that is not being paid for. Briggs and Hollis (1997) have compared the practice of assessing Web advertising based on click-through by evaluating TV-ads for automobiles based on the number of people visiting the showroom the following day. Consequently, the purpose of the second study is to elaborate on the aptness of using pay-per-click as a pricing method for Web advertising.
10
Cost per thousand – M is the latin letter representing one thousand.
18
Outline of the dissertation
Chapter two of this dissertation provides a theoretical background and sets the stage for the study to be conducted. It also features the theoretical framework of the dissertation mainly examining theories of advertising and attention. Chapter three contains a summary of the theoretical argumentations leading to the formulation of hypotheses to be tested. In addition, a restructured model that originates from the attention research ?eld is inserted into an advertising setting. Hypotheses are formed to test the model. Chapter four provides the methodology, the overall design and way of conduct of the various experiments. This chapter is also addressing issues regarding how and what type of analysis that will be conducted on the collected data sets. A presentation of gathered data will kick-off chapter ?ve and initiates the analysis of the data. Various methods for analysis are being used. Chapter six comprises of conclusions, a general discussion and guidelines for further research. Following this a section with references and appendices is presented.
Outline of the Dissertation
Chapter 1 Chapter 2 Chapter 3 Chapter 4 Chapter 5 Chapter 6 Introduction, Research problem and purpose Theoretical background and theoretical framework Summary of theory and hypotheses Research Method Results and Analysis Conclusion and discussion Figure 1.3
Figure 1.3. Outline of the dissertation.
19
Chapter Two
Theoretical Framework
In this chapter the theoretical framework will be presented. The framework will address theoretical issues associated to the research problem. While in the previous chapter a brief portrait of the relevant theories has been given, this chapter will provide a presentation of a more elaborate theoretical overview.
20
The origin of Advertising
From a historical perspective, advertising can be seen as originating from the public criers on town markets (Dyer, 1982) in ancient cities such as Ur, Babylon and Jerusalem. The public crier served the function of providing timely, and some times persuasive, information about goods and services available on the market. In a similar fashion, the modern advertising of today is communicating to customers about available offers. The means of the public crier and those of modern advertising may differ but the basics remain the same. Daniel Starch argues that advertising in some form probably has existed “since the time when men lived in communities and competed with one another for the necessities and luxuries of life” (Starch, 1923, p. 17). Without going any further into the archives of advertising, it can be concluded that advertising is not just a recent phenomenon but has been an intrinsic component of society for quite some time. Advertising as a scienti?c area for academic study shares similar characteristics as the marketing subject. Advertising is a subject that is often applying theories and ideas from other disciplines. For instance, the in?uence of advertising on consumers is understood in light of theoretical perspectives brought in from linguistics, literature, psychology, sociology, anthropology, economics and other subjects (MacInnis, 2005). Methods like ethnography, survey research, experimentation, quantitative modeling and the like are being used to assure a diversi?ed understanding of the advertising phenomena. “Through this multimethod, the multi-theory perspective, we (the “blind men”), gain insight into the “elephant” known as advertising” (MacInnis, p. 14, 2005). Albeit this research does not span over all the areas and use all the methods stated above, it is clear that this research is an intersection where multiple subjects unite. In this sense, this study is following a traditional path in marketing and advertising research. Advertising, being one of the means which can be used to market an offer by communicating to an audience, is intimately connected to primarily two areas. One area is marketing, which has the purpose to create, communicate and deliver value to customers (Darroch et al., 2004). Another area is communication,1 which is the process of establishing a commonness or oneness of thought between a sender and a receiver (Schramm, 1955) or the
1
The area of communication theory has to some extent been subject to controversies regarding origin, theoretical foundation etc. In marketing, Schramm’s model of communication is often used as a starting point and it is perhaps important to consider the following; Schramm’s model is congruent with Claude Shannon’s general theory on signal transmission developed in the electronics industry and more speci?cally at Bell Telephone Company. According to for instance David Richie, Weaver (Shannons “co-author” in 1949) in speculative manners extrapolated Shannon’s theory on signal transmission into the social sciences as a theory of communication. Though, this is by no means Shannon’s wrong doing or as Ritchie states “To criticize Shannon’s model as inapplicable to the complexities of human communication is to criticize a rowboat because it is not a whale” (Ritchie, p. 280 1986). Shannon’s signal transmission theory de?nes the information rate of signal transmission and not the information content of a message. “The semantic aspects of communication are irrelevant to the engineering aspects” (Shannon, 1949) and since the purpose of Shannon’s theory was not communication it should not be used for other purposes. Though, other researchers like for instance Gonzales (1988), Finn & Roberts (1984) see it as revitalizing to bring the theory to other situations that are not strictly analogous. 21
process by which individuals share meaning (Dibb & Simkin, 1991). Duncan & Moriarty (1998) argue that marketing theory and communication theory share common roots and thereby they are enriching each other. Amalgamating marketing and communication consequently results into the area of marketing communication. In a marketing communication context, information is distributed to inform, persuade, motivate and to make potential customers aware of an organization’s offering (Keller, 2001). Marketing communication comprises of a variety of activities that aims at communicating with the company’s customers. Marketing communication tools are typically divided into four or more de?ned areas and these tools or sub functions are often referred to as the promotional mix. The tools in the promotional mix are advertising, personal selling, sales promotion, public relations and can also be extended with for instance direct marketing, packaging, point of purchase display and event marketing (Keller, 2001). Of these mentioned tools, advertising will be scrutinized in greater detail since it is the main topic at hand and of special interest in this theoretical framework.
What is Advertising and What is a Web Advertisement?
In the literature, a number of descriptions and de?nitions of advertising can be found. An early de?nition by Daniel Starch (1923, p. 5) suggests, “The simplest de?nition of advertising, and one that will probably meet the test of critical examination, is that advertising is selling in print. Or to put it more completely, commercial advertising consists in presenting a commodity in print to the people in such a way that they may be induced to buy it”. Starch’s de?nition is easy to comprehend but is perhaps not the most elaborate one. A further problem is that the de?nition is too narrow in that it merely recognizes advertising in print as advertising. This means that it is excluding all other types of contemporary media channels for advertising. Yet, it ought to be kept in mind that Starch’s de?nition of advertising was presented at a time when print was indeed the major means to convey advertising messages. In a recent discussion on the topic of advertising de?nitions, the editor of journal of advertising, Faber (2002, p. 1), argued that de?nitions of advertising usually embrace components such as “paid’’ messages, “from an identi?ed sponsor” in “mass media” with the goal of trying to “persuade or in?uence”. The same year as Faber published his notions, Richards and Curran (2002) screened a number of advertising de?nitions in order to capture the essence and put together a “current de?nition” which is as follows; “Advertising is a paid non-personal communication from an identi?ed sponsor, using mass media to persuade or in?uence an audience”. Richards and Curran’s de?nition of advertising aims at being general and thereby harbors all kinds of advertising, including Web advertising. Though, Web advertisements are somewhat different than for instance print advertisements. It is therefore appropriate to elaborate on what a Web advertisement is and what kind of Web advertisement is being targeted in this study.
22
Lei (2000) described the early banners as mini outdoor billboards that could not convey complex messages. This limitation that Lei pointed at in year 2000 has been eradicated and the modern Web advertising does not suffer from it any longer. On the contrary, Web advertisements can now provide quite complex messages as they are not constrained to being button advertisements or mini-billboards as they once were. Visually, the Web advertisements are quite similar to print advertisements but can in fact exhibit greater complexity, especially when animated. It is therefore misleading to speak about “Web banners”, when referring to advertisements on the Web. The word banner, in its basic form, denotes a rather primitive type of advertisement where usually a brand name or company name is printed on a large rectangular piece of fabric with diminutive room for anything else to be communicated. Figure 2.1 below shows what is often accredited to be the very ?rst banner ad used on the World Wide Web. The advertisement in this ?gure is an archetypal example of the very simple banner-like advertisements that were once frequently used on the Web. This type of advertisement is becoming less common today and is not to be studied here.
One of the First Web Advertisements
Figure 2.1
Figure 2.1 Depicted is the ?rst banner ad used on the World Wide Web by AT&T in the on-line magazine Hot Wired in 1994. The advertisement is reprinted with AT&T’s kind permission (2006).
Concurrent with the development of the Web banners, an entire industry of search engine generated keywords or “AdWords”2 have surfaced on the World Wide Web. These are clickable words in a text or small “AdWords” next to search results on a search engine. Albeit a majority of these “text advertisements” can qualify and be de?ned as advertisements, they are not of focal interest in this study. Instead, the following depiction may serve as a ?nal guideline of what is meant by a “Web advertisement” here: The Web advertisements of today, contrary to the early Web banners, often have some basic properties such as an image or description of a product, a brand name, copy text and a headline put together into an advertisement entity – the Web advertisement. Visually, the Web advertisements are similar to print advertisements and they are both intended to convey advertising messages. However, dissimilar to print advertisements, the Web advertisements are frequently published on Web pages on the World Wide Web. Web advertisements
2
See for instance Google. 23
Contemporary Web Advertisements
Figure 2.2
Figure 2.2. Depicted are two contemporary Web advertisements next to each other. On the left in the ?gure, an image advertisement by British Petroleum is presented. Both advertisements are in fact animated using multiple messages that cannot be seen here. On the right in the ?gure, an advertisement by Avis is presented. The advertisements are reprinted by kind permission of British Petroleum Corporation, Great Britain and Avis, a wholly owned subsidiary of Cendant Corporation USA (2006).
can be static (like print ads) but can also be animated or use a pop-up function. In ?gure 2.2, an example of a contemporary Web advertisement is reprinted. As is apparent, the two advertisements in ?gure 2.2 do not have a great deal in common with the old banners of the nineties depicted in ?gure 2.1. These types of advertisements are also especially dissimilar to “AdWords” and clickable text from search engines such as Google or the like. Instead, these advertisements are not very different from print advertisements that are being used in newspapers and magazines. It is this type of advertisement that is of focal interest in this study. One question that can be raised at this point is where to draw the line between simple animation and full streaming video. The answer is not uncomplicated especially considering the fact that more and more animation is being used in Web advertisements and there are a number of techniques for producing motion or animation on the Web. Only a few years previously, most animated Web advertisements were alternating between two or three visual presentations of the advertisement. At present, the advertisement designer is free to choose, from static to full motion picture, making the distinction between simple animation and full motion picture (advanced animation) irrelevant or arbitrary. To confuse things even further, there are Web advertisements that occasionally use both simple animation and streaming video. Considering that animation means moving picture, it is perhaps more relevant to discuss in terms of more or less animation employed and leave it at that3.
3
The exact animation frequency used in the experiments in this dissertation is described in detail in the method chapter. 24
The Web is growing important
Traditional media like TV, Radio, billboards, newspapers, magazines et cetera are well-established channels for advertising. The Web, conversely, has appeared more recently yet is increasing in importance for advertisers as well as publishers. Figures are pointing at a rapid growth of the Web (IWS) and advertisers are also expecting the Web’s importance as a media channel to increase4 Dréze & Zufryden (1997) stated a decade ago that one of the reasons for the increased importance of the Internet is the “tremendous growth” that it has undergone in recent years. The growth after 1997, when Dréze and Zufryden made their remark, has indeed been rapid with an astonishing 1086 million users in the World in October 2006 (Internet World Stats). Hoffman and Novak (1996), who often are considered as forerunners in the area, argued that research in this area is still in its infancy and that there is a lot of research to do to map the terrain. Ducoffe (1996) and Ducoffe et al (1996) declared early that, “continuing developments in the area of new media technology represent the most important in?uence on the future of the advertising industry over the next 10 to 15 years”. Zhou and Bao (2002) argue that it is not merely the Internet’s tremendous growth that makes it interesting among researchers, but furthermore its “distinctive characteristics”. These distinctive characteristics of the Internet, sets it apart from other types of media channels since it is a type of media with hybrid properties (Dréze & Zufryden, 1997). The distinctive characteristics that are being referred to are for instance the ability to use animation, pop-up features, radio, streaming video, TV, interactivity and the like in one and the same media and all at the same time, if one so wishes. Coupey (1999) expresses the possibilities with the Internet and the Web, stating, “The unique features of the media can provide a focus to extend or to create theory, methodologies, and the discovery of new phenomena.” Altogether, there seem to be reasons to believe, that the World Wide Web is an important type of media to study and that research in advertising connected to the World Wide Web is in a state of renaissance.
Advertising is taking place in a setting
Marketing activities such as advertising are taking place in a social setting. These are in actual fact taking place ubiquitously in society, between companies and consumers. For instance, when an individual is starting the day by reading the morning paper she will be exposed to a number of advertisements. These advertisements are placed in a certain setting. Consumers are constantly exposed to brand names, company names and window displays affecting them in various ways.
4
At an industry seminar with representatives of the top advertising agencies in Sweden all participants unanimously expected the Web and the Internet to increase in importance at the expense of other media channels. No one expected TV or any of the other traditional media to have a similar development. Industry seminar (Stockholm, autumn 2005) 25
The number of potential encounters with advertising and other marketing activities are too numerous to describe. To the long line of traditional encounters, the digital ones are now also added. A typical “digital encounter” is when an individual is sur?ng the Web and being exposed to a Web advertisement that may be part of an even greater advertising campaign. This encounter is taking place in a setting that may or may not in?uence the effect of the advertisement. Kakkar and Lutz (1981) have elaborated on how the situation and environmental aspects may affect consumer behavior. According to their view, the situation can have a great deal of impact on how people receive process and make sense of communication. The signi?cance of situational variables has also been considered important by Ward and Robertson (1973, p. 26) who argue that it is likely that situational variables account for more variance than what actor related variables do. For instance, many buyer behaviors are only exhibited under speci?c conditions, in speci?c situations (Lavidge 1966). According to Belk (1975, p. 157), a situation is a point in time and space and furthermore all those factors observable “that do not follow from a knowledge of personal (intra-individual) and stimulus (choice alternative) attributes and which have a demonstrable and systematic effect on current behavior” (Belk, 1974a). To identify the boundaries of what constitutes a situation, Belk (1975, p. 159) puts forward ?ve groups of situational characteristics representing general features. Physical Surroundings – the most apparent features of a situation. The physical surrounding includes such aspects as geographical and institutional location, décor, sounds, aromas, lighting, visible con?gurations of merchandise or “other material surrounding the stimulus object” (Belk (1975, p. 159)). Social Surroundings – which provide additional depth to the situation such as other persons present, their properties and roles, interpersonal interactions going on and the like. Temporal Perspective – a dimension of situations speci?ed in time units and related to some other entity. Task De?nition – “features of a situation include an intent or requirement to select, shop for, or obtain information about a general or speci?c purchase. In addition, task may re?ect different buyer and user roles anticipated by the individual” (Belk (1975, p. 159)). Antecedent States – is another group of factors that can characterize a situation. These are moods or short termed conditions immediately antecedent to a situation. They are states that the individual brings into the situation and not the result from being in a certain situation.
26
Albeit Belk’s characterization of a situation is not targeting the Web case per se, it is apparent that there are similarities. The previously mentioned individual in the “digital encounter” who is reading the news on a Web page will be exposed to the “physical surrounding” of the Web site with its distinct features. It is naturally also relevant to consider the social surrounding as there may be other individuals present. When visiting a Web site, it is applicable to see the event in terms of a de?ned task. The Web surfer who is visiting a Web page has the intention to do something – to solve a task. This task solving can, in more common words, take the form of reading an article, ?nd the weather information for the following day or to purchase something. To de?ne and regard the Web interaction as a task is an appealing way to describe the user – Web engagement. This notion will be further elaborated in the attention section of this chapter and used as an instrument to describe the advertising setting. In a number of studies (see f.i. Lavidge, 1966; Ward & Robertson, 1973; Belk 1974a, 1975; and Kakkar & Lutz, 1981) the signi?cance of situational factors has been deemed imperative. Extending these notions to the advertising area would mean that advertising surrounding us and reaching our minds through our senses, is taking place in a setting where situational or contextual factors evidently are at play. This means that the design of a Web site, the colors, the framing of various areas and, in particular, the complexity of the Web site may have an impact on how people attend to, receive, process and make sense of communication. One setting may be more effective, from an advertising point of view, than another setting. In ?gure 2.3 below it is illustrated how the “total” environment, in which a Web advertisement may be inserted, is built up by external factors outside the screen; the Web environment, the focal area and other surrounding areas on the screen. If the screen environment is described in detail, the following components can be found; the screen frame and the screen. The visual part of the screen comprises of the task bar, which is part of the operating system that is being used. The Web browser is also visible including the navigation toolbar and other toolbars that can be customized. When a Web page is loaded, the Web page and the information presented on the Web page will be covering most of the screen. The Web page itself can be seen as an environment which can be altered in an in?nite number of ways. It is this environment that is being referred to, when stating “Web environment”, in this study. Using the terminology that has been discussed hitherto it is appropriate to state that the Web environment can be seen as a task environment in which a task (for instance to read the Web news) is being solved.
27
The Web Environment
Advertisement Focal area
Figure 2.3
Surrounding areas
The web screen
Other factors (external, outside of the screen).
The web user is reading this and is affected by other areas of the screen as well.
Web environment
Figure 2.3 The environment in which the advertising is presented.
In ?gure 2.3 the various parts that is building up the setting is put together. It is in this context that Web advertising is taking place. This setting is divergent to other settings as it includes new elements that may or may not in?uence the effects of advertising. It is a computer mediated communication situation with a different kind of format as opposed to for instance the print media. It is a different kind of interaction since the user is not turning pages in a paper but instead clicking on links leading to additional information and additional possibilities to present the advertisements – for instance by using animation, pop-ups and the like. This model will be elaborated on at a later stage in this chapter in order to describe its various components in theoretical terms.
Outcomes of Advertising
Hitherto, it has been discussed how advertising is de?ned, that the World Wide Web is a different kind of environment and that advertising is taking place in a setting. Equally important is to add the outcomes of advertising. Companies, governments, non-pro?t organizations, political parties, and other entities employ advertising because they want speci?c outcomes from it. An advertising campaign is generally undertaken in order to pursue certain goals. The goals that are pursued are often formulated by the organization or an external partner, which is designing the advertising. The reason why an organization does advertising varies but it is often done to achieve some basic effect. It can for instance be to inform (Keller, 2001) the public or a target market about services available, or to change the attitudes (Petty & Cacioppo, 1984) towards some object or activity, for example, to not drink alcohol and drive. Other examples of desirable effects are to create interest (Smith & Swinyard, 1982) towards the product that a company is promoting or to motivate or persuade (Richards & Curran, 2002) the consumers to choose offer A before offer B. For companies that are striving to make pro?t it is important to achieve action or an actual purchase (Lavidge & Steiner 1961) which frequently is the ultimate goal in a chain of events. However, before
28
coming to the very end of the purchase process it is imperative, as a ?rst step in this chain, to capture the attention (Janiszewski and Bickart, 1994) of the audience targeted. From a communication perspective, the organization ought to start by assuring that it has the consumers’ attention. Only then it can be effective in its endeavor to transfer information or to present an offer in a favorable and appealing way. If the advertiser does not capture the attention of the consumer, then the opportunity to communicate with her is lost. With an ever-increasing number of advertisements, competing for existence the consumers’ attention has become a scarce resource (Adler & Firestone, 1997; Davenport & Beck, 2001; Pieters & Wedel, 2004). From an individual advertiser’s perspective, the task is to break through the clutter of advertising messages and other kinds of information that is calling for the consumers’ attention. Krugman, one of the researchers who have focused quite some effort on attention in marketing communication and advertising, has argued that consumers are screening “[...] the advertising and distributing their attention selectively” (Krugman, p. 48, 1988), with closer attention to some advertisements and with less to others. Krugman has also emphasized the role of capturing the audience’s attention and pointed it out as particularly important. Janiszewski and Bickart (1994, p. 329) are also pointing out the initial phase of effects, attention, as important. Horace Schwerin (1967, p. 56-57) states, in line with Krugman, that “The opening sequence of any commercial is of key importance, since advertisers must capture and hold the attention of viewers to retard them from mentally or physically tuning out”. Krugman has in his research also studied how to measure attention effects adequately using measures such as recall and recognition. Unfortunate for organizations using advertising to communicate with their target audiences, there are not only desired effects that occur but every now and then also non-effects or even worse, undesired effects. To this category of unwanted effects belong, for instance, mistakes on the part of the advertiser (messages out of tune) so that the audience does not recognize the organization, its offer or brand. It can also be mistakes on the part of the media-planning agency or others involved in the process of bringing the message to the audience. Among all potential errors, and error sources, is an additional error that has been highlighted as the World Wide Web has grown and developed to become a part of everyday life. It is connected with the advertisers’ ambition or overambition to get the attention of the audience. On the Web, it can be termed the “pop-up effect”, which is when Web advertisements are calling for too much attention and thereby become intrusive or annoying to the user sur?ng on the Web site where the advertising is presented. As mentioned in the previous chapter, Brackett, Benjamin and Carr (2001) found examples where the Web surfers considered advertising on the Internet being irritating, annoying, and even insulting to people’s intelligence.
29
This kind of negative effect is not what organizations generally want to be connected with, since it may instill negative5 attitudes in the audience. Attitudes are central when discussing around both positive and negative outcomes of advertising. An attitude is a latent variable that cannot be directly observed. The attitude refers to a tendency of psychological nature that is expressed by evaluating a certain entity with some degree of favorability or unfavorability (Eagly & Chaiken, 1992). The entity referred to, often called an attitude object, can be virtually anything that can be discriminated by an individual. Hence, an individual can have favorable or unfavorable responses towards an attitude object. The responses are generally characterized as being cognitive, affective or conative. These variables, cognition, affection and conation are measurable. Therefore, the researcher is observing these variables in order to state what attitude an individual has. Cognition refers to thinking, affection to emotions and feelings while conation refers to actual behavior towards an attitude object. Using the attitude terminology, it is adequate to state that a company can use advertising to initiate, alter or reinforce a consumer’s thoughts -cognitionaround a particular product or product attribute. Furthermore, it is at times desirable for advertisers to create certain emotions or feelings -affectiontowards a product or its attributes. The last part is to make the consumers take action -conation- in relation to the attitude object, for instance to purchase the product and to use certain product attributes. Even though it may seem obvious that attention getting can produce negative attitudes, other effects can work in the opposite direction. Part of the attention getting will produce brief exposures that have the potential to create an effect in the opposite direction. For instance, a Web advertisement is perceived as annoying, due to aggressive attention getting, while the brief exposure takes place. On the other hand, the consumer may forget about how annoying the advertisement was when recognizing the product and brand in the store. Since the consumer has been exposed to the brand a number of times, the brand will have a greater “familiarity” than other unfamiliar brands. Mandler, Nakamura and van Zandt (1987) suggest that mere exposures can have the potential to produce a kind of vague feeling of familiarity that later could be interpreted as liking. In a recent study, Monahan, Murphy and Zajonc (2000) arrived at the conclusion that “enhancement of one’s mood state can occur by virtue of repetition of exposure” (Monahan et al., p 465). This would indicate that repetition alone has the potential to in?uence liking. In a speci?c Web advertising case, the effects mentioned above are working in two directions. On one hand, too aggressive attention getting may evoke negative attitudes and on the other hand repeating the message may in?uence liking.
5
There are of course also cases where the advertiser wants to evoke negative attitudes towards an attitude object. This can, for instance, occur when a political party is portraying its opponents in a negative way. 30
In addition, when the consumer is in fact about to do the purchase the aggressive attention getting may be forgotten whereas the familiarity and liking may persist, in line with Monahan, Murphy and Zajonc’s ?ndings. There is nevertheless, the risk that aggressive attention getting in the advertising will affect the Web site where the advertising is presented in such a way that users may avoid the Web site or dislike it. These mentioned effects are important when interpreting the results of the studies to be conducted and especially when transferring the ?ndings into a real world situation. In this research, subjects will be exposed only at one occasion whereas in real world situations repetition is obviously a dimension to consider. At the same time, it is also worth to bring to mind that there are signi?cant methodological differences between Monahan et al’s study and the one presented in this study.
Putting the parts together
To summarize what has been discussed this far and to tie various parts together the following can be stated; Advertising is part of marketing communication and has, in recent years, come to be used on the World Wide Web. Initially, a certain kind of very simple “banner ads” were used but they have been developed into a “contemporary Web advertisement” that is described as: The Web advertisements of today, contrary to the early Web banners, often have some basic properties such as an image or description of a product, a brand name, copy text and a headline put together into an advertisement entity – the Web advertisement. Visually, the Web advertisements are similar to print advertisements and they are both intended to convey advertising messages. However, dissimilar to print advertisements, the Web advertisements are frequently published on Web pages on the World Wide Web. Moreover, all types of advertising are taking place in a setting. Web advertising is taking place in a World Wide Web setting, which has, as has been argued around, somewhat different features than for instance print advertising. The Web advertisement and the setting in which it is presented are part of what can be seen as advertising input to the consumer. The last piece to this puzzle is the outcome of this advertising input and in particular certain aspects of this outcome, namely attention and attitudes related to attention getting. Attention, being a construct primarily originating from and connected to psychology, is frequently used to describe the initial phase of advertising effect. In advertising models, this is often visualized when a consumer is exposed to an advertising message. Learning from experimental psychology, this way of describing attention in advertising can be elaborated and extended. It is thereby possible to increase the understanding of attention processes and how attention operates in an advertising setting. Especially when the advertising is presented in a parallel way as it is for instance on the World Wide Web, which is the case here. The parallel way of presenting information can be framed in the same way as is frequently done in research in experimental psychology. A setting where an individual, the consumer, is exposed to an advertising
31
message can be reinterpreted using an alternative terminology. The setting can be described as a task environment where the individual is engaged in some kind of activity, for instance reading or the like, when an advertisement is presented at the same time. Even though most consumers presumably do not see this situation as a task, it can be framed as a task, where there are two or more objects calling for attention. In focus is, from the Web user’s perspective, the text and pictures that the user wants to read and look at. Outside of this focus is the advertisement that is calling for attention. The advertisement is a peripheral and secondary object struggling to catch the user’s attention. The next section will focus on how attention is modeled in advertising models. The concepts that are mentioned hitherto will be further elaborated on in detail in order to get a clear understanding of how attention to advertising in Web environments is in?uenced by various factors.
Central concepts in advertising
Mainstream advertising research originates from the very end of the nineteenth century. It evolved from personal selling in general and from the AIDA model in particular, which in fact was a model of personal selling. It was adapted by researchers in advertising and became a foundation for further research within the area (Strong, 1925, p.76). According to Coolsen (1947) E. St. Elmo Lewis was one of the pioneers in advertising, publishing his book Financial Advertising already in 1908. Lewis argues in his work that good copy ought to attract attention, awaken interest and create conviction. Preston (1982) writes that Lewis actually formulated what was to become the AIDA model already back in 1898. The AIDA construct (Attention - Interest - Desire - Action) was the ?rst formal advertising model (Preston, 1982). From this model emerged the class of persuasive hierarchy models. This model and other hierarchy of effects models have had quite an in?uence in the advertising ?eld. In a study of how advertising works (Vakratsas & Ambler, 1999) the authors are using a model framing some central concepts in advertising, as can be seen below.
32
How Advertising Works
Advertising Input: Message content, media scheduling and repetition
Figure 2.4
Filters: Motivation, ability (involvement)
Cognition
Consumer: Affect Experience
Consumers behavior: Choice, consumption, loyalty, habit and so forth Figure 2.4 How advertising works (Vakratsas & Ambler, 1999).
In the model in ?gure 2.4, the starting point is the advertising input comprising of the message content i.e. the advertisement. Part of the input is the media scheduling and repetition, which together determines the reach and frequency with which the message content will ?nd its way to the consumer. Intermediate factors that affect the consumer response are ?lters such as motivation and ability (involvement) to process information. Part of the model is also cognition, affect and experience where experience represents memories of prior experiences of the advertised product or other products similar in function. Common for many studies in the advertising area is that they often focus on one or a few of the concepts presented in the model above. Also common is that authors tend to use some kind of hierarchical model even though for instance Vakratsas and Ambler claim to ?nd little support for a sequential effect in their study. Two of the early proponents of the hierarchy-of-effects model, Lavidge and Steiner (1961), were elaborating on the functions of advertising. They meant that the effects of advertising actually are long-term but if something will have a long-term effect “…something must be happening in the short run, something that will ultimately lead to eventual sales results” (Lavidge and Steiner, 1961, p. 59). Lavidge and Steiner also argued that this effect should or could be measured in order to evaluate advertising effectiveness. According to their view, advertising could be thought of as a “force”, which is guiding people through a number of stages. The ?rst stage would be potential customers who are unaware of the product or the product offer. The second stage is consumers aware of the product’s existence but is still far from buying. The next stage consists of consumers who know about the product and the offer. The fourth stage in the sequence comprise of those who like the product in question, i.e. having a favorable attitude toward the product. At the ?fth stage, people prefer the product to other alternatives. At the following stage, consumers are not only preferring the product but also convinced that this is the best option and ?nally the last stage is where all previous stages are translated
33
into a purchase. This model described is rather known to marketers as a way to determine what communication objectives that might be appropriate in different stages. Lavidge and Steiner (1961) elaborated further and meant that these stages were an indication of three main functions of advertising namely the cognitive function – the intellectual, “thinking” part, the affective function – the emotional and feeling factor and ?nally the conative function – the action oriented component. Their view was that these three components were crucial to advertising and that the aim of advertising would be to move consumers through these stages, ultimately leading to a purchase. Since Lavidge and Steiner’s article there has been a tremendous development within the advertising area and various models have been proposed, hierarchical and non-hierarchical, focusing on different aspects of advertising and taking into account some concepts while leaving out others. In the next section, a few models that are relevant to this study will be discussed and highlighted.
Models describing advertising
Not surprisingly, there is not just one model used by researchers that describes advertising. On the contrary, there are many and they tend to focus on different aspects of the advertising phenomena depending upon what the author of a particular model has chosen to focus on. The advertising models have similarities and dissimilarities and a few of these models will be looked upon in detail. Since the focal interest is the concept of attention, extra attention will be given to what the models are contributing in that particular respect.
The Classic Hierarchic Model of Advertising Effects
Ad Characteristics Stimulation Credibility Liking Personal relevance Familiarity Figure 2.5
Attention
Memory
Attitude towards the ad
Attitude towards the brand
Purchase Intention
Figure 2.5 The Classic Hierarchic Model of Advertising Effects (Attention, Memory, Attitude and Conation: A Test of the Advertising Hierarchy, Thorson, Chi and Leavitt, 1992)
The model presented above is a typical hierarchic model of advertising effects. The sequential process starts on the left in the model where the characteristics of the advertisement are presented. Stimulation refers to the advertisement’s ability to stimulate the modalities of the consumer such as vision, hearing or more uncommonly, smell, taste as well as touch. The purpose with an advertisement is to stimulate and that is also why advertisements are sometimes referred to as stimuli. Credibility relates to how credible the message is perceived. This is of course subjective since different individuals will have different opinions about whether the advertisement is credible or not. Furthermore, personal relevance, liking and familiarity are important variables that have an impact on advertising effect.
34
The next step in the model is attention, which is part of the consumer and is described as the amount of attention distributed to the advertisement by the consumer. After that comes memory, which is a measure of how well the consumer remembers the advertisement. From the advertiser’s perspective it is naturally preferable if the target group remembers the advertisement. Next, is attitude towards the ad (A.ad.) and attitude towards the brand (A.br.) leading to the ?nal stage namely, purchase intention (P.I.). Purchase intention is the conative component where the consumer acts upon the advertisement and initiates a purchase of the advertised product. The model presented above gives a quite good overview of what has been considered important concepts in advertising. When discussing advertising effects, it is often a matter of measuring variation in these constructs and their relation to each other. As we can see, the model is sequential in nature and it is thought that there is no or limited interaction or feedback between the different stages. In the literature of advertising theory, this model or derivatives of it, has been widely used (Thorson et al. 1992). Despite its wide use it has been criticized and many studies have found little or no support for its design (Thorson et al. 1992). In a study by Thorson et al (1992) it was found that advertisement characteristics did indeed affect purchase intention but the effects were both direct and partially mediated. This mediation did, however, not ?ow through attention and memory to attitudinal and conative responses. Instead, some characteristics such as credibility and liking operate through (A.ad) attitude towards the advertisement and (Abr) attitude towards the brand on purchase intention. Furthermore, it was found that attention affects neither purchase intention nor attitude towards the advertisement. Consequently, conative responses are not driven by attention. In addition, they found that the measure recall was not a good index of persuasion. Their result provided evidence that attending to an advertisement makes viewers remember it better but it had no in?uence on whether viewers would like the advertisement or brand. These conclusions seem to follow a general kind of logic since mere attention to an advertisement can hardly have an impact on liking6. Gibson (1983) has, in line with this logic, criticized the use of recall measures to evaluate advertisements since recall can only have little or no correlation with consumer choices in the market. Interesting still, is Thorson’s ?nding that none of the ?ve advertisement characteristics drive attention or memory despite the fact that they drive attitude towards the advertisement, attitude towards the brand and purchase intention. The impact of advertisement characteristics on attitude towards the ad and attitude towards the brand is an important research area that often has been debated among researchers (Brown & Stayman, 1992). This, since the purpose of advertising often is meant to in?uence attitudes. Attitude has in advertising contexts generally been described as “a learned predisposition to respond in
6
Just the fact that a viewer watches and attend to a commercial should have little effect on liking since there are numerous 35
the consistently favorable or unfavorable manner to advertising in general” (Lutz, 1985). In the classical model above attitude towards the advertisement and attitude towards the brand is, not surprisingly, part of the sequence and has also been subjected to extensive research (Brown & Stayman, 1992). Attitude towards the ad and attitude towards the brand are two measures that are being studied on the dimension favorable/unfavorable or as a composite of other belief items such as informational value, entertainment and the like (Mehta, 2000, Bauer and Greyser, 1968). Millar (1987) has studied how thought prior to an attitude assessment in?uences the correlation between attitude and behavior. In brief, it can be said that thoughts (prior to an attitude assessment) can both increase and decrease the correlation between attitude and behavior. In a study by Wilson et al (1984, in Millar, 1987) one experiment group had a correlation of 0,17 and the other test group a correlation of 0.54 for attitude correlation depending on whether they had been involved in cognitive effort or not prior to the attitude assessment. In Thorson’s study (Thorson et al, 1992) they concluded that stimulation and credibility drive attitude towards the advertisement, credibility, liking and familiarity drive attitude toward the brand and credibility and personal relevance drive purchase intention. Next, two other advertising models, which are slightly different compared to the previous model will be examined. These two models are derived from the classical model but instead of having one single route from advertisement characteristics to purchase intention they have two routes. They are also focusing on advertising from an emotional/non-emotional perspective.
The Two-Route Hierarchic Model for Emotional Commercials
Attention Ad Characteristics Stimulation Credibility Liking Personal relevance Familiarity Memory Figure 2.6a
Attitude towards the ad
Attitude towards the brand
Purchase Intention
Figure A, The Two-Route Hierarchic Model for Emotional Commercials
36
The Two-Route Hierarchic Model for Non-emotional Commercials Figure 2.6b
Attention Ad Characteristics Stimulation Credibility Liking Personal relevance Familiarity Memory
Attitude towards the ad
Attitude towards the brand
Purchase Intention
Figure B, The Two-Route Hierarchic Model for Non-emotional Commercials Figure 2.6 The Two-Route Hierarchic Model of Advertising Effects: Emotional vs. Non-emotional commercials. (Attention, Memory, Attitude and Conation: A Test of the Advertising Hierarchy, Thorson, Chi and Leavitt, 1992)
In the above ?gure, the two-route hierarchic model of advertising effects is displayed. As can be observed, these two models differ from the classical model in the sense that they have two separate routes. In the ?rst model, ?gure A, it can be observed that there is one route that is in?uenced by the advertisement characteristics and a second route that is not in?uenced by the advertisement characteristics. In the second model, ?gure B, both routes are in?uenced by the advertisement characteristics. These two models have in common the “main route” comprising of the attitude towards the advertisement, attitude towards the brand and purchase intention. The attention – memory sequence in the models is treated differently for emotional commercials compared to non-emotional commercials. The reason for this is that an advertisement that elicits emotional response in the viewer is supposed to create a memory engram that is enhanced as opposed to nonemotional conditions (Squires, 1986). It is more likely that there is a direct link between memory and attitudes for emotional advertisements than for non-emotional advertisements (Thorson, Chi and Leavitt, 1992). In the nonemotional version of the advertising model there is no link between memory and other constructs, except for its input from attention. It is rather interesting when examining these two models (and other similar models) from an attention point of view in the sense that the attention and memory constructs seem to fall outside of the model. In ?gure A, the ?rst model, the attention construct does not correlate with the advertisement characteristics and in the second model, ?gure B, the memory construct does not correlate with the attitude towards the advertisement construct. Hence, no uni?ed model can be reached. Instead, different models apply to different kinds of stimuli i.e. commercials. Here, the question can be raised whether it is possible to separate attention and memory from attitude formation in such a way as in these models above. Without any kind of attention to an object and without short-term, long-term, verbal or pictorial memory it becomes hard for the subject to have any knowledge or attitude towards a speci?c advertisement and its characteristics.
37
In the models above attention and memory have, to some extent, been described as processes not intrinsically related to the “Aad, Abr and P.I.”-sequence. At the same time, data have not supported the classic hierarchy of advertising effect model where attention and memory is part of the main sequence. Considering that attention is a construct that is inherently connected with whether input will reach the central nervous system or not, it is indeed disturbing that the attention and memory constructs have a subordinate route or sequence such as can be seen from the models A and B in ?gure 2.6. At the same time, data in some studies do not support the classic hierarchic advertising model. The reason why attention and memory are modeled aside of the main route is because data have not shown to have any correlation between attention and memory and the other constructs. Before going any further into this potential contradiction, an additional model will be introduced. A fourth model which is relevant in this context and which has been widely used, namely, The Elaboration Likelihood Model of Persuasion will be discussed.
The Elaboration Likelihood Model of Persuasion
High involment processing Cognitive Response Belief and attitude change Figure 2.7
Behavior change
Central route Communication (source, message channel) Attention and comprehension Peripheral route Low involvement processing Belief Change Behavior Change Attitude Change
Figure 2.7 The Elaboration Likelihood Model of Persuasion.
The Elaboration Likelihood Model of persuasion, ELM, has dual routes depending upon whether the subject will be engaged in a low or high involvement processing. The low involvement processing comprises of a peripheral route starting with belief change, behavior change and ?nally attitude change. The high involvement processing takes the central route comprising of cognitive responses, belief and attitude change followed by behavior change. The ELM is a model that is describing attitude formation under different levels of involvement and processing. Authors such as Karson and Korgaonkar (2001) are describing the ELM as one of the leading models for how involvement in?uences communication effects. Despite the wide use of the ELM even the inventors of the model, Petty and Cacioppo, later criticized the model since they had found that research on persuasion and involvement indicated that neither the central nor the peripheral approach alone could account for the multiplicity of attitude change
38
that has been observed (Bitner and Obermiller, 1985). Bitner and Obermiller also criticized the Elaboration Likelihood Model mainly by pointing at the following ?ve perceived problems with the model: 1) Central cues – peripheral cues which are which? 2) How does peripheral processing in?uence affect? 3) Are there differences in the strength of peripherally and centrally processed attitudes? 4) Are central and peripheral processing interactive? 5) Can the central processor make do with peripheral cues? The model has also been targeted by Vakratsas and Ambler (1999) criticizing the ELM model together with other persuasive hierarchy models (the AIDA group) chie?y because: this category of models assumes that the “[…] brain works through a series of stages, as if it were a primitive serial computer. The brain actually works in parallel” (Ambler, 2000, p. 305). And also because: “these models ignore the consumer’s experience of previous product usage” (Ambler, 2000, p. 305). Ambler is a critic of the strong theory (as the main theory) stating that the notion that advertising equals persuasion is an ingrained belief and that most advertising “does not persuade consumers to change their minds but strengthens new behavior once changed” (Ambler 2000, p. 300). Instead he is pointing at the advantages of the weak theory (reinforcement) brought forward by Ehrenberg (1974) which has, according to Ambler, received less attention, even though the weak theory is perhaps more applicable to real world situations (Ambler, 2000). Without going further into the complexities of the Elaboration Likelihood Model, we will instead shift focus and look at the Elaboration Likelihood Model of persuasion and examine what it and the other models have to say about attention and comprehension. In ?gure 2.7 we can observe that the attention and comprehension construct precedes the two routes independently of whether it is a low or high level of involvement. Intuitively, this is a more appealing situation and solution as opposed to what was seen in the two previous models. The treatment of the concept of attention in the Elaboration Likelihood Model is rather similar to how it is treated in the model in ?gure 2.5. What have been witnessed this far in the discussed models, is an attention construct that was downplayed in ?gure 2.6 A since attention was considered as having an indirect path leading to memory and attitude towards the advertisement in the end. In ?gure 2.6 B attention and memory was separated from the main route. As stated earlier, it is rather paradoxical when observing that neither attention nor memory has any impact on the other constructs in the model. Consequently, attention and memory is of limited importance in forming attitudes and a purchase intention, if this would be assumed to be correct. This feature of the model in ?gure 2.6 B does not coincide with, for instance, Mack and Rock’s (1998, p. 14) notion that there is no perception without attention. Mack and Rock are stating that if there is a
39
reduced attention, there will be a reduced perception, which in turn will lead to less attitude change or attitude reinforcement. Neurophysiologic evidence is supporting this last position suggesting that there is a reduced processing of unattended stimuli, that is to say, stimuli that a subject has not been attending to (Eysenck & Keane, 2000). Tulving and Thomson (1973) are stating “Only that can be retrieved that has been stored, and … how it can be retrieved depends on how it was stored.” (Tulving and Thomson, 1973, p. 359). Hence, recall as well as recognition will suffer if the stimulus was unattended. Taken together, it becomes problematic with hierarchical models where the attention construct is detached from the attitude formation part and does not precede the attitude formation7. As a result, it seems commonsensical to state that, ?rst of all, an input in the form of a message has to be attended to, at least at some basic level, in order to be perceived. Moreover, the message or parts of it has to be placed in memory in the form of, for instance, an attitude towards an advertisement and an attitude towards a brand. If there is no kind of memory of this (unattended) stimulus, it becomes dif?cult to comprehend how the consumer will be able to recall or recognize a brand. A model that does not include attention in the initiation of a communicative process does not follow the logic of attention to a great extent and it becomes even more problematic when memory is set aside. When studying the two models in ?gure 2.5 and 2.7 it is apparent that these models do have a serial perspective and an attention construct placed in a serial route. The attention construct is placed before memory, as in the case in the classical hierarchical model, or before the central and peripheral routes as in the case with the Elaboration Likelihood Model. From these two models it can be deduced that the process stops if there is no attention. This is logical. When recapitulating the above discussions and in particular when taking into account Belk’s notion, there are factors in the surrounding, the environment or in the situation that has the potential to affect the communication and the attention to the communication. Since there are no constructs in the classical hierarchical model or the Elaboration Likelihood Model that are considering this, a general model where this construct can be found will be discussed.
It is at this point also relevant to mention that an attitude comprises of the cognitive, affective and cunative components (Tesser and Schaffer, 1990) and is a predisposition that is expressed by evaluating for instance a product with some degree of favor or disfavor. Furthermore attitudes are considered as being rather stable over time.
7
40
Summary Model of Web Ad Effects
Ad Processing Attention Learnning Emotions Acceptance
Figure 2.8
Ad attitude
Action 1. Brand inquiry 2. Brand purchase
Exposure (intended, content & Structure Content 1. Content nodes 2. Content node enhancers Structure 1. Micro structure (a) Links (b) Link Effectors (c) Molecules 2. Macro structure (a) No. of molecules (b) No. of links between molecules (c) Intended order (d) Navigation aid
Web and schema
Moderators 1. Web navigation ability 2. Category need 3. category expertis 4. Situational factors
Communication Effects Identification 1. category needs 2. Brand awarness Evaluation 3. Brand attribute 4. Brand attitude 5. Brand intention 6. Brand purchase facilitation
Figure 2.8 Summary model of web ad effects. (From John R. Rossiter and Steven Bellman, 1999)
The model in ?gure 2.8 is indeed detailed and takes into consideration additional aspects of the advertising phenomena, as opposed to the previously discussed models. The model is in fact an adaptation of a general advertising model that Rossiter and Percy presented in 1997 (1997). This model is dissimilar to the previously studied models in the sense that it is speci?cally designed to model Web advertising. Rossiter and Bellman are stating in their article that Web advertising has different characteristics than other kinds of advertising. Mainly, since the consumer is “actively skipping about searching for interesting items of information, often pausing to be entertained and perhaps missing or deliberately bypassing content that the advertiser thinks is important.”(Rossiter and Bellman, 1999, p. 15) This is an exquisite observation capturing a general picture of Web behavior among Web surfers. What they are describing in essence, is the attentional system in action, which is constantly engaged in selecting, including and excluding information when guiding through the environment – the Web site, where advertisers are competing for attention. This behavior is not unique for the World Wide Web. In fact it is very similar8 to the behavior when “sur?ng” through print material, such as newspapers, magazines and the like. In the model by Rossiter and Bellman above, the situational factors are taken into consideration. The situational factors are moderators affecting the Web ad schema and the advertisement processing where the concept of attention can be identi?ed. Even though the model is mentioning these constructs it is not speci?cally targeting the interrelation between attention and context represented as a task environment.
8
There are of course differences in the interactive possibilities though. 41
Moreover, it is not stressed in any of the other models either. Rossiter and Bellman’s model is the only one, of those examined here9, where situational factors are taken into account. Following the ideas of Belk, situational factors may have an impact on the process of attending to different stimuli and ought to be examined further. Prior to elaborating on the relation between attention and situational factors, it is important to discuss around advertising and the World Wide Web.
The World Wide Web as an advertising vehicle
The development of marketing communication on the Internet has been rapid, driven by a variety of companies exploring the possibilities on the Web. According to Hoffman and Novak (1997b) the Internet is a new “marketscape” that calls for a new marketing paradigm. This, since consumers are more active on the Internet than is the case with other media (Hoffman and Novak, 1996). That a new marketing paradigm is needed as a consequence of the emerging use of the Internet and the World Wide Web is perhaps to stretch matters too far. Nevertheless, the development of the Web poses a challenge to companies in a variety of industries and in particular to organizational functions related to marketing communication and advertising. An opposite stance that some are taking, compared to for instance Hoffman and Novak, is that the Internet and the World Wide Web is not very unique; it is just another communication channel among a number of channels. This stance is, on the other hand, perhaps not quite acknowledging the versatile nature of the Web or the radical impact that it has had in most parts of society. With the Internet, companies have gained a new platform to communicate to and with their target audience. This is possible through various means such as e-mail, chats, instant messaging, discussion groups, news groups, on-line gaming, hypertext links, short message services through gateways to GSM networks, ip-phone communication and the like. One of the most important and current possibilities with the Internet is to use it for advertising on Web sites and Web pages. Advertising on the World Wide Web has increased the last few years giving new opportunities for marketers to reach their customers. Advertising can be conducted in different ways on the Web ranging from static banners and advertisements, click through banners, traditional as well as interactive commercials “broadcasted” with streaming video, pop up advertisements (interstitials), triggered advertisements, or a combination of two or more of these advertising tools. The variation and combination of different kinds of advertisements are greater on the World Wide Web in comparison to any other type of media.
9
Rossiter and Bellman’s model is not the only model in marketing communication and advertising that have situational factors present, see for instance Belk. 42
Advertising on the Web has both differences as well as similarities when comparing to established advertising media such as TV, radio or print. A salient property distinguishing Web advertising from other kinds of advertising is that it has hybrid characteristics combining properties from print, broadcast, outdoor, and direct response media (Dréze and Zufryden, 1997). As a whole, these properties give Web advertising designers room for creativity and development of new combined types of advertising. However, advertising on the Web can be distinguished from other broadcasting media (television & radio) since the advertising message is integrated within the editorial material in the same manner as in print media (Chandon, Chtourou and Fortin, 2003). In television and radio, advertising messages and program content is presented sequentially. That is, program content is presented in a separated block which is interrupted by a block of commercials and subsequently a new block of program content is followed by a block of commercials and so on. Thus, the commercial block creates a “pause” in the program content that can be used for various activities10 by the consumers. The pause represents a possibility to divert the attention from the media and thereby the advertising as well. On the Web, advertisements are competing with the Web page content for the consumers’ attention in a parallel way. The consumers’ attention is thereby continuously directed to an area close to the presented advertising material. In the case of parallel presentation Web advertising is similar to print where advertisements are presented together with program content or the editorial environment. Nonetheless, the dissimilarities between Web and print that are often pointed at are most of all that the Web has a greater interactivity, greater complexity and a high level of ?exibility (Bruner & Kumar, 2000, Karson and Korgaonkar, 2001). There are of course differences in the visual appearance as well.
Web advertisements are placed on the World Wide Web
How to place and where to place a banner or an advertisement on a Web page has been studied by many researchers (see for instance Bruner; Shamdasani, Stanaland and Tan; Dréze and Zufryden; Novak, Hoffman and Yung) with the aim to obtain an understanding of how to gain the best effect out of an advertisement. The complexity of a Website has been a concern since prior research has suggested that increased complexity could reduce the effect of an advertisement (Stevenson, Bruner, and Kumar, 2000). Since the developments in Web design has been racing towards greater complexity with competing sights and sounds found on Web pages (Bruner G., 2000) it is crucial to understand how this affects advertising. One of the main reasons why Web design is heading towards greater complexity is because of the possibility to transfer larger amounts of data per time unit.
10
To fetch some snacks, go to the restroom, watch another program or the like. 43
This development of hardware infrastructure speeding up data networks has given Web designers new opportunities to improve Web design as well as Web advertisements. As a result, advertising on the Web today offers greater possibilities than it did just a few years ago. Increased complexity on Web sites (Bruner G., 2000) is created by using more graphic images, more text (even blinking or the like), more hyperlinks, more sound, more animation and also more full-motion video. The added complexity on Web sites makes them more attractive and enhances the perceived entertainment value and exclusivity. Still, as Web designers add these features to attract users, some studies from psychology and advertising suggest that interest may grow with complexity but not necessarily pleasantness. Already early research (Berlyne, 1960) has shown that pleasantness is optimal at some low to moderate level of complexity. Thus, the new way of designing and presenting Web sites may have negative consequences for the users’ perception and attitude towards the Web site (Bruner G., 2000). Berlyne (1960), was furthermore stating that people, in general, are more motivated to explore complex stimuli when they appear in a setting of simpler stimuli which is in line with the idea that increased complexity may affect the impact of advertising in a negative way. The Web sites and their various environments that is referred to here is to some extent a mirror image of the editorial environment which is the term often used for print media (Appel, 2000). Advertisements in print media are placed into the editorial environment and the editorial material surrounding an advertisement has been considered important as a factor in?uencing the effectiveness of the advertisement. Appel has elaborated around this issue and he is arguing that there is a strong relationship between editorial environment and advertising effectiveness. In Appel’s article he is arguing that when the credibility of editorial material in a magazine was low, it also affected the effectiveness of an advertisement in a negative way. When the editorial environment on the other hand was credible, the advertisement also scored higher on perceived credibility. Appel attributed this effect to composition of audience partly but mainly to the editorial environment. It is plausible that this “editorial environment” effect that Appel is discussing around is present in a Web setting as well as in print. In line with Appel’s ?ndings, there are studies showing that the same source delivering the same message to the same audience can produce different effects depending on the editorial context in which the message was presented (Chaiken and Stangor, 1987; Cooper and Croyle, 1984; Collins, and Miller, 1969). Context has also been shown to be important since it can affect recipients’ judgments of advertised products (Puto, 1987; Smith 1996; Woodside and Singer, 1994) Shamdasani et al (2001) has studied the vehicle source as a variable affecting the effects of an ad. They are stating that “source variables … essentially moderate
44
the impact of a persuasive message by in?uencing the audience’s con?dence in the message endorsed by the source” (Shamdasani et al 2001, p. 8). There are in other words variables that are connected to the source and not the message that affect the impact of the message. Source credibility is important since when credibility is high, consumers counter-argue less with advertising claims and are therefore more in?uenced by the message (Grewal, Gotlieb, and Marmorstein, 1994). When source credibility is low, conversely, consumers discount the arguments made in the message. Thus, messages presented by a highly credible source are more readily accepted and induce greater attitude change in consumers than a source with low credibility (Harmon and Coney, 1982; Sternthal, Dholakia, and Leavitt, 1978). In a study of Web advertising recall and recognition (Mullarkey and Danaher, 2003) the authors found that factors such as duration of page viewing, Web page context factors and viewing mode to be crucial and part of the determinants of advertising effect. The impact of the duration of Web page viewing is consistent with studies conducted of televised advertising. Researchers (Krugman D., Cameron and White, 1995, and Krugman H., 1986) showed that the more time spent viewing and attending to an advertising media, the more advertising content respondents tended to remember. The connection between attention to advertising and viewing duration is tied to the information processing that the consumer is occupied with. A greater opportunity (reinforced by time) to attend to an advertisement and to process information result in a positive effect on recognition. This is also being supported by studies using neuroimaging (Rossiter, Silberstein, Harris and Nield, 2001). The viewing mode in the study by Mullarkey and Danaher (2003) is described in a similar manner as in the study to be conducted here. Mullarkey and Danaher considered it reasonable to divide Web surfers into two categories representing two different modes, i.e. “goal directed” and “sur?ng”. Their terminology is brought in to the advertising area by Janiszewski (1998) who is arguing that there are two broad types of behavior namely, goal directed search and exploratory search. Goal directed visual search behavior has been shown to reduce attention to peripheral stimuli, whereas exploratory search behavior gives peripheral stimuli a chance to be processed. Mullarkey and Danaher argue in line with Janiszewski’s ?ndings stating that when consumer viewing or use is directed toward a particular task there would be a reduction of the impact of advertising. A less goal directed and exploratory sur?ng behavior would on the other hand result in more browsing of media content, including peripheral content such as advertisements. However, a limitation of their study was nonetheless that the respondents never engaged in real Web sur?ng activities since the experiments were designed, and limited, in such a way that the respondents merely watched Websites that were automatically switched every 20, 40 or 60 seconds. Thus, the real hands-on sur?ng never occurred in that study. Since there was no hands-on sur?ng there were no interaction, clicking or sur?ng either which are the hallmarks of Web – user interaction.
45
Web page context factors were also addressed in the Mullarkey and Danaher’s study and in line with other authors (see for instance Bruner G., 2000) the context factor was described as the design and layout of the Web site. The main context effect taken into consideration was how design of the Web page and how competing stimuli “…and potential page cluttering or excessive complexity could affect reception of advertising” (Mullarkey and Danaher, p. 255, 2003). Similar studies of Web context effects have been focusing on Web page background complexity (Bruner and Kumar, 2000, Stevenson, Bruner and Kumar 2000), and also the Web page complexity’s impact on page access (Dreze and Zufryden, 1997). In summary it is adequate to say that Web advertisements are placed in a Web setting. The Web setting is essentially very similar to its print counter part “the editorial environment”. The general development on Web sites the last few years has been towards greater complexity and more competing stimuli resulting in a developed Web setting. The added complexity on Web sites makes them more attractive and enhances the perceived entertainment value and exclusivity. This is an evolution that can on the one hand increase pleasantness but on the other hand reduce overview and access. Simultaneously, being a context variable, complexity is affecting the task environment moderating how attention is distributed towards advertisements. Thus, changes in complexity and environment have occurred and are occurring that may have disadvantageous in?uence on advertising.
The advertisement
Another obvious factor in?uencing the total advertising effect is naturally the advertisement itself. Hitherto, we have not been dedicating much attention to the advertisement itself other than factors surrounding it and affecting the advertising effect indirectly. Next, focus will be on the advertisement and the properties of an advertisement that affect attention, among other effects that can be measured. One of the main principles with advertising is to attract attention (Pieters & Wedel, 2004), in order to get the opportunity to present the company offer to the consumer. However, it is becoming increasingly dif?cult to attract attention because of increased clutter and noise. For advertisers the quest to reach customers is more demanding today when, for instance, half of the material in a typical magazine consists of advertisements. Consequently, the inability to capture consumers’ attention has become more important and prioritized (Adler & Firestone, 1997; Davenport & Beck 2001; Pieters & Wedel, 2004). Some are even claiming that “the power of marketing is eroding … from lack of attention” (Sacharin 2001, p 3. in Pieters & Wedel, 2004). The pursuit to attract attention is a complex one and has many paths. Part of it is the message that is being transferred to the consumer and in particular the properties of the advertisement. That size plays a major role and has a positive correlation with attention is a general notion in advertising (Finn,
46
1988; Abernethy & Laband, 2004) and has recently also been supported by research on Web site advertising (Baltas, 2003). Except for size, there are many other characteristics of an advertisement that have been studied by researchers. In the work of Finn, other characteristics (for print ads) have been shown as in the ?gure below.
Advertisement Characteristics
Size and location Ad size Cover position Facing: ad / editorial Right / left page Layout and pictorial Color Illustration size Photo art Bleed / no bleed Copy Amount Readability Benefits Other Characteristics Headline: Words Phrases Nouns Verbs Adjectives Determiners Type size Product reference Question form Benefit Product as object Figure 2.9
Figure 2.9 Advertisement characteristics. Adapted from Finn (1988).
Of the characteristics listed in ?gure 2.9, not unexpectedly, advertisement size was in fact the property that had shown a signi?cant effect most frequently in the material that Finn reviewed. However, aside of the characteristics stated above later researchers have pointed at a few central characteristics of an advertisement that are important to take into consideration. According to Pieters and Wedel (2004) there are three key elements that can be identi?ed in an advertisement, the brand, the picture and the text. Their position is that these three elements have distinct effects on attention. In Pieters’ and Wedel’s study they found that the pictorial was superior in attracting attention, independently of its size. Other researchers are also reporting that the pictorial is by far the “most important structural element in magazine advertising” (Rossiter & Percy, p. 295, 1997) and studies (with eye movement tracker) are in fact showing that around 90% of the viewers ?xate the main picture in an advertisement before they start focusing on the copy text (Kroeber-Riel, 1984). The enduring effect after viewing the advertisement is the formation of a visual memory of the advertisement that enables subsequent recognition (Finn, 1988). These ?ndings presented here should lead us to convert to the standpoint that advertisements should seldom be designed without adding a pictorial element to it. In that case (without picture) we would suffer the bene?t of base line attention to the pictorial element. Furthermore, the pictorial element is also supposed to lead the reader or viewer to the other elements of the advertisement. The purpose is not only to attract attention but to propel readers to engage in reading the brand and text elements.
47
Not only the pictorial elements but also the text elements are considered as being important and even assumed to be able to capture attention. Text elements are usually divided into two main categories namely, the headline and the copy text. The headline is supposed to be an eye catcher capturing the readers’ attention and being interesting enough to get the reader to continue with the copy text in the advertisement. One of the proponents of the importance of headlines is David Ogilvy. One of the more well-known statements in advertising originates actually from Ogilvy saying that “The wickedest of all sins is to run an advertisement without a headline” (Ogilvy, 1963, p. 130, author’s italics). Some advertisers hold the view that the headline is often perceived as the most vital advertisement component (Pieters & Wedel, 2004). This position among practitioners is however challenged by researchers such as Kroeber-Riel who is pointing out that respondents have a preference for attending to pictorial elements and that they in fact avoided, if possible, “the cognitive effort of reading the verbal elements when skimming ads” (Kroeber-Riel, 1984, p. 593). The text elements are though important in that they can communicate speci?c information about the product offer that is more complex and that cannot be conveyed by the pictorial element. The ?nal element, the brand, comprises of the brand name, trademark and logo (Pieters & Wedel, 2004). The brand name is a potential carrier of image, value and status communicating this information in a subtle way to the audience. There are scholars arguing both for maximizing and minimizing the brand size in an advertisement. Those in favor for a brand that has a prominent position in an advertisement argue that the brand is important to remember and also that it should be easily identi?ed. If the advertisement stands out in comparison to competitors it will attract more attention and generate favorable brand communication effects (Pieters & Wedel, 2004). These properties as a whole will for instance help the consumer recognize and choose the brand when entering a buying situation. Those in favor of a less salient brand placement in an advertisement claim that the brand element signals that the message is an advertisement and since the consumers are not interested in advertising it will be counter-productive. Hence the brand strategist or the advertiser ought to downplay the size and location of the brand in the advertisement. It is worth to notice that it is commonly assumed that consumers are not interested in any advertising whatsoever. As a matter of fact it is assumed that consumers “hate” advertising (Aitchinson, 1999 p. 61. in Pieters & Wedel, 2004). This is a rather extreme position. In a print advertisement context and potentially also in a Web context it is important to take these three key elements (the brand, the picture and the text elements mentioned previously) into consideration when designing an advertisement. Therefore, the three elements discussed have been taken into consideration when developing and designing the experimental advertisement so that the experimental advertisements have these generic elements. However, these elements will not be manipulated since they are not unique to Web
48
advertisements. Instead it is the speci?c Web dimension of advertisements that will be altered. In Web advertising there are additional properties than the brand, the pictorial and the text that may generate increased/decreased attention to the advertisement. A Web advertisement can have properties that reach beyond the list presented in ?gure 2.9, properties that are, to some extent, unique for Web advertising. It is these properties, animation and motion together with abrupt appearance that will be manipulated. In the following it will be elaborated around this extra dimension that Web advertisements have compared to for instance print advertisements.
The Web dimension of advertisements
When studying different World Wide Web advertisements and what various authors have written about Web advertising11 one will soon discover a diversity of types of advertisements and also a general lack of classi?cation. Most authors are referring to and studying the banner advertisement, hitherto few are describing its properties12. Yet, development of the WWW has brought new advertisement tools, which in turn has made it evident that these tools have to be classi?ed. Classi?cations have however been poor or at least not up to date with the rapid development in terms of advertisement design and also development of information technology infrastructure, making it possible with larger and more demanding advertisements on the Web. Despite the rapid development in some parts of the world, researchers are still referring to the banner advertisements as, “…the most common form of advertising currently” (Danaher and Mullarkey, 2003, p. 253), which is a small billboard-like graphic that appears on a Website that is clickable (Hoffman and Novak, 2000), and furthermore, “…is a rectangular shaped image typically located at the top of a Web page” (Chandon, Chtourou and Fortin, 2003, p.217).
11
There is a wide variety of advertisements and advertisement formats that is being used on the Web. The more frequently used advertisements will be discussed with a main focus on those types of advertisements that are the most relevant for this study. 12 A reason why few authors bother de?ning web advertisements might be the fact that the banner advertisement was the ?rst kind of advertisement that was used on a larger scale and they are now taken for granted. However, development of the WWW has brought new advertisement tools making it inappropriate to classify all web advertisements as banner advertisements. Banners were the ?rst commercial form of online advertising and were created by the staff of the magazine ‘Wired’. Banners quickly became adopted as an advertising standard generating revenues for publishing and content sites (www.rightonthespot.com). 49
Web Advertisements
Advertisement in the shape of a small square
Figure 2.10 Avertisements in the shape of traditionional banners Advertisements with a normal rectangular shape
Figure 2.10 Web advertisements in different sizes and shapes.
The notion that a banner advertisement is a rectangular, small billboard-like graphic and also the most common form of advertising currently, does not ?t well with the kind of advertisements that is frequently being used on many north-European Web pages as well as international ones. Pilot study number two pointed in the direction that Web advertisements nowadays have little or no resemblance of their banner counterparts13 from the mid-nineties. The one-colored, static banners that were being used at that time are infrequently used on today’s Web pages. Since modern Web advertisements are becoming increasingly similar to print advertisements as opposed to banners, one could argue that it is about time to rename these modern Web advertisements. In addition, when taking into consideration the fact that banner advertisements have the purpose to evoke click-through at least for high-involvement and functional products (Dahlén, Ekborn, and Mörner, 2000; Dahlén and Bergendahl, 2001; Dahlén, 2002) and that the click-through rates have fallen to below 1 ‰ (or 1 in 1000), it seems as if banner advertisement’s prime time has reached its peak. Still, there are banner advertisements that are being used and that will be used although new sizes and formats are also appearing. Even though many researchers are still holding on to the banner advertisement, the Interactive Advertising Bureau - IAB, is leading the way forward and have presented a new system for categorization of advertisements, which include more updated formats14.
13
The ?rst banners employed on the WWW were often one colored with just a simple text with the brand name on. 14 The IAB is working to “…familiarize publishers, marketers, and agencies with Interactive advertising standards and guidelines formulated and developed by their peers in association with the IAB. Use of these standards and guidelines directly organize the industry so as to foster an environment in which the Interactive medium is implemented with ease, thus allowing the industry to buy more and capture value.” (IAB Standards and Guidelines 2004). 50
The Interactive Advertising Bureau have developed general standards, classifying advertisements into two broad categories, the “In-page” advertisements and the “Out-of” page advertisements. An In-page advertisement is a Web advertisement presented directly on the Web page itself15. An Out-of page advertisement is an advertisement that is not presented directly on the Web page. An Out-of page advertisement could, for example, take the form of a popup advertisement that will appear as soon as one has clicked on a launching link or when entering a Web page or the like. Out-of page advertisements have furthermore been subcategorized into “Over the page advertisements” (pop-ups), “Between page advertisements” (transitional advertisements) and “In-stream advertisements” (Streaming video commercials, see Appendix E). The Out-of page advertisements have in common that they are advertisements presented in a separate window which is separated from the actual Web page that the user is using and visiting. In-page advertisements can have different sizes, shapes and any property that can be presented on a Webpage. It is usually presented in conjunction with program content on the Web site and is thereby competing with the program content for the user’s attention (Chandon, Chtourou and Fortin, 2003). In-page advertisements are either static advertisements or animated advertisements. The use of static Web advertisements are declining and in Chandon’s study only 7.2 % of over 1200 advertisements studied were not animated. Their study, that measured click-through rates, indicated that animation is bene?cial to advertisers since it appears to improve click-through rates. Kisielius and Sternthal (1984) found that vivid information attract user’s attention easier due to its sensory effects. On the Web this means that animation may increase the likeliness to attend to the advertisement (Zhou & Bao, 2002). Another study conducted by Lohtia, Donthu and Hershberger (2003) indicated that the animated advertisements increased the click-through frequency and the authors are strongly recommending the use of animation. In a study by Sundar and Kalyanaraman (2004) arousal and memory were studied as a function of animation speed. Sundar and Kalyanaraman’s ?ndings are pointing in the same direction. That is, animation has an impact on measures such as arousal and recall with an increase when animation is speeded up. Their study deviates from the former studies since it is not based on click-through frequencies. Instead, their study used a physical measure, namely arousal which is measured via electrodes during the session, and also recall which is a measure often employed in advertising research. In these reported studies, recognition has not been used as a main measure, a measure that for instance Krugman is advocating when measuring advertising effects (Krugman, 1986). An out-of-page advertisement is a group of advertisements comprising of “overthe-page” advertisements, “between-page” advertisements and “in-stream” advertisements. In-stream advertisements and between-page advertisements, on the other hand, will not be covered in this study.
15
A banner advertisement would of course belong to the category In-page ads. 51
An over-the-page advertisement is a category of advertisements with the characteristic that the advertisement will appear over the Web page that is being viewed by the user. The main advertisements in this category are the well-known pop-up advertisements (also called interstitials) and also the ?oating advertisements. Floating advertisements are advertisements that appear when one enters a Web page, and the advertisements “?oat” or “hover” over the page for a short period of time, typically 5-10 seconds. Depending upon the attention getting strategy chosen, the ?oating advertisement might be placed either in a peripheral position or in a more central position obscuring the view of the Web page (Pilot study 2). The second type of advertisements belonging to the over-the-page category is the pop-up advertisements. The characteristics of the pop-up advertisements are that they “[…] utilize a Web-browser initiated additional window to deliver an advertisement impression either directly above or below the existing browser experience”16 (IAB report, 2004). The pop-up advertisements have two features that distinguish them from their static counterparts, namely some degree of surprise since they tend to appear from nowhere when the Web user enters a Web page or clicks on a certain hyperlink or some similar event. These features contribute to enhance the attention getting effect. The pop-up advertisement also adds motion or visual disturbance to the Web page area where it pops up. This is equivalent to the effect of animation even though the pop-up itself might not even be animated. According to Beard (2001), the pop-under advertisement has become widely used in the industry. The reason for this is that the pop-under is a salient piece of stimulus since it appears on the screen in an unexpected manner and without the user’s consent. In doing so, it is more effective compared to traditional banner advertisements. The attention getting features of the pop-under advertisements are equally relevant for pop-up advertisements as they appear on the screen in an unexpected manner and without the user’s permission. One of very few systematic studies on pop-up advertisements, by Diao and Sundar (2004) found that pop-up advertisements were superior in achieving ad-recall. At the same time they did not observe this effect for ad-recognition, contrary to their own hypothesis. On the other hand, they used a somewhat different methodology in comparison to this study. They used for instance heart rate in beats per minute to measure the response of the respondents. Furthermore, they relied on data that could be confounded. For example, they did not take into consideration differences in size between advertisements that were compared. Such a procedure is risking the validity of the results.
16
When presented under the existing browser it is usually called a pop-under advertisement.
52
Research on basic forms of animation found that moving or ?ashing objects could be useful in attracting attention from one area of the screen to another as a consequence of its visual distinctiveness (Cropper and Evans, 1968, Smith and Goodwin, 1971). An alarming negative ?nding regarding animation was that it could be “distracting”, “obtrusive”, “disruptive” and “fatiguing” (McCormick 1970, Stewart 1976). Later research has, to some extent, supported these ?ndings stating that the effect of animation at times can be “annoying”, “irritating” and even “evil” (McGalliard 1998, Nielsen 1996, 1997, Spool et al. 1999). Hong, Thong and Tam (2004) have studied the phenomena of animation from an information systems approach and found that ?ash indeed is a salient feature that has the potential to capture attention. A negative aspect (from the perspective of an advertiser or a Web site publisher) is that this attention getting technique can, under certain conditions, lower users attitudes’ towards using the Web site in question (Hong, Thong and Tam, 2004). Another problem less focused upon is the aggregate effect of attention getting techniques on the users of the World Wide Web as a whole. The more attention getting techniques used, the more mental effort that will be used to avoid advertisements, and in particular advertisements that are non-relevant from the user’s perspective. One should not simply assume that there are no countermeasures against attention getting techniques. The users’ attention system will use more resources to screen out advertisements. This could then result in what Cho and Cheon (2004) call “Banner Blindness”, which is a negative development for Web advertising in general. Regarding the pop-up advertisement, which is one of the most attention getting advertisements, there is the possibility to close them by clicking on the small “x” in the upper right corner on the pop-up window. According to one study, a majority of Web users are closing down the pop-up window within 3 seconds (Eyetrack, 2004). Another strategy is to not close the popup down but rather to avoid looking at it and after about 25 seconds to click somewhere else, which will cause the pop-up to disappear (Eyetrack, 2004). The pop-up advertisements are not only becoming a victim to Web users that are clicking them away. There are other initiatives taken by organizations where the purpose is to avoid advertisements (in this case pop-ups). Software producers developing Web browsers have sometimes added functions that are screening out pop-up advertisements. In Windows XP, for instance, the pop-up suppressing function is on, in default mode. This has of course been counter balanced by new versions of pop-ups that are passing these ?lters. It was indicated earlier that practitioners are in favour of animated advertisements compared to static or non-animated advertisements. The non-animated advertisements’ share of all Web placed advertisements have decreased and was about 7,2 % in 2003 when Chandon et al conducted their study. It is believed that the static advertisements are less effective as opposed to animated in-page advertisements and pop-up advertisements. The pop-up
53
advertisement is in fact a special case of an animated advertisement (no matter whether the content in the pop-up window is animated or not) since it “uses” motion in the area where it appears, but is different compared to its in-page counterpart as it can be moved around or clicked away. Advertisers are eager to use attention getting techniques to attract users to their advertisements yet there are risks connected with this as well. It seems that the advertiser has to walk the ?ne line between getting attention and evoking fury, at worst. In summary, Web advertisements have additional features, compared to for instance print, that can be used to attract attention. These attention getting techniques brought up here is: a static advertisement – where no technique is being used, an animated advertisement – where some kind of motion picture is being used and ?nally pop-up – where the advertisement is being abruptly presented somewhere on the Web browser creating disturbance and motion. The purpose with these attention getting techniques is to attract the attention of the audience. In the previous it has been discussed around the attention construct to quite some extent without going very deep into the area. Next, a deeper review of the attention construct will be done in order to gain further insight into this imperative aspect of advertising.
The concept of Attention
The concept of attention is central to advertising. In order for the consumer to perceive an advertising message she will have to distribute at least some attention to it. Otherwise the message will go by unnoticed; an unwished event by most advertisers. Therefore, attention is important for advertisers and their aim is that the audience a ttends to the commercial messages that they are communicating. Thus, advertisers commonly use attention getting techniques such as animation, pop-up, ?ash or other means in order to attract attention. With this said it is appropriate to examine further what attention is and how attention can be related to advertising. What is attention? And how is attention de?ned and described? William James once wrote:
“Everyone knows what attention is. It is the taking possession of the mind, in clear and vivid form, of one out of what seem several simultaneously possible objects or trains of thought.Focalization, concentration, of consciousness are of its essence.”(William James, 1890, pp. 403 – 404 in Eysenck and Keane, 2000).
James description of attention is a rather early one nevertheless like others of more recent date he is pointing at selection as a core feature and also at the bringing into consciousness. Mack and Rock (1998, p. 25) are de?ning attention as a “process that brings a stimulus into consciousness. It is in other words, the process that permits us to notice something.” This view and description of
54
attention is the one used further in this study and is also in line with the notion that has been expressed when discussing the advertising models previously. A stimulus can for instance be an advertisement or any other event that can be brought into consciousness. The fact that Mack and Rock are stating “the process that permits us to notice something” implicates that a reduction of the process of attention will at the same time also reduce what is being noticed and brought into consciousness. The major conclusion of Mack and Rock’s very extensive research in the area of attention is that there is no perception without attention. Perception in turn is the means “by which information acquired via the sense organs is transformed into experiences of objects, events, sounds, tastes etc” (I. Roth 1986 in Eysenck and Keane, 2000). In order for information to enter the internal system of a subject this individual will ?rst have to distribute attention to the information so that it can be perceived and thereby be further processed. From an advertising perspective this would clearly mean that subjects ought to attend to an advertisement, at least at some minimal level, if we are to expect them to be able to recognize an advertisement or product at a later stage. Consequently, we have to capture the audience attention so that they can perceive what we are communicating. However, what means do we use to capture the audience’ attention? Earlier in this chapter it was highlighted that advertising researchers have found that advertisement properties such as picture and headline are said to capture attention. The pictorial in particular seemed to be an important means in attracting attention and Kroeber-Riel’s study indicated that subjects actually tried to avoid the text part in favor for the picture part. It was also discussed that size had been reported to have a positive correlation with the capture of attention. Furthermore, motion, ?icker and ?ash have also been identi?ed as properties that can capture a subject’s attention. The properties of objects that can capture attention have been studied extensively within the theoretical realm of visual search. In the area of visual search quite some research has been conducted in regards to investigate what a basic feature is and how humans are performing visual search tasks. A stimulus that supports both effortless texture segmentation and ef?cient search is usually perceived as being a basic feature. Effortless texture segmentation can, for instance, be described as when a region of green spots in a ?eld of red spots is instantly segmented from the background or when spots at one stereoscopic depth among spots at another stereoscopic depth are immediately identi?ed (Wolfe, 1999). Ef?cient search, on the other hand, is when an increase in set size (set size is the number of distracting objects in a visual search task) does not produce an increase in reaction time. When looking closer at the various stimuli and their properties it can be observed that color is a typical example of a basic feature and is one of the best ways to make a stimulus “pop out” (Bundesen & Pedersen, 1983).
55
Generally, a greater difference between colors will make it easier to discriminate between targets and distracters (Nagy and Sanchez, 1990). Orientation is another basic feature and refers to, for instance, the orientation of multiple lines, where one line deviates in orientation from the other lines (Foster and Ward, 1991a). Curvature and vernier offsets are also considered as basic properties and refer to a curve among straight lines and a broken line among straight lines respectively (Treisman and Gormican, 1988; Levi et al., 1985). Size, is yet another basic feature that has been extensively studied in various ways and shown to possess the properties to be easily identi?ed (Bilsky, Wolfe and Friedman-Hill, 1994). Motion, is an uncontroversial basic feature as it is easy to ?nd a moving stimulus among static ones whereas the opposite, to ?nd a static among moving spots, is more dif?cult (Dick, Ullman and Sagi, 1987). Other basic features are depth and gloss since both produce ef?cient search times (Enns and Rensink, 1991; Nakayama and Silverman, 1986; Wolfe and Franzel, 1988). These basic features can be used to guide the subsequent deployment of attention (Pashler, 1998). The deployment of attention can in turn be stimulus driven (also called bottom-up) or user driven (top-down). Stimulus driven, means basically, that the external target is suf?ciently different from its distracters so that the stimuli “pops-out”. User driven, means, that we deploy the attention to a stimuli that we have decided being worthy of our attention (Pashler, 1998). Information from top-down and bottom up analysis of stimuli is used for attentional prioritization where attention will be directed to the item with highest priority (Wolfe, 1994). Braun has studied the bottom-up and top-down distinction by designing an experiment with a demanding task at ?xation to see what attributes of a peripheral stimulus could be identi?ed. Many of the basic features described above could withstand this test (Braun 1993; Braun 1994; Braun and Sagi, 1990). This is taken as evidence that attention can be seen as active for the central task while it is passive for any peripheral task (Braun and Julesz, 1994). It could also be argued that this means that preattentive processing continues across the visual ?eld while attention is busy somewhere else (Wolfe, 1998). The basic features have also been of interest in regards to attention capture and the properties that attract attention have been under scrutiny in experimental psychology and debated elaborately (see for instance Yantis, 1993 a & b, Yantis and Jonides, 1996, Gibson, 1996, Turatto, Galfano, Gardini and Mascetti 2004, Franconeri, Hollingworth and Simons, 2005). The perspective has been somewhat different and much of the interest has been directed at understanding basic properties and their in?uence on attention. Some of these basic features that have been studied are color, numerosity, motion, ?icker, shape, location and texture segregation (Mack and Rock, 1998). An early notion was that some of these properties were pre-attentive, perceived without attention, but was reinterpreted to mean that “numerosity, location, color and motion were properties that could capture attention” (Mack and Rock, 1998, p. 14). However, later research with more sophisticated methods, under an
56
inattentional paradigm, showed that not even these properties were able to capture attention. Instead, Mack and Rock argued that among a long array of features it seemed as meaningfulness and size were the properties that could capture attention. Other researchers in the area like Yantis (1993a) and Yantis and Jonides (1996), have argued that what is really capturing attention is the appearance of a new visual object in the visual ?eld. What is already in the visual ?eld will be part of the stable visual setting and only new visual objects will have the capacity to capture attention. Yantis and Jonides are however stating that abrupt visual onsets may not be unique in capturing attention. When studying the experimental research done in psychology, it is important to remember that the experiments are conducted with visual searches for stimuli with basic features. Yet, in the real world most people are rarely looking for a “T” among “X’s” or a “red dot” among “orange dots”. Moreover, we also have to keep in mind that these studies that have been referred to often use very short duration times of the stimulus presentation. Typical duration times are well below 1000 milliseconds. Despite this obvious dissimilarity between how attention capture is described and studied in experimental psychology, it is of importance to have an understanding of the basic features that are considered to have the power to capture attention. Remarkable, typical properties that are believed to capture attention, such as motion, ?icker or color were not found (for instance in Mack and Rock’s work) to attract attention at a 200 millisecond exposure. Instead, size together with meaningfulness was found to be decisive. Meaningfulness was for instance measured in an experiment where the respondents’ names were exposed and compared to other names. A particularly surprising ?nding was that subjects did not observe a “word”, presented for 200 milliseconds, almost indistinguishable from their own name; i.e. if a person’s name was Jack and it was changed into Jeck then the subjects were largely blind to the stimulus. This strongly indicates that a stimuli with a high level of meaning has the capacity to break through and capture attention (Mack and Rock, 1998). On the other hand, one should have a somewhat ?exible view when considering these results since a typical advertising exposure can last for much longer than 200 milliseconds. It is not uncommon that an advertising exposure will last for minutes and under these circumstances properties such as motion, ?icker and color may have a different impression on subjects. Presumably, because a longer time period will signi?cantly increase the chance a person distributes some degree of attention to a certain area. That is, the area where the advertisement is presented. In addition, interesting in this context is the fact that Mack, in Mack & Rock’s study of approximately 5000 respondents, never found any gender differences. The belief is that if the critical stimulus is equally salient and meaningful for women as for men, attention capture will be equal for women and men.
57
Following is a discussion into the area of how attention works when screening the world around us. It will be explored how attention works when conducting everyday tasks, such as reading some news on a Website while being exposed to an advertising message.
How the attention system operates
The world around us ?lls our senses with information, but only a small fraction of it is relevant to our goal-driven thought and behavior (Milliken and Tipper 1998, p. 191). As a consequence, we have to process information selectively since we otherwise would be overloaded with inputs. An intelligent behavior demands selectivity and that we direct our attention to certain targets that sometimes are external objects and at other times internal thoughts. This selectivity appears to be an enhancement of processing of relevant information and seem to dominate our mental activity (Milliken and Tipper, 1998). In recent years, consensus seems to have been reached that inhibitory processes play a vital role in selective attention. Broadbent’s (1958) ?lter theory, Walley & Weiden’s concept of lateral inhibition (1973) and Moran & Desimone’s (1985) single cell recordings share the view that a crucial function of the attention system is to prevent central mechanisms from interference and overload due to irrelevant input (Milliken & Tipper, 1998 p. 195). This means that the processing of unattended information is impeded and conversely that the processing of attended information is enhanced. Mack & Rock (1998) have also studied the phenomena of inhibition in detail. As has been discussed earlier, in order to achieve a deeper processing of information to take place, the subject ought to attend to an object. Merely looking at an object does not guarantee the fact that the subject is attending to it. As a matter of fact, they found that inhibition has the potential to be the greatest for stimuli presented at ?xation. They are also stating that since evidence of inattentional blindness17 is greater at ?xation, “attention can actively be inhibited from operating on input from some particular spatial location, and this was con?rmed in a series of experiments”.
17
A related phenomena that Mack & Rock were studying at the time. Subjects failed to detect a supratreshold stimulus presented at fovea and this phenomena was named inattentional blindness. 58
Focused and Peripheral Attention
Center of Attention
Figure 2.11
Less processing of events away from C.A.
Figure 2.11 When attention is directed to the cross within the circle then mental effort and processing will be impeded outside of this area. Less focus will be directed to events that are taking place for instance at the vertical line on the right of the circle.
Not only can attention be inhibited, Milliken and Tipper (1998) argue that attention can be used to generate expectation, to concentrate on ongoing processing demands of a primary task or used to direct behavior selectively toward an event among competing events. Still, trying to perform too many tasks at once in a certain task environment will have a negative impact on how these tasks are being performed. The best performance is achieved when the tasks are done one at the time. Attention researchers are often studying this kind of phenomena by presenting a primary task to the subjects to solve and then measuring the performance on a secondary task that overlaps temporally with the primary task (Milliken and Tipper, 1998). Of interest is to study to what extent the response to the secondary task suffers (Pashler, 1984). Kahneman (1973) states that attention refers to the mental effort or cognitive capacity allocated to a task. In this case a limited pool of resources is used by cognitive processes to perform tasks. The more demanding a task, the more cognitive resources is needed for the task. Therefore, performance of a secondary task will be constrained by the resources needed for the primary task. Another view of the phenomenon is that inef?cient performance in dual-task situations is due to an attentional bottleneck that is inherently serial in nature (Milliken & Tipper, 1998). The poor performance on a secondary task would thereby be explained by the fact that the primary task is blocking the serial stage (the bottleneck) forcing the secondary task to be delayed. This kind of explanation would suggest that individuals are not quite capable of diverting mental capacity to two or more tasks at the same time. Instead, it is in fact a matter of sequential processing.
59
The Limited Pool and Attentional Bottleneck Contrasted
Attentional bottleneck Serial view Activity A Activity B Activity C Activity D
Figure 2.12
Limited pool of resources Parallell view Activity A Activity B Activity C Activity D
time
0,0 msec
200 msec
Figure 2.12 The limited pool and attentional bottleneck concepts contrasted. Note that this is just a schematic view to describe the different perspectives.
Common for the limited pool and attentional bottleneck view is that a certain task environment may constitute of a primary task, setting the stage for any other task in that environment. A secondary task will be affected by the demand structure of the primary task in such a way that a more demanding primary task will reduce performance on the secondary problem and other events calling for attention. Many researchers (see for instance Broadbent, 1958, Deutsch & Deutsch, 1963, Treisman, 1964, Walley & Weiden, 1973, Kahneman, 1973, Moran & Desimone, 1985) have modeled attention elaborately. Among these theories on attention Treisman’s model seem to be more appreciated and adequate as opposed to other models (Johnston & Heinz, 1978). Treisman’s view on attention is that there is a sensory register that allows stimulus input to pass, an attenuator that reduces or attenuates processing of information outside focal attention. As a result, there is a limited amount of information that can reach short-term memory where information is stored temporarily and used for different mental processes (Treisman, 1960, Eysenck and Keane, 2000). The components of Treisman’s attenuation theory are sketched in ?gure 2.13 below.
60
Treisman´s Attenuation Theory
Sensory register Attenuator
Figure 2.13 Short term memory
Limited capacity
I n p u t
Figure 2.13 Treisman’s Attenuation theory. (Eysenck and Keane, 2000)
Treisman’s model is different from Broadbent’s selective ‘?lter’ model. Broadbent’s idea was that non-attended input was mechanically screened out (Eysenck and Keane, 2000), which was shown to be incorrect by Treisman (1960). The point with Treisman’s notion is that there is some kind of a barrier that is allowing attended input to pass through whereas non-attended input is attenuated. In her article from 1960 Treisman refers to, as mentioned above, an ‘attenuating’ mechanism. In her view this mechanism may signi?cantly reduce the incoming input, but not necessarily block information in the same way as in Broadbent’s ?lter (Treisman, 1964, 1960). In a more recent work, Pashler refers to the ?lter or attenuator (if using Broadbent and Treisman’s terms) as a “?ltering mechanism” (1998). This term, “?ltering mechanism”, will be used in the following since it is straightforward. From an advertising model perspective it is adequate to state that the model in the ?gure above is opening up the box of attention and giving an elaborate description of how incoming stimuli is being processed. This kind of detailed explanation and modeling infused into an advertising frame can provide greater insights and understanding about how attention operates in advertising. In addition to the descriptions of attention that has been presented, it is also important to consider the different impact of intentional and unintentional attention to an advertisement. From an advertising point of view the best case is, most often, when the individual is intentionally distributing attention to the advertising message. In that case, the individual will have time to read, see and comprehend the entire message communicated. There are of course different levels of intentional attention. One individual may explicitly direct the attention to the advertisement and study it in detail, whereas another individual may brie?y, and intentionally, glance at it. In the case of intentional attention, the advertisement is in the center of attention and not a peripheral object. This means that there is a much greater chance for processing the advertising message.
61
Unfortunately for advertisers, this type of intentional and high level of attention to a message is rarer and increasingly less common in advertising. Instead, the majority of encounters with advertising are brief exposures (Krugman, 1986). These exposures may be unintentional and the individual may not even be aware of the fact that she has seen or perhaps even deny having seen (Krugman, 2000). Brief or involuntary exposures can take place, for instance, when screening a page with a number of advertisements placed together with program content. In this screening process, the individual is trying to localize where the program content is in order to direct focal attention to this area. When engaged in this screening the individual will, simultaneously, be exposed to advertising content when the gaze is wandering over the surface to be screened. In relation to the phenomena of brief exposures Krugman is stating that;
“Thus, to some small but measurable degree, one must note, perceive, or identify what one will not attend to in order to reject it. I suggest that this “pre-attentive process” leaves some trace of the rejected material, certainly not enough to be recalled, but perhaps enough, with repetition of the process and repeated rejections, to be recognized later on.” (Krugman, 2000, p. 53)
From this quotation there are two points that can be made. First, not attending to or avoiding, through rejection, an advertising message will generate an effect. Secondly, to measure this effect the researcher ought to use an observation method that is sensitive enough to capture this effect, i.e. recognition. In Krugman’s argumentation on this issue he is also pointing at that it only takes a very short look to effectively convey the advertising message (Krugman, 2000). As a consequence, the unintentional attention will generate an advertising effect through brief or partial exposures (Krugman, 1986). Janiszewski is arguing in the same way by stating that pre-attentive mere exposures can encourage an individual to have more favorable attitudes even though she cannot recall the initial exposure (Janiszewski, 1993). According to Janiszewski the pre-attentive mere exposures belong to the category of incidental exposures. This is synonymous to the unintentional exposures that have been discussed. The pre-attentive processing of information can be seen as a parallel processing system that is parsing the environment into units that are meaningful, selecting adequate subgroups of units to receive additional attention (Janiszewski, 1993). Wolfe (1998) is describing preattentive information as something that is there to be used and in particular, that pre-attentive processing can be used for the subsequent deployment of attention. The pre-attentive information could therefore be considered as a basic processing of the environment which takes place prior to attending to a particular area. Transferring this discussion into an advertising context would mean that the entire setting is pre-attentively processed at a basic level. This pre-attentive
62
processing includes advertising present in the setting. The individual thereafter selects an area where to deploy the attention, for example the headline of an interesting article in the program content. When brand names and product offers (in the shape of advertising) are included in the secondary information in an environment, they are processed in a pre-attentive manner (Janiszewski, 1993). Extending Treisman’s, Kahneman’s, Mack and Rock’s, Krugman’s and others’ notions of attention to an advertising context gives a richer theoretical picture of the scene presented in the starting point model in ?gure 2.3. The context or media where an advertisement is presented can be seen as a task environment where the primary task is the program content transmitted to the receiver. The receiver is most often interested in the program content and is intentionally directing her attention to this material. There will also be other events, – tasks –, calling for the attention of a subject. Advertising can clearly be seen as one of these tasks that the subject can either attend to or inhibit. Attending to stimuli means more processing of the information whereas inhibition means limited processing of the same stimuli. The individual can also attend to the advertisement/s, which is becoming increasingly less common. Yet, even when the individual is not explicitly attending to an advertisement, brief, partial or mere exposures can have an impact. These unintentional exposures are also important and are making up the bulk of advertising exposure as Krugman states (1986).
Advertising, Attention and Media
A number of media, for instance TV and Radio, that are routinely being used for information transfer of advertising messages are most often being used in such a manner where the program content is interrupted (a break for advertising) and then the advertising messages are being sent. In this case the message stream is serial in nature18. In other media such as newspapers, magazines or the World Wide Web, advertising messages are being presented in connection to the program content. To the viewer or reader, the program content present is calling for the attention of the subject. At the same time, the advertising material is competing with the program content for the attention of the subject. In the latter case, but not in the former, stimuli are presented in a parallel way. This constitutes a setting very much similar to the one in ?gure 2.3. By describing the setting in this way a task environment has been de?ned from an attentional point of view comprising of the media, its program content and the advertisement, see ?gure 2.14.
18
There are however also situations where there are advertisements or logos present within the program content, for instance when it comes to football games, ice hockey or the like. 63
The Web Task Environment
Center of Attention Media vehicle
Figure 2.14 Secondary task
Site Name, Headlines etc Headline Text fields etc Headline Text fields etc Advertisement
Figure 2.14 Viewing the media as a task environment with a primary and a secondary task.
Part of the complete task environment and also valuable to take into consideration is the user activity mode or task mode of the subject being exposed to the stimuli. As mentioned before, attention can be either externally or internally driven. The task environment described in ?gure 2.14 comprises mostly19 of the external task environment. However, a second ingredient, the internally driven aspects have to be considered as well. A user’s activity mode can be set by the mood, time of the day, etc (see for instance Belk 1975) or an intention (Milliken & Tipper, 1998). An intention to solve a given task in a particular way is thereby also part of the task environment and sets the subject in a certain mode when interacting with the media vehicle. When visiting a Web site (a task environment) a person may be in different modes (just as one reader is reading ‘The Times’ to educate herself, another one is reading it to kill time and a third is searching for a speci?c information to quote) and different modes in turn affects the perception of presented stimuli in different ways. The mode a certain Web surfer is in, is dependent on what intention the user has. The user mode can also be altered or triggered by the functions or tasks that the subject encounters on the Web site or as a consequence of the way information is presented. A less goal directed mode might be to just surf around without any speci?c goal, perhaps solely for fun and pleasure, allowing the subject to wander from one place to the other in the media. It is plausible to assume that a user activity mode of this latter kind would make it more likely that attention, to a greater extent also will be directed to the secondary task. In other words; a user activity mode that is highly focused on the primary task, may direct less attention to the secondary task. A user activity mode, which does not require such a focus on the primary task, may open up for more attention to the secondary task.
19
An external environment can however evoke a certain, internally driven, mode.
64
Levels of Involvement
Media, context and ad
Figure 2.15 Media, context and ad
A person in a dual task environment under high involvement
A person in a dual task environment under low involvement
Figure 2.15 In?uence on advertising effect as a consequence of different user activity modes. The user is engaged in activities connected to the web page.
Janiszewski (1998) has studied differences in goal-directed search tasks and exploratory search tasks. He notes that goal-directed search is a process involving selective attention and that the “demands of the search task, organization of the information in the search environment, and the taskrelevant experiences of the searcher determine the ease/dif?culty of the search activity” (Janiszewski, 1998, p. 291). In addition, he is stating that exploratory search routines are used to “monitor the environment when a person is not actively searching for a piece of information” (Janiszewski, 1998, p. 291). Moreover, it can be a screening process on the look out for targets that can be subjected to goal directed search. In this screening process, the exploratory search have procedures to evaluate to what extent attention will be directed to a certain location in the ?eld of view before going to another location. Janiszewski’s view has characteristics that are similar to the view that Web users are either activated in a mode where they are actively looking or searching for some piece of information or just screening, sur?ng, the Web environment to see whether there is something appealing that they can direct their attention to. Sheehan (2002) has in his typology of Internet users, described users in the same way as here. Sheehan argues that there are Web users that have “a speci?c goal in mind; for example, an online session spent searching for speci?c information…” (Sheehan, 2002, p. 63) and also that there are users that “use the Internet for diversion, escape, and/or relaxation”(Sheehan, 2002, p. 63). The ?rst case quoted above coincides with the search concept whereas the second case quoted coincides with the surf concept. Sheehan is developing the typology and describes even more re?ned categories that will not be further developed in this study. Other researchers have identi?ed similar categories as those discussed above and common to them is that they have searching and sur?ng as broad categories in one way or the other (Korgaonkar and Wolin, 1999, Rodgers and Sheldon, 1999, Stafford and Stafford, 2001). Taken together, this reinforces the argumentation that searching and sur?ng indeed can be seen as broad and
65
separated behaviours exhibited by users when using the World Wide Web. Worth to note is that in general there is interplay between the external task environment and what activity mode the user will be set to. When visiting a search engine it is more probable that one is in a search mode. When visiting a Web site for general entertainment, on the other hand, it is plausible to assume that the visitors to a greater extent are sur?ng around for amusement, enjoyment, entertainment or the like. Multi-mode environments are naturally also possible and common, as well as environments used for other purposes.
Measuring Advertising - Click-through and Pricing
Before ending the chapter we will study how Web advertising is measured and how it is being priced. The various advertising means, mentioned earlier, have different purposes as well as effects. The old Web banners, for instance, that recently were much appreciated among Web advertisers had the purpose to generate click-through and bring the user to a target Web site. These kind of Web banners have witnessed click-through rates drop the last few years reaching frequencies below 0.1% (Khermouch and Lowry 2001). Despite this, click-through rates have been and still are widely used as the main measurement of the effect of Web advertising. While the focus on click-through banner advertisements may not disappear, it is likely that advertisements, including animation and video, will be used to a greater extent in a near future to come (Weaver, 2000). This is in fact already taking place. In one of the pilot studies conducted, ?ndings suggest that static banner advertisements are being replaced with more vivid banners on Web sites. Furthermore, since fewer users are actually clicking on the banners, the promotional content has to be presented straight on the host page as an advertisement and not on the page where the hyperlink leads to. Consequently, the decreasing click through rates and other developments,20 have driven the industry to develop advertisements that are more similar to what is being used in print media. There is also empirical evidence suggesting that click-through measures are likely to undervalue the Web as an advertising medium (Briggs and Hollis, 1997). Furthermore, the Interactive Advertising Bureau (Interactive Advertising Bureau, 1997) found that the use of simple involuntary exposure to a banner advertisement without click-through generates an increase in advertisement awareness. There are therefore reasons to believe that there is indeed an effect of advertising beyond the click. At the same time, clickthrough frequency has been, and still is, widely used as a method for pricing in Web advertising. The click-through measure has been a starting point for the development of the pay per click method (PPC). Except for the pay per click method there are also other pricing methods such as pay per view (PPV) or pay per time
20
And also due to technological development such as greater bandwidth, new methods of delivering digital content and the like. 66
unit (PPTU). Pay per time unit is in essence an agreement between the Web publisher21 and the advertiser to use a certain space during a certain time. This might for instance connote that the advertiser buys the upper right corner of a Web site for a period of one month or any other time period. Pay per time unit can also be a package of different locations on the Web site during a prede?ned time period, where some of the locations are prominent and others being less salient. Pay per view or sometimes called “CPM” (Cost per thousand, Hoffman & Novak, 2000) is a pricing method based on impressions or exposures (Mangàni, 2003). In the same way as with conventional broadcast and print advertising the pay per view method is taking into account “[…] the amount of advertising delivered” (Hoffman & Novak, p. 180, 2000). The “amount of advertising delivered” is sometimes used for broadcast as well meaning that the advertiser buys airtime at a certain time when a certain number of individuals are supposed to be reached. The “delivered amount” is however a more accurate measure in the Web case compared to broadcast. An example can reveal the difference. If an advertiser buys airtime the publisher will ask a certain price that depends on for instance the duration of the commercial, the time of the day and whether the spot is to be aired in connection with a certain program or the like. After taking these aspects into consideration the publisher will calculate and present a ?nal price.22 This price will on the other hand not (usually) take into consideration whether it has only been 13.8 million individuals that watched the show and the subsequent commercial instead of the expected 15.8 millions that normally watches the show. The pay per view pricing method on the other hand is taking each and every visitor into account since the pay per view measure, like the pay per click, is with a high degree of precision counting every exposure. There are not only “pure” pricing methods like those described here but also hybrids that combine different aspects to suit the needs of clients, publishers or both. Using the pay per click method as a means for pricing does not appear to be the most rational choice when the click-through frequency measure has an additional communication effect as measured by, for instance, recognition or recall, that is not accounted for by the pay per click method. This has the inherent consequence that those buying advertising space on Websites are receiving a service at constant discount, a discount that we do not know the magnitude of. From a Web publishers’ perspective this is a state that ought to be recti?ed. It is, however, perchance explicable that there may be a widespread malpractice in pricing matters considering that research in the area has not been able to keep even pace with the development. Hence, practitioners have been left to ?nding their own way.
21 22
From the perspective of those selling advertising space or the like.
There may also be other factors taken into consideration such as ?rst time discount, bulk discount, repetition and the like. 67
In this framework it is worth to mention that for example the search engine industry is using the pay per click pricing method to a larger degree. The industry is also increasingly successful in their endeavor to generate revenue. In the case of the search engine industry the pay per click method is tied to the number of clicks on keywords or Adwords that are presented as search results. In connection to the issue on what pricing methods are relevant to use, it is appropriate to also include how advertising is being measured and how it has been measured historically. Click-through has never before (before the World Wide Web came in to being) been a means to measure advertising effect. To measure clicks was a possibility that came along with the inherent properties of the information technology used for the World Wide Web. The click measure was ?rst and foremost a means to register activity on a Web site. When using links connecting a banner to a target Web site it became possible to measure how many users that clicked on the banner and were redirected to the target site. Soon it became a widespread practice to count the number of clicks as a measure of advertising effect. However, the perhaps most prominent researcher in the area of how to measure advertising is advocating other means than the click. Krugman has elaborately studied the use of recognition and recall as measures of advertising effectiveness. He is suggesting that recall and recognition are actually measuring two different things. Recall, which is when an individual is being asked if she can recall seeing an advertisement in a certain media and tell what brand name it was, is part of measuring verbal memory. Memory for words is though relatively low with a high level of forgetting (Krugman, 1985). Recognition on the other hand is measuring pictorial memory. When measuring recognition, an individual who has been exposed to an advertisement (the target ad) is presented with a number of irrelevant or phony advertisements of which one is the target ad. The main purpose is to see if the subject recognizes the target advertisement. This, is an “easier task” as opposed to when testing recall since right-hemisphere memory for pictures and images is exceptionally high with very little forgetting (Krugman, 1986, Krugman 1985). Even though both recall and recognition can be used for measuring advertising, Krugman is most of all arguing in favor of recognition as the main measure. Krugman is stating:
68
“Advertising researchers have argued about the relative merits of using recognition or recall as measures of advertising effect. Because the criterion of recognition is much more easily achieved than that of recall, it has been criticized as being less sensitive. Underlying this “technical” controversy, however, is the fact that the use of recognition justi?es modest advertising expenditures. I would reposition the recognition versus recall problem with this proposed addition to the theory of involvement – i.e., the nature of effective impact of communication or advertising on low-involvement topics, objects, or products consists of the building or strengthening of picture-image memory potential. Such potential is properly measured by recognition, not by recall. The use of recall obscures or hides already existing impact.” (Krugman, 2000, p. 53)
When using the click-through to measure advertising effect neither recall nor recognition is measured; only those who have clicked are accounted for. This may severely underrate the effect of an advertisement. Consequently, when a publisher uses click-through as a basis for pricing, the entire advertising message delivery will not be taken into account. A more sensitive measure ought to be used, i.e. recognition. It is therefore the pay per view method that is argued for at the expense of pay per click method. The pay per view method is closer related to actual exposures and to measure recognition is an appropriate way to measure exposures. Since the industry is frequently using the click-through measure, it is pressing to make known that there may be more adequate means when advertising effect is to be measured. It is always good to remember that “The practice of evaluating Web advertising on the basis of click-through is like evaluating TVads for automobiles on the basis of how many people visit the showroom the next day” (Briggs and Hollis, 1997, p. 33).
This chapter
This chapter set off by identifying the origin of advertising with the broader concept of marketing communication and marketing in general. Afterwards, it was elaborated around the essence of advertising and how advertising has developed into the World Wide Web. A number of issues that has bearing on the research questions that were formulated in the introduction chapter have been covered, placing attention to advertising on the World Wide Web into a theoretical context. Theories from disciplines other than advertising have been brought in to the theoretical framework to complement and develop the knowledge at hand. In the next chapter the different parts of the framework will be integrated in order to formulate relevant hypotheses taking as a starting point the research questions.
69
Chapter Three
Formulating Hypotheses
With the research problems as a starting point and the theoretical chapter as a guiding tool, parts of the theory that are relevant will be integrated and used for the formulation of the hypotheses for this study.
70
First section - Attention effects of context
In the theoretical chapter it has been elaborated on how the setting or the context in which the advertising resides can in?uence the impact of advertising. It is important to study this contextual dimension to understand how it affects advertising and whether it is powerful or can be neglected. Extending Kahneman, Janiszewski, Milliken, Tipper and Treisman’s theories of attention to a advertising context may provide insight to how attention is affected by context and activity mode. In the theory section, the Web environment was described as a task environment where the primary task is the program content transmitted to the receiver. The secondary task in this context is then the advertisement. Furthermore, a secondary task will be affected by the demand structure of the primary task in such a way that a more demanding primary task will reduce performance on the secondary task and all other events calling for attention. Therefore; H1: A situation where respondents are exhibiting a less goal oriented sur?ng behavior will elicit more attention to an advertisement than what a more goal oriented sur?ng behavior will. To be rejected… [H0H1 A situation where respondents are exhibiting a less goal oriented sur?ng behavior will elicit equal or less attention to an advertisement than what a more goal oriented sur?ng behavior will.] If H1 is correct, we can conclude that individuals set to search mode, pay less attention to an advertisement than what a respondent in a surf mode does. Another contextual dimension that has been elaborated on is the physical appearance of the environment where an advertisement is presented. This physical environment that was described as a task environment may also in?uence advertising. One aspect of this task environment is its inherent complexity which can be manifested in a number of ways. One such way is the complexity of the search system on a Web site and the search depth that it has. Given what has been learned in the theory it is plausible to state that; as complexity increases, the demand on the primary task will increase and thereby reducing the attention to the secondary task. Hence: H2: A Web environment that is less complex, with regard to its search depth, will elicit more attention to an advertisement than what a more complex Web environment does. To be rejected… [H0H2: A Web environment that is less complex, with regard to its search depth, will elicit equal or less attention to an advertisement than what a more complex Web environment does]
71
If hypothesis H1 and H2 will be supported then there is reason to consider an alteration of the advertising models displayed in the theoretical chapter. If H1 and H2 are correct then the models studied can be elaborated. More speci?cally, the attention construct should be described more in detail taking into account what is happening between the advertisement, context, attention getting etc and the attention construct. What is lacking in the models is a more elaborate description of the attention construct. If Thorson’s models are taken as an example to compare with, it can be said that what is lacking is an environment or setting construct that takes into account the environment’s effect on an individual’s attention. There is also a lack of a construct that can impede the incoming stimuli. These issues that have been pointed at in the previous paragraph will be further addressed under the section “Modeling attention to advertising”.
Second section - Advertisements and their respective attention effect
In the previous chapter it was argued around different attention getting techniques and that one and the same advertisement using different attention getting may receive different amounts of attention from individuals. Chandon’s study indicated that animation gave a greater response measured as click-through frequency compared to non-animated advertisements. He also recognized that many advertisers had stopped using non-animated advertisements. Lohtia, Donthu and Hershberger received results that pointed in the same direction and they strongly recommended the use of animation as a means to attract attention to the advertisement. Sundar and Kalyanaraman’s ?ndings were also in favor of animation. Common for the studies above is that none of them used recognition or even recall when measuring results, instead they used click-through frequency as a measure. In the attention section a number of features were shown to have the ability to attract attention. Mack and Rock had studied how color, numerosity, motion, ?icker, shape, location and texture segregation were performing when attracting attention. After a great number of experiments Mack and Rock concluded that properties such as meaningfulness and size indeed could attract attention. Yantis and Yantis & Jonides were arguing that in particular new objects presented in the visual ?eld had the capability to attract attention. These studies referred to here were studying the phenomena of attracting attention on a very basic level where these basic features were studied. In advertising these basic features are put together on a greater scale and the exposure times are much longer. It is therefore important to understand how attention getting works on a “macro” level. An animated advertisement is actually using some of the basic features that our experimental psychologists have studied in attention experiments. An animated advertisement will by nature have the possibility to present new objects in the visual ?eld since the animation itself is not stationary. It also has motion or ?icker since animation is adding motion to an object that otherwise would be static.
72
Taken together it seems plausible that animation will evoke greater attention than what a non-animated advertisement will. Hence: H3: An animated advertisement will elicit more attention than a static advertisement. To be rejected… [H0H3: An animated advertisement will elicit equal or less attention than a static advertisement] As was stated already in the theoretical framework, a pop-up advertisement is merely a special case of an animated advertisement from a visual perspective. When the pop-up advertisement appears in a certain location it creates “motion” where it appears. This could be seen as an “abrupt onset” as Yantis described it theoretically, since the pop-up appears in an empty area (sometimes also in other areas than empty areas). Cropper & Evans and Smith & Goodwin found in their study support for animation as a means to attract attention from one area of the screen to another area because of the “visual distinctiveness”. Relevant in the case of pop-up advertisements, since it can be perceived as a special case of an animated advertisement, is of course all the arguments presented in favour of animation in hypothesis 3. With this said: H4: A pop-up (traditional pop-up, over-the-page) advertisement will elicit more attention than a static advertisement. To be rejected… [H0H4: A pop-up (traditional over-the-page pop-up) advertisement will elicit equal or less attention than a static advertisement.] It is also expected that that a pop-up advertisement will elicit more attention than an advertisement that is just animated. The reason for this is that the popup advertisement is not only animated but also that it appears in an empty area creating extra “motion” in the area where it appears compared to its animated counterpart that is just animated and staying in the same area the whole time. It is expected to be easier to inhibit the animated advertisement than what it is inhibiting a pop-up advertisement that is unexpectedly popping-up in the viewers’ visual ?eld. Therefore: H5: A pop-up advertisement will elicit more attention than an animated advertisement. To be rejected… [H0H5: A pop-up advertisement will elicit equal or less attention than an animated advertisement]
73
The Second Section Extended
During pre-analysis of hypothesis 5 that was conducted during the data collection, a few interesting ideas surfaced that ultimately lead to extend the study somewhat. This part was mostly driven by a few theorems that originated from Yantis, Jonides, Mack and Rock. When analyzing hypothesis 5 it seemed obvious that the pop-up advertisement would be more effective than the animated advertisement, and so it was. And the result would clearly follow Yantis and Jonides idea that a new element in the visual ?eld would attract more attention. However, before jumping to premature conclusions there was though, one additional aspect that could be taken into consideration, namely the frame. The frame of a pop-up advertisement is an additional element to consider which was not considered in isolation initially, since previous studies have not taken this into account. The frame on the pop-up advertisements that were used in the treatments added 50 pixels on the height and 25 pixels on the breadth. This means that the pop-up advertisement in fact comes with a frame that makes it 21 075 pixels or 29.2% larger than the animated advertisement. At the same time we have to remember that it is not the advertisement itself that is larger, they are of the same size, however the difference is just in the frame. The difference in effect that was observed in both the pre-analysis and the ?nal analysis of hypothesis 5 may (or may not) originate from the frame around the advertisement in the pop-up and not because of the “abrupt presentation” of the advertisement. It could of course be that there is an effect from both of these factors – the abrupt presentation and the frame. These ideas were considered of such an importance that a 6th hypothesis was formulated based on the reasoning in the two paragraphs above. In order to study the effects of the frame separately, a new set of experiments had to be designed and conducted. This hypothesis demanded the study to be extended with 292 more respondents and the hypothesis is an extension complementing the original hypotheses. It complements the 5th hypothesis in such a way that we may obtain an answer to the “frame effect versus the abrupt presentation effect controversy”. What we will in fact do here is to put in an advertisement that is animated, but “popping up” only as an in-page advertisement and not as an over-the-page advertisement like traditional pop-ups. By doing, so we will have a pop-up advertisement that does not have a frame. Thereby we can compare the effect between a pop-up with a frame with a pop-up without a frame. By using this procedure we will perhaps be able to resolve the controversy. This can be formulated into hypothesis number 6a) and 6b). H6a: A traditional pop-up advertisement will elicit more attention to the advertisement than a FL1 pop-up (in-page) advertisement.
74
To be rejected… [H0H6a: A traditional pop-up advertisement will elicit equal or less attention to than a FL pop-up advertisement.] And furthermore: H6b: A static advertisement will elicit more attention to the advertisement than a FL pop-up (in-page) advertisement. To be rejected… [H0H6b: A static advertisement will elicit equal or less attention to the advertisement than a FL pop-up advertisement.] A result that would be in line with hypothesis 6a) and 6b) can give some information about the “abrupt presentation effect” and the “frame effect” so that we perhaps can resolve the controversy. However, there will still be obstacles in our way if we want to attribute effects to the right construct. The point is that not even an answer to the “abrupt presentation – frame” construct pair will be suf?cient for a comprehensive understanding. The problem is the dual property of the construct under examination here. Let us elaborate further in order to decipher what this means. First, the abrupt presentation effect has to do with the popping up of the advertisement and will be measured in the hypotheses above. The frame effect has to do with the frame of the pop-up advertisement and can be divided into subordinate constructs that is building up the frame. One can of course measure and analyze the entire frame effect and get an answer of that effect and this will be measured in the hypotheses above. However, one can also analyze the subordinate constructs. From a stimulus point of view an advertisement that is popping up with a frame has at least two properties that are relevant here. We can talk about the frame as having a size component and a “frame-meaning” component. The inherent properties of a pop-up advertisement (when it comes with a frame) will inevitably increase the size of the advertisement. Just the fact that this popup package is larger in size may have an effect when measuring the response from this stimulus. This has to be considered. Furthermore, there may be a second effect originating from the frame that does not have to do with the size of it. This second “frame effect” might be connected to meaningfulness or formulated differently; the frame may represent a distinct meaning compared to an advertisement without a frame. The rationale for this is that a frame that is popping up represents an intrusion in the visual ?eld of
1
The special pop-up advertisement will hereafter be called FL (FrameLess) to make it easier to separate it from the traditional pop-up advertisement. 75
a Web user and this intrusion may have a certain meaning to the user and that could then be the reason why they attend to the advertisement. The meaning that it represents is most likely not a positive one, but a negative. So instead of applying Yantis idea of new objects in the visual ?eld to the described situation it might be even more fruitful to lean towards Mack and Rock’s notion of meaning, or perhaps both. Following Mack and Rock’s notion of meaning and late selection then a popup advertisement without a frame should carry “less meaning” from a user’s perspective than what a pop-up advertisement with a frame carries. This would clearly mean that the frameless pop-up advertisement will elicit less attention to it than the pop-up advertisement with a frame. Nonetheless, as we may recall, Mack and Rock were discussing around what actually attracts attention and they were pointing at “meaning” as one factor. They also discussed around a late selection theory to explain why subjects for instance saw their names when spelled “Jack” but not when spelled “Jeck”. Translating this to the advertising situation; can it be that they see the advertisement when they see the frame since it is a carrier of negative “meaning”? Without carrying on any further we can already here say that we will obtain the effect of both size and frame-meaning at the same time2 when we are measuring the frame and then we can not with certainty attribute the effect to the right construct. Though, the fact that we will not obtain an answer to the frame-meaning issue is however not a reason to not pursue this endeavor, on the contrary. As we have discussed earlier we may reach a greater understanding of the boundaries of the pop-up effect and to resolve the “frame effect”3 versus the “abrupt presentation effect” issue. The assumption of hypothesis 6b is that if 6b is correct then the abrupt presentation for the FL pop-up advertisement plays a minor role when comparing it to the static advertisement. In this experiment the two tested advertisements have indeed equal size. The very reason why the static advertisement is expected to elicit more attention is because it has longer exposure duration. If the abrupt presentation, on the other hand, is strong this should more than well compensate for the shorter exposure time for the FL pop-up advertisement and the attention to the FL pop-up advertisement should be greater than that for the static advertisement. Finally, we should be open for the fact that there may be other not yet known effects that may have an in?uence on the effects that we are
2
What is clearly needed to potentially resolve this issue to the fullest and to be able to attribute the effect to the right ”location” is a pop-up advertisement with a frame and where the size of the frame should not exceed the size of the advertisement X. Then the size effect would at least be equal but the advertisements would be unequal. A higher effect for the pop-up advertisement with the frame would then indicate that it is indeed the frame and not the size that is behind the increased effect. This could however not be tested in this study due to fund and time constraints. 3 With the sub constructs not separated. 76
observing. Nevertheless, here is an opportunity to reach a bit further in our understanding of this phenomenon, especially since no research has been done in this particular area.
Modelling attention to Web advertising
Vakratsas’ and Amblers’ model (see ?gure 2.4) is the most similar to the model and ideas to be presented here. However, there are some differences that are related to the in?uence of the environment/context dimension and also the attention getting techniques. Furthermore, from the theoretical description of attention in the previous we know that the attention system is an intrinsic part of cognition and the point here is to open up the cognition part of Vakratsas’ and Amblers’ box somewhat. Drawing upon Treisman’s attenuation theory and Vakratsas’ schematic model it can be visualized what is taking place when taking into account the studied factors’ (the ones in hypothesis H1 and H2 as well as H3 - H6b) interplay with attention. Based on this the model is to be built and tested. Starting at the left (in ?gure 3.1) there is a contextual dimension, the Web environment/mode (hypotheses H1-H2), in which the advertisement is placed in. Given that it is a question of a Web advertisement the advertising message has a potential “propellant”, the extra attention getting technique, which can enhance attention (hypotheses H3-H6b). The environment/mode and the attention getting technique together with the advertising message constitute the input. As we have seen in the theoretical chapter there are a number of terms that are being used for the construct that Broadbent initially called a “?lter”. Treisman has used the term ?lter too in her work but also pointed out that her “?lter” is an attenuator which is attenuating the input. She has also used the term “barrier”. Still other researchers have used the term bottle neck and of course also the limited pool that can be seen as a barrier. No matter what terminology that is being used one could say that there is a ?ltering mechanism, borrowed from Pashler, which is located between the input and attention. The ?ltering mechanism is placed as a barrier in the center of the model. This construct is actually a part of the attention system so the separation is merely to visualize how the system works. The ?ltering mechanism can allow advertising messages to pass through to a greater or lesser extent depending upon how the contextual dimensions affect the mechanism. On the right there is unintentional and intentional attention leading to memory. The two different kinds of attention are covering both the situation when an advertisement has been attended to in a peripheral unintentional manner and when it has been more intentionally attended to. If isolating and focusing closely on the critical constructs that have been discussed, it could take the following form;
77
The Proposed Model
Filtering Mechanism Unintentional Attention Intentional Attetion Unintentional Attention Affecting the permeability of the filtering mechanism
Figure 3.1
Contextual dimension influencing
Attention Getting The advertising Technique message Contextual dimension influencing
Memory
Figure 3.1 A model of attention in Web advertising.
A straightforward description may summarize the relationships between the various parts of the model. The advertising message, on the left in the ?gure above, is the source of stimuli input and may be propelled with an attention getting technique. The advertisement is placed into a context that may impact the advertisement’s ability to penetrate the ?ltering mechanism in order to be attended to, intentionally or unintentionally. The advertising message will at the end be delivered to memory. From the author’s point of view the proposed model in ?gure 3.1 is a plausible translation of relations between concepts, given that the hypotheses are supported. That is, that support is generated for hypotheses H1, H2 and the general notion4 expressed in H3-H6b.
Third Section – measurement and pricing
This third section of the hypotheses is targeting how advertising on the Web is being measured and can be tested directly with hypotheses. The second area is not being tested directly with hypotheses but the results from the ?rst area will have implications for the second area since they are intrinsically tied together. In one of the pilot studies that were conducted, ?ndings suggested that banner advertisements, with the sole purpose to redirect users to their own Web sites, have to quite some extent been replaced by advertisements containing a complete message on the advertisement itself. This result is supported by other studies conducted, which has been reported in the theoretical chapter. Furthermore, since fewer users are in fact clicking on the banners, the promotional content has to be presented straight on the host page as an advertisement and not on the page where the hyperlink leads to. In recent years this is actually what has happened and recent studies are indicating that Web advertisements are increasingly designed for this purpose as well.
4
That there are attention getting techniques in Web advertising, as described, that enhances attention to an advertising message 78
However, despite this development, advertising effect is still being measured as click-through frequency. Krugman, one of the prominent researchers of advertising effects, have advocated the use of recall and most of all recognition in order to better account for the effects of advertising. This is in contrast to a widespread practice in the Web advertising industry. In order to investigate whether it is a correct practice or not to use click-through as a measure of advertising effect, the following hypothesis is formulated. H7: A Web advertisement will generate more respondents being aware of the advertisement as measured by the recognition measure than it will generate clicks on it. To be rejected… [H0H7: A Web advertisement will generate equal or less respondents being aware of the advertisement as measured by the recognition measure than it will generate clicks on it.] If H7 is correct it would imply that there is an advertising effect taking place that is not accounted for if only the numbers of clicks are measured. Adequate measures ought to be used in order to really know the effect of advertising on the Web. This would be in line with for instance Krugman’s view on how to measure advertising effect and contrary to a widespread practice among contemporary practitioners only measuring click through frequencies. It would also have implications on models for pricing in the area of Web advertising. This would indicate that the basis for the pay per click- pricing method (PPC) is inadequate and should be altered or dropped in favor of pay per view (PPV) models.
Other effects regarding Web advertising
In the theoretical chapter it was pointed out that it is not entirely positive when advertisers are over ambitious in their striving to capture attention. In some cases there are of course individuals that gladly will look and spend time and attention on viewing an advertisement, perhaps even clicking on it. This is though not always the case. Research by for instance McGalliard, Nielsen, and Spool et al. show that some subjects do not appreciate advertising to such a great extent and that attention getting techniques can be perceived as “annoying”, “irritating” and even “evil”. Based on this it is reasonable to expect that the more attention getting techniques employed the more annoyance evoked. In this experiment three different attention getting techniques will be tested against each other. H8: The risk for an advertisement to be perceived as annoying will be greater when more attention getting techniques are employed in the advertisement. This would mean that a static advertisement will be the least annoying, an animated advertisement will evoke intermediate annoyance and a pop-up advertisement will be the most annoying.
79
To be rejected… [H0H8:All other alternatives.] A rejection of H0 will tell that there is a difference between the various treatments. This difference would indicate that attention getting comes at a “price”, the price of the advertisement being perceived as annoying by the audience. With further analysis, it may also tell what kind of attention getting that is perceived as more or less annoying by the respondents.
Summary
The hypotheses in this chapter are summarized in the following: H1: A situation where respondents are exhibiting a less goal oriented sur?ng behavior will elicit more attention to an advertisement than what a more goal oriented sur?ng behavior will. H2: A Web environment that is less complex, with regard to its search depth, will elicit more attention to an advertisement than what a more complex Web environment does. H3: An animated advertisement will elicit more attention than a static advertisement. H4: A pop-up (traditional pop-up, over-the-page) advertisement will elicit more attention than a static advertisement. H5: A pop-up advertisement will elicit more attention than an animated advertisement. H6a: A traditional pop-up advertisement will elicit more attention to the advertisement than a FL pop-up (in-page) advertisement. H6b: A static advertisement will elicit more attention to the advertisement than a FL pop-up (in-page) advertisement. H7: A Web advertisement will generate more respondents being aware of the advertisement as measured by the recognition measure than it will generate clicks on it. H8: The risk for an advertisement to be perceived as annoying will be greater when more attention getting techniques are employed in the advertisement. This would mean that a static advertisement will be the least annoying, an animated advertisement will evoke intermediate annoyance and a pop-up advertisement will be the most annoying.
80
Chapter Four
Research Method
In order to answer the research problems an adequate method is needed that can develop the information required. In the following chapter the research method is outlined and elaborated in detail, showing the path of the studies conducted.
81
Approach
This study is in its essence a study of social activities. Social activities can be studied in a variety of ways employing different methods to achieve understanding and to capture and create knowledge or knowledge statements. Independently of the research strategy chosen, it can be seen as a map or a tool for the researcher on her way through the research process to ful?ll the purpose of the study (Bengtsson, 1994, p. 41). The researcher is though not a blank sheet when initiating the research. Instead, she will have all prior knowledge and views as a luggage that may tilt the map in some way. While some researchers are more prone to view the world in a more objective way prioritizing certain kinds of knowledge, others on the other hand may be more interested in other more subjective aspects of the social world. My view is that there is a whole spectrum of possible views of the world and that they all can contribute with different kinds of understanding. Czarniawska (1999 p.9) states that different views contribute in their own respective ways, and it is reassuring that, many versions of the world are possible. Johansson-Lindfors (1993 p.41), when reasoning around Morgan’s views, is stating that objective as well as subjective views on reality can be relevant when studying different research problems in social science. Bonnedahl (1999, p.39) is arguing in line with this stating that knowledge about reality to a greater or lesser extent can be seen as either objective or subjective. Following Bonnedahl’s eloquent phrasing, this study is not aiming at providing an objective truth but rather descriptions that can be inter-subjectively understood. The research problem and the purpose of the research is also of importance in this context since they, to quite some extent, will be guiding the research process and also reveal what kind of knowledge the researcher is aiming at. The research problem is also to at least some extent narrowing down the potential approaches and methods that can be used (Yin, 1994 p.4). However, a given research problem can be approached by using either a qualitative or a quantitative approach, or a combination of the two. Furthermore, one might use a survey, case studies, interviews, experiments or some other strategy in order to answer the research problem and to extract information. These strategies have various advantages as well as disadvantages that would make a certain strategy very suitable in one case while perhaps completely inappropriate in another setting. Whether to choose one strategy, at the expense of another strategy, is a delicate decision that the researcher has to make. For instance, the qualitative approach is perceived to provide rich descriptions of the social world (Denzin & Lincoln, 2002 p. 10). According to Denzin (2002, p. 3) the qualitative research is a “situated activity that locates the observer in the world” and furthermore that “qualitative researchers study things in their natural settings”. The quantitative approach, on the other hand, makes it manageable to study a larger selection of individuals and thereby provides, to a larger degree, knowledge statements that can be generalized to a larger population.
82
The research strategy chosen for this study has been based upon a few considerations that were taken into account. First, a quantitative approach is appropriate as one of the aims is to be able to identify differences between treatments in a selection and with the ambition to generalize to a greater population. A qualitative approach is not quite suitable for that task, a notion supported by Johansson Lindfors (1993, p. 61-65). Furthermore, theory suggested that some of the phenomena to be studied are assumed not to occur frequently in a population. When an infrequently occurring phenomenon is to be studied and compared with other occurrences then a quantitative approach with a fairly large sample is preferable. Another reason is that, in one part of the study the objective is to ?nd (or not ?nd) variation and rates of occurrence between groups and within groups, which makes the qualitative approach less suitable. Consequently, the research strategy chosen is quantitative in nature and includes a survey to extract information from a large number of subjects. Considering that the primary aim of the study is to investigate differences in response between subgroups subjected to different treatments then some kind of test has to be conducted. In light of this ambition the experimental approach is a natural choice. In the words of Cook and Campbell;
“The word experiment denotes a test, as when one experiments with getting up two hours earlier to see if this makes one’s working day more productive.” (Cook and Campbell, 1979, p. 2)
This study is however not a matter of testing productivity but to test a number of speci?c advertising messages under some speci?c conditions. The purpose of the various experiments is to answer and highlight different aspects of the overarching research problem. A number of hypotheses have been formulated, originating from the research problem. These hypotheses are derived from theory and represent statements, which “speci?es how two or more measurable variables are related” (Churchill and Iacobucci, 2002, p. 93). Formulated differently one can say that the hypotheses are a possible answer to the research question (Aaker, Kumar and Day, 2001, p.49). Multiple experiments were designed and set up to arrive at a comprehensive understanding of the many facets of the research problem. How the experiments were conducted and how they were designed will be elaborately described in the following section.
The essence of an Experiment
Compared to other methods experiments demand rigorous planning in advance in order to avoid pitfalls and obstacles that may lead the researcher astray. An example will best explain the essence of an experiment. For instance, Ivan Pavlov, a Nobel Laureate in 1904, conducted his well known experiments in such a way that he rung a bell and subsequently gave food to his research subjects, the dogs, for a period of time. After a short “learning”
83
period he could just ring the bell and afterward measure an increase in the saliva produced in the dogs’ digestive glands. In that case Pavlov could with a degree of certainty say that the measured effect (increase of saliva) originated from the ring in the bell and no other occurring events. The reason for this was because Pavlov controlled the environment so that there were no other events that could explain the effects that were observed. Hence, Pavlov could rule out other potential “z-variables” (other variables that could have explained the event) as having any impact on the observed effect. Pavlov was “so concerned with isolating the cause of the dog’s salivation that he constructed a ‘Tower of Silence,’ with eight soundproof rooms separated by spiral staircases” (Becker, 2003, p. 88). Pavlov had thereby shown, without the interference of other variables, the cause (ringing bell) and effect (salivation) of conditioned re?exes and at the same time paved the way for a more objective way of studying behavior. In another rather well known experiment1 a horse was claimed to have the capacity to perform some simple calculations. When asked by his master, Wilhelm von Osten, what is the sum of 3 plus 2, the horse tapped his hoof ?ve times. This and other tests of the intelligence of the horse (which went by the name “Clever Hans”) were taken as evidence of an understanding of human language as well as simple calculus. Albeit many researchers looked into the case it took quite some time before it was found that it was in fact another variable that had an effect on Clever Hans’ counting skills. Oskar Phungst, a psychology researcher, found that Hans was able to respond to subtle physical cues (more or less equivalent to body language) so he stopped “counting” when he reached the predestined number, not because Clever Hans could count, but because he could read the body language of the person who was asking the question (Fernald, 1996). When Clever Hans reached the right number the body language of the one asking the question was slightly altered and then Hans stopped. This unexpected variable “z” gave an explanation to the whole phenomena and at the same time disproved that the horse had an unprecedented intelligence2. In short, what is preferred is the situation of Pavlov and to avoid the situation with Clever Hans. It is for that reason imperative in experimentation to control the environment in order to be able to reach non-confounded conclusions. This and many other problems are the focus for the researcher’s effort to ensure reliable and valid results. When taking the discussion above and formulating it in a more formal way, then an experiment is, in its essence, a test of what the outcomes are under
1 2
Actually the events started out as a show for the public but eventually led to tests conducted by researchers.
Even though Clever Hans was not intelligent in calculus, he was very skilled and intelligent in reading subtle physical cues. 84
certain criteria. It is distinctly stated “certain criteria” for the reason that the researcher often controls the environment and wants to be assured that it was actually and solely due to the treatment X that the change of Y occurred and not for any other reasons. The researcher has a strong wish to be able to exclude other explanations Z which could interfere with the conclusion that Y happened because of X and nothing else than X. The general idea with an experiment is to observe the variation in one variable as a consequence of the manipulation of another variable, while other variables are being kept stable. If this endeavor is accomplished then the researcher may reach a state where she can say, with some degree of certainty, that X caused Y. To conduct research in a natural setting is often what researchers aim for. It is though not always achievable. There are many reasons for this, for instance scarce resources such as time and money but also practical problems for instance how to study a phenomena and to have control over disturbing or distracting z-variables in a natural setting. Another way to conduct the study, which is often more manageable, is to conduct research in a laboratory setting where it is more uncomplicated to control variables. The strength with the laboratory rests on the possibilities it provides to keep the environment stable in the way the researcher wants it to be. It thereby becomes easier to be certain that the observed effect can be attributed to the right cause. A typical experiment is often staged so that the researcher is ?rst doing a pretest where the treatment group is being measured on the variable to be examined. Afterwards this group will receive a treatment followed by a posttest to see what effect the treatment had. Usually the researcher will have a control group as well where respondents are not receiving any treatment or receiving an irrelevant treatment, i.e. placebo. This will put the researcher in a position to compare the treatment group with the non-treatment group. In the posttest the researcher can notice whether there is a difference between the groups and if so, how big the difference is. When properly conducted, the researcher can then conclude that the effect observed in the posttest is caused by the treatment. In this study, this strategy described was, however, not used. Instead, a slightly different kind of experimental design was employed; a design which is very frequently used in advertising research as well as and in other areas. The design included multiple treatments since the overall idea was to do multiple comparisons. This experimental strategy is sometimes called “planned variations” (Cook and Campbell, 1979, p.349) or more commonly “factorial design”. Compared to the “traditional” experiment described in the above paragraph the multiple treatments experiment does not include a no-cause baseline “making it dif?cult to test hypotheses about “absolute” cause as opposed to hypotheses about differential impact” (Cook and Campbell). This is in line with the purpose of the study, which is to investigate differences between treatments, rather than to state the amount of the absolute level of a certain treatment in comparison with the placebo group. A non-treatment group would therefore not have contributed to the interpretation of the
85
results. Albeit the lacking of an explicit or separate control group, it should be kept in mind that each of the treatments “acts as a control for the others” (Aaker, Kumar and Day, p. 345, 2001) making a traditional control group redundant. This procedure is adequate since, as noted earlier, the purpose of the experiments is to compare differences between multiple treatments. However, a non-treatment group could have provided some information about the level or how many of the respondents that falsely claimed to have seen an advertisement in a treatment when they actually did not see one. This problem was however overcome by using an alternative procedure. The respondents were not simply asked whether or not they had seen advertisement “X” but instead they had to point out which one out of ?ve different alternatives they had seen. Rothschild and Singh (1983) are stating that response bias may be eliminated by using multiple alternatives when the stimulus item in the recognition test is presented together with one or more distracter items. By doing so, the respondents could not falsely claim to have seen an advertisement they had not seen. If the respondents guessed which advertisement had been presented their responses would have been evenly distributed between the ?ve alternatives making it possible to calculate the effect of the guessing and to subtract this from the results. However, this never became a problem indicating that the ?ve alternatives was an adequate means for deterring the respondents from guessing. Another noticeable difference, compared to traditional experiments, is that there were no pretests in order to measure the level of the dependent variable or any other variable. Since the treatments were ?ctitious innovations then we know with certainty that all respondents in all treatments were equivalent to “zero”3 before they received a treatment. Consequently, any pretests would have been super?uous.
3
They were “zero” in that respect that no one had ever seen or attended to the treatments before. The attention to the advertisements in the various treatments was the variable to be measured. A pre-test would not provide additional information and would also be logically inconsistent to implement. 86
Overall design
The research design comprise of a pre study phase, which is followed by the main study. Due to the nature of the research problem and also because the area of Web advertising is a fairly new one, pilot studies were undertaken in order to get the required pre-knowledge to do the main study. The studies, which vary in their width and time span, were conducted in the beginning of the project. The studies can be described as here below:
General Description of the Research Design
Pilot study 1, How are Websites being used, Web advertising and Web categories
Figure 4.1
Pilot study 2, Test of Web experiments and on-line questionnaire etc.
Main study, Three experiments of advertising effects
Figure 4.1 General description of the research design.
The pilot studies were targeting different areas of the subject. The purpose of pilot study I was to obtain an overview of different kinds of Websites, advertising and advertising means on the Web. The intention was to identify frequently used Web functions and design elements. This provided basic data for an elementary Website categorization on how Websites are being used by Web users. Pilot study II was a test of the developed experiments, the various treatments and the on-line questionnaire. It was a live test on a group of 30 respondents with the purpose of revealing problems with questions, concepts, inconsistencies or any other ?aws with the research instrument. Findings from pilot study I was used to develop and to design the experiment and the questionnaire whereas pilot study II had the purpose to re?ne and ?ne tune the main study. The main study comprised of three separate experiments. The experiments were however conducted more or less simultaneously in order to save time. The experiments were followed by a survey to measure the effects of the different treatments. These both parts, the experiment and the survey, were integrated into one single HTML design that was presented sequentially to the respondents.
87
The three experiments in the main study can be portrayed as in ?gure 4.3.
The Research Design
Attention getting A site & A instruction Ad 1 Ad 2 Ad 3 Ad 4 Goal orientation Ad 3 + B-site
Figure 4.2
Task environment
Ad 3 + B-site
Figure 4.2 The research design with its treatments in three dimensions.
Figure 4.2 is visualizing and providing a schematic overview of the dimensions studied in the experiments. The horizontal dimension in the ?gure is targeting the attention getting effects between various advertisements, i.e. ad 1 through 4. Initially this experiment comprised of ad 1 through 3 but was extended with ad 4 (treatment 6), the “frameless ad”. This experiment used Web site “A” and instruction “A”. The vertical dimension in ?gure 4.2 is aiming at the impact of the setting or task environment in which the advertisement is presented. In this experiment the advertisement variable, Ad3, was kept stable and instead the Web environment was altered. In the last dimension, which is diagonal in the ?gure, goal orientation or mode was the main issue. In this experiment the advertisement variable, Ad3, was kept stable as well. At the same time the task environment was kept stable. The experiment was to test different levels of goal orientation manifested through two separate instructions given as treatments.
The design of the experiments
The three experiments in the main study were conducted in order to answer the research questions posed. The reason why experiments were employed was to investigate if any differences between advertisements and Website environments could be found. The expectation of differences is natural when doing experiments. Graziano and Raulin (1993) are stating that the expectation of variation is logical when using experiments – if there is no variation there are no differences to test. Four different Web advertisements and two complete Web sites were set up. The Websites were designed4 in line with how a typical Web site in that particular
4
The Websites were designed by the author together with Web designer Magnus Lindgren and two computer engineers who were solving some speci?c problems. 88
category is designed. The design of the two Websites was guided by information collected in pilot study I. In the same way, the advertisements were designed5 in line with the appearance of a typical Web advertisement. The design of the advertisements was guided by information collected in pilot study I. The experimental Web site was a complete news Web site with the same breadth and scope as an average national Swedish news Web site. The Web site had more than 130 individual Web pages that could be accessed by the respondents. Mullarkey and Danaher (2003) did a study that was similar in some respects as the study at hand. However, in their study the respondents could not surf as one normally does when visiting a Web site. Instead the respondents were restrained to merely watching a Web site that was changing pages automatically with different time intervals. In the experiments conducted here the respondents were able to actually interact and surf in any way they wanted. Thereby this experimental design is an improvement that makes the situation for the respondents more realistic and comparable to a real world situation. The experiment Web site design included one Web version (hereafter called the “A-site”) that could be seen as representative for Swedish news Web sites and a second Web site (hereafter called the “B-site”), actually a derivation of the ?rst one, which was less complex with respect to the search depth. Both Web sites had exactly the same content, news, headlines, name and visual appearance. The only difference was that the B-site had a shallower search depth and was, consequently, less complex. If we take a look at the outside world we can see that the A-site is similar, when it comes to the search depth structure, to for instance the Swedish Web sites “DN.se”, “HD.se” and “SVD.se”, to mention a few. The B-site is more similar to the search depth structure of “Aftonbladet.se” and “Expressen.se”. The search depth system employed in the experiments was thus not arbitrary but an imitation of what is commonly used in real. The Web sites were designed so that they would look similar to real Websites and this was achieved by using the same kind of links, subjects, concepts and general structure as can be expected in a traditional news Web site. The two Web sites had the same color theme, the same background color, the same frame colors and text fonts. Furthermore the same system for links and buttons were used. The Websites were speci?cally designed in such a way that there would be as small differences between them as possible except for the search system and search depth. The search depth was an operational variable to be tested and used as a measure of the complexity construct.
5
The advertisements were designed by the advertising agency 89
In the advertisement experiment, targeting attention getting effects, four different advertisements were used. The advertisements were exhibiting brands, products and offers that were entirely ?ctitious so that there would not be any problems with bias (for instance, that some respondents would recall or recognize the advertisement beforehand while other respondents would not recall or recognize the advertisement). All advertisements had in common that they had the same brand name, picture, headline and copy text. This general advertisement was then used in different ways. Here is a short description of the advertisement: the brand name was “Dentax”, there was an image of a female mouth covering the mid section of the advertisement, the headline stated “För en hälsosammare mun”6, the copy text stated “Nu ?nns den första tandkrämen som motverkar tandsten och samtidigt ger vitare tänder”7. The advertisements can be found in appendix A. The ?rst stimuli, Ad1, was a static version of the general advertisement placed to appear on the right upper part of the screen. The size of Ad1 was 283 times 250 pixels. The second stimuli, Ad2, was a dynamic animated banner advertisement based on the general advertisement. It was placed at the same location as Ad1. The size of Ad2 was 283 times 250 pixels. The difference between Ad1 and Ad2 comprised solely of the animated effect added to Ad2. The animated effect was accomplished by using a java script producing a “slideshow” where 5 images were presented for 3000 milliseconds each. The “slideshow” produced an animation that altered the mouth that was displayed on the advertisement. This created an effect so that it appeared as if the mouth was moving and forming into the shape of a “kiss”. The third stimuli, Ad3, was a dynamic animated banner advertisement based on the general advertisement. The animation was equivalent to Ad2. Ad3 also had pop-up properties distinguishing it from Ad1 and Ad2. Ad3 was programmed to appear in the same area as Ad1 and Ad2. The size of Ad3 was the same size as Ad1 and Ad2, 283 times 250 pixels. Though, since Ad3 was a traditional pop-up advertisement it came with a standard Windows pop-up frame that added 50 pixels on the height and 25 pixels on the breadth. This meant that the pop-up advertisement in fact came with a frame that made it 21 075 pixels large or 29.2% larger than Ad1 and Ad2. This size difference was an unwanted but necessary difference in order to study advertisements of the same size. To control for the size effect just mentioned there was also a fourth advertisement, Ad4, designed. This advertisement, Ad4, was similar to Ad3 since it was a pop-up advertisement. However, it did not have the frame that Ad3 had. The treatment with Ad4 was primarily designed to investigate the effect of abrupt presentation and the effect of the frame presentation.
6 7
In english: For a healthier mouth In english: The ?rst toothpaste that ?ghts tartar and gives you whiter teeth at the same time.
90
The advertisements Ad1 and Ad2 were presented on the Web sites throughout the whole experiment session of 6 minutes. Ad3 was a pop-up advertisement and was launched after 15 seconds, 60 seconds and after 180 seconds. In the second round of experiments the average exposure time was recorded for the respondents subjected to Ad3 and was less than 90 seconds. The exposure time here is stated as “less than” 90 seconds and the reason for this is that the automatically set exposure time for Ad4 was 90 seconds. So as long as the time for Ad3 was less than these 90 seconds we can be sure that the average exposure time for Ad3 has not been in favor of Ad3 compared to Ad4. The hypothesis for the comparison between Ad3 and Ad4 expects Ad3 to be more effective and a longer exposure time for Ad3 would jeopardize the conclusions. An equivalent or shorter exposure time would on the other hand ensure us that exposure time is not a factor here. As a matter of fact, in a majority of cases the exposure time was considerably less than 60 seconds for Ad3 making the exposure time a less important factor in this case. This exposure duration ?ts well with for instance the study reported in the theory chapter where a majority of respondents clicked away pop-ups within 3 seconds after appearance or clicked on some other link or area within 25 seconds making the pop-up disappear. Advertisement Ad4 was also a pop-up advertisement and was automatically launched after 15 seconds, 60 seconds and after 180 seconds, the same time intervals as Ad3. Ad4 was only used in the second round and gave the respondents a total exposure time of 90 seconds. The ?rst experiment aimed at testing the following treatments: Treatment 1 comprised of Ad1 presented in Web site “A”. (static) Treatment 2 comprised of Ad2 presented in Web site “A”. (animated) Treatment 3 comprised of Ad3 presented in Web site “A”. (pop-up) Treatment 6 comprised of Ad4 presented in Web site “A”. (ad without a frame) The internal drive experiment was designed to study how internal drive or goal oriented behavior impacts the effect of attention to advertising. In order to test this hypothesis an experiment was created where the entire design actually was based on a difference in instructions. The respondents were given slightly altered instructions in order to induce an internal drive and to make them more goal oriented. This was accomplished with a few very simple questions that could be answered with a simple yes or no. The instructions that were given can be found in appendix B. The standard instruction that was given is called the “A-instruction” and the alternative, internal drive inducing, instruction is called the “B-instruction”. The group that got the B-instructions was treatment 4.
91
This experiment aimed at testing the following treatments: Treatment 3 comprised of Ad3 presented in Web site “A” with the “A-instructions”. Treatment 4 comprised of Ad3 presented in Web site “A”. (B-instructions, goal oriented behavior) The Web environment experiment was designed to study how the physical task environment or the Web environment impacts the effect of attention to advertising. In order to test this hypothesis the experiment was designed so that two different Web environments, with respect to the search depth structure, were given as treatments to the respondents. The two task environments differed in their complexity and this was the variable to be studied. One environment was less complex, Web site “B”, than the other environment which was more complex, Web site “A”. Hence, the advertisement stimulus was placed in two different contexts. The Web environment experiment aimed at testing the following treatments: Treatment 3 comprised of Ad3 presented in Web site “A” with the “A-instructions”. Treatment 5 comprised of Ad3 presented in Web site “B” (less complex environment) with the “A-instructions”. In the ?rst round of experiments treatment 1 through 5 were run and in the second and extended round of experiments treatment 1, 3, 5 and 6 were conducted.
The main measures used
In order to measure the effects of the various treatments a number of means to assess the effects of the stimuli were employed. The main measures that were used were recognition and recall. The adequacy of using these measures has been elaborated on in the theoretical chapter but in addition to those descriptions the following statement can be added:
“The difference between recall and recognition is that in a recall test a subject is given a set of information and is later given some minimal cue and asked to retrieve and reconstruct the original information, whereas in a recognition test the subject is confronted with the original material and asked whether it has been seen or heard before. Hence, for recall the individual must describe the stimulus which is not present; for recognition the stimulus must merely be identi?ed as having been previously seen or heard” (Singh & Rothschild, 1983, p. 235).
92
In some sense it is “easier” to recognize than to recall and recognition is more important than previously thought (Bettman, 1979). Krugman has also commented on the issue:
“[…] the inability to recall something does not mean it is forgotten or that it has been erased from memory. The acid test of complete forgetting is if you can no longer recognize the object” (Krugman, 1972, p. 14)
The recognition measure is thereby a more sensitive measure that can detect effects where recall cannot. In connection to the experiments both recognition and recall was used to measure the effects but recognition was used as the main measure in the analysis of the data. Recall was measured by asking respondents whether they could recall if they had seen any advertisement on the Web page that they had visited. If they answered yes they were presented with four additional questions where they could write the brand name, headline, copytext and/or describe what the advertisement had portrayed. In order to measure recognition, those who could not recall (and also those who could recall) the advertisement were thereafter presented with images of the target advertisement and four distractor advertisements. The respondents were then asked to indicate if they could recognize any of the ?ve advertisements presented. An additional measure that was used was the click-through frequency. The clickthrough frequency was measured when a respondent indicated in the survey that she had actually tried to click. The reason why there was no possibility for actual click-through to a target Web site was that these respondents would have gotten a different stimulus treatment than the other respondents, had this possibility existed. However, it has been mentioned previously that the clickthrough measure has never been a legitimate measure of advertising effect, which is also one of the reasons why it is being included and studied. In addition to these main measures described above attitude measures were used. These questions were asked using a 7 point likert scale where the respondents could indicate whether they had a positive or negative position towards the statement or question. These questions were carefully developed and ?ne tuned based on input from the tests of the questionnaire.
Selection
The procedure of selecting the respondents for the main study is mainly guided by an interest of targeting individuals with skills relevant for the study. Since the experiments require familiarity and skills in how to use the Web, the Internet and also knowledge of how to use computers, a younger selection was preferred. Even though computer literacy is fairly high in Sweden, younger individuals have greater skills than those who are older. Furthermore we also
93
know that students (compared to non-students) are forced to use computers on a day to day basis as part of the various educational programs. It is therefore more likely that students, as opposed to other groups, will satisfy the basic requirements for the experiments. Support for these notions can be found in other researchers’ studies of the World Wide Web and the Internet. For instance, Gallagher, Parsons and Foster (2001) conducted a series of studies in laboratory environment as well as home environment and used student samples as well as random samples. They also replicated their studies. Their conclusions were that there sometimes are circumstances when a student sample is particularly suitable to use. They state that “…university student samples may provide the best available indicators of where the general population will be when the Web becomes as familiar as traditional media, such as print and television.” With these above arguments at hand it was decided to do the study on a selection of students and in particular students at the Umeå University. By selecting students, by de?nition a convenience sample, for the study there were also additional advantages achieved. A major problem that could be anticipated was a reluctance to participate, i.e. non-response. A random selection of individuals from the Umeå area could have resulted in a nonresponse level well above 90%8. It is not unreasonable to believe that it would have been perceived as quite an obstacle from the respondents’ perspective to travel on average 5 kilometers and sometimes up to 30 kilometers just to participate in an experiment. The concern for a low response level was an additional consideration and concern when it was decided to use students for this study. By targeting students the problem with travel and ?nding the location was assumed to be reduced since students know the locations and frequently visit the area where the experiments were being conducted. Umeå University has around 28 000 students and is located in a community that has the highest Internet penetration in Sweden (Liljenäs and Öberg, 2001) and most likely also in the world. Already in 2001 there were more than 80% of the households that could access the Internet and were using the Internet on a regular basis (Liljenäs and Öberg, 2001), which is a fairly high percentage when comparing internationally. Of these 80% more than 50% of the households are connected to the Internet thru a broad band connection at a speed greater than 256 Kilo bits per second. Typically the transfer speed is 1.0 or 10 mega bit per second and sometimes even as high as 100 megabits per second. Taken together the use of Internet, with high capacity, in Umeå is thereby more widespread than what it is in any other place in Sweden. To study students in Umeå community was thereby a very suitable opportunity.
8
40 phone interviews, with individuals randomly selected from Eniro’s Umeå section, showed that less than 10% were willing to participate. The ?gure would most likely have been even lower if the request to participate would have been sent by letter, which was the initial intention. 94
In order to attract respondents to the experiments posters were put up on bulletin boards and the like at the University. On the posters students were asked to sign up for an experiment in “informatics and social science”. In addition to this a number of classes were approached and information was given to the students that they had a chance to participate in an experiment in “informatics and social science” and that they could sign up for participation. The classes that were approached were from all faculties at Umeå University the faculty of science and technology, faculty of social sciences, faculty of arts, faculty of medicine and faculty of teacher education. The selection comprised initially of 410 respondents but was extended with another 292 respondents. In summary, a total of 702 respondents (about 3% of the students at the University) were selected on the basis that they wanted to participate out of free will. The respondents were informed that they would receive 100 SEK as compensation9.
The experimental setting – the lab
The experiments were conducted in the A wing of the Naturvetarhuset at Umeå University (The building for science and technology)10. At my disposal for the experiments I had an entire corridor with seven identical of?ces and one board room. In ?ve of these of?ces equipment for the experiments were installed. The equipment in every of?ce comprised of 1 Dell stationary computer, 1 keyboard, 1 mouse, 1 mouse pad, 1 IBM 17” CRT screen and the necessary cables. Windows 2000 was installed on the computers as well as the internet browser Internet Explorer 6.0. The interior of the of?ces were similar, a desk with a calendar, a plant in the window, a bookshelf and an of?ce chair. The doors to the of?ces were kept open so that it was possible to observe the respondents. Though, they were not aware of that they were being observed. The corridor where the experiments were conducted was de?ned as restricted area and unauthorized individuals were dismissed. The reason for this was to avoid any unwanted interference with the experiments11. In connection to the corridor with the lab rooms there was a waiting room where respondents could sit down and wait for their turn.
9
In the evaluation of the experiment and the questionnaire the respondents received 50, 70 or 100 SEK. The respondents did however not know that it was an evaluation. 10 The choice of location is perhaps even reducing bias since respondents will not build up expectations about social science and advertising in Naturvetarhuset’s environment 11 At only one occassion the experiments were interrupted. Some chemistry students refused to follow instructions and consequently the experiments were terminated and cancelled for that entire day. They were however resumed the following day. That particular class had been chosen to participate in the experiments but was replaced with another class in order to avoid any negative in?uence on the experiments. 95
Instructions and procedures
First the respondents were welcomed and asked to take a seat in the waiting room. When it was time to start the experiment they were called in from the waiting room and asked to stand in the door way to the of?ce and experiment that they had been randomly assigned to. Thereafter the respondents received instructions that were read to them to them. The instructions were given to the respondents by either the author or by an assistant to the author. The instructions were read straight from a paper without deviation and with the same tone of voice. Since the respondents were Swedes the instructions were given in Swedish. However, the instructions have been translated into English and are reproduced in the appendix. The standard instruction to the respondents can be found in appendix B and is called the “A”-instructions. The goal oriented instruction given to the respondents can be found in appendix B and is called “B”-instructions. After the respondents had gotten the instructions they were asked to take a seat next to the computer. They were asked if they were sitting comfortably. They were also not to do anything before they got the command from the instructor. At the command of the instructor they commenced the experiment. After the experiment they were asked to turn around on their chair so that they would not look at the Website anymore. Thereafter the respondents were given the on-line questionnaire so that they could ?ll it out. When the respondents had ?lled out the questionnaire they submitted their answers to the database and were asked to come out of the lab. Some of the respondents were then chosen to participate in the “Exit poll”. Those who were not chosen for the exit poll went straight to the debrie?ng. All respondents were debriefed after the experiments and the exit poll; see appendix B “Debrie?ng”. The ?nal activity was to give the respondents 100 sek for their participation and they also got to ?ll out some personal information to participate in a lottery and to receive information about the results of the study at a later occasion.
96
The respondents’ journey through the experiment can be described as in the ?gure below.
Procedural Steps for the Experiments
Welcome Instructions Have a seat Start the experiment now! On-line questionnaire 6 minutes Figure 4.3
Exit Poll Debriefing Payment Thank you and have a nice day!
Figure 4.3 Schematic view of the procedural steps for the experiments
The survey
After conducting the experiment on the computer an on-line questionnaire immediately appeared on the screen for the respondents to ?ll out. The on-line questionnaire comprised of a total of 60 questions12. The questionnaire was designed in such a way that 5 to 10 questions appeared at the same time on the screen13. The respondents were to ?ll out these questions and could continue to the next 5 to 10 questions after clicking on a “continue” button. At the end of the questionnaire they pressed on a “submit” button to ?nish the on-line form. The questionnaire took on average 13 and a half minute to complete. The questionnaire is enclosed in appendix H.
The “exit poll”
After the respondents had conducted the survey 100 of them were randomly chosen to participate in an “exit poll”. The reason for this was to follow up some issues of interest for the study. Considering that the advertisement in the experiments was about tooth paste it was of interest to ?nd out to what extent the respondents could be considered as being in the target group. Only 2% answered that they had not bought tooth paste the last 12 months. These 2% stated that a parent was providing this product for them. A second question was whether they perceived the entire Web sur?ng experience to be realistic or not on the scale unrealistic-neutral-realistic. 89 percent stated
12 13
In the questionnaire the questions are numbered up to 63. Despite this fact there are only 60 questions. See Appendix H for an example. 97
realistic, 7 percent stated neutral and 4 percent stated unrealistic. This ?gure corresponds fairly well with the results from the survey. Notice that the questions are slightly different. A third question was asked to respondents that had not recognized the advertisement. These respondents were asked if they had had an intention to click on any advertisement on the Web site. Out of 42 respondents 0 respondents or 0% answered that they had had such an intention. A fourth question was asked regarding how dif?cult or easy they thought it was to “do what they had been instructed to do on the Web site”. The respondents that got the A-instruction responded almost exclusively “very easy” whereas the respondents who got the B-instructions most often responded “easy”. A ?fth question that was asked was what kind of expectations that they had on the experiment and what they thought that it was all about before it started. 4 % thought that it had something to do with business administration and none (0%) expected something in marketing communication (remember the information given when students were invited to participate and also the location where the experiments were conducted). A sixth and ?nal question was asked to hear out the respondents whether or not it had worked out well to conduct the experiment and to ?ll out the questionnaire. All respondents stated that it had gone smooth and without problems.
Validity and Reliability
Validity comprises of a collection of subdivided concepts with the overall purpose to describe or ascribe quality to a study. Validity, in its different guises, are providing a set of instruments to evaluate whether or not the ?ndings from a study can be stated to be valid and reliable in a number of respects. Internal validity Internal validity addresses to what extent we can infer that a relationship between two variables is causal or not (Cook & Campbell 1979 and Calder, Phillips & Tybout 1982). It should be further noticed that when we say “relationship” we are dealing with the relationship between the research operations “irrespective of what they theoretically represent (Cook & Campbell, p. 38, 1979). Thus, internal validity has to do with the relationship between X and Y and not the abstract constructs that X and Y may represent. Nevertheless, when the researcher is estimating the internal validity of a relationship the “investigator has to systematically think through how each of the internal validity threats may have in?uenced the data.” (Cook & Campbell, p. 55, 1979) The process is very much one of ruling out different threats to validity. Among the many threats to internal validity we can immediately rule out a number of them as a consequence of how the experiment was designed and
98
conducted. History effects, Maturation effects, Testing effects, Instrumentation effects, Regression effects, Mortality effects, Ambiguity about the direction of causal in?uence effects, Selection effects, Selection maturation effects and the like are not posing any problems since the respondents were randomized in this study. That is to say that they were randomly assigned to the treatment they received. This position is supported by Cook & Campbell (1979). Randomization makes causal inference easier. Randomization is however not ruling out all potentially harmful effects, for instance, imitation of treatments, compensatory equalization, compensatory rivalry and demoralization in groups receiving less desirable treatments. Demoralization can be ruled out because no one was in the position to know anything about potentially less desirable treatments and even if they would have known anything it is doubtful whether any of the treatments could be perceived as less desirable. Compensatory effects can also be ruled out since there were no reasons for respondents or a group of respondents to compensate in anyway for a treatment that they were subjected to. Actually it is hard to see how compensation in its normal form would take place in this case. The problem of one treatment imitating another treatment can also be ruled out in the same way as the compensatory effects. However, one risk that potentially could have impacted and negatively in?uenced the study is if the respondents violated the instructions and shared information between respondents. The potential effect of this information sharing could however be diminished since no one knew before hand in which treatment they belonged. Moreover, they were not informed of that even after the experiment either14. A speci?c point with the debrie?ng was to inculcate in the respondents the importance of not sharing information about the experiments with others (those who had not yet done the experiment) by stating the following in the last three sentences of the debrie?ng: “Furthermore, please, do not talk about or tell about the content in this experiment to anyone during the next 4 weeks since more than 600 students at Umeå University will participate. It is preferable if those who will participate do not know in advance about the content of the experiment since this will in fact ruin the experiment. The cost for this study is more than 1 million Swedish crowns so it is good if this study can be conducted in a correct and undisturbed way”. (Instructions in the debrie?ng, translated into English) Construct Validity Construct validity is what researchers are concerned with when they are trying to evaluate if the results are confounded or not. Confounding refers to the possibility that “the operations which are meant to represent a particular cause or effect construct can be construed in terms of more than one construct” (Cook & Campbell, p. 59). The example we saw before with
14
The respondents were briefed in a general way what they had been subjected to. 99
Clever Hans highlighted the issue – are we measuring the right thing? What one investigator interprets as a causal relationship between A and B might be interpreted by another researcher as a relationship between X and B or perhaps even X and Y. Measures to increase construct validity and to avoid confounding includes thorough planning and pre-experimental testing so that the constructs are clear and de?ned. Throughout the study this has been achieved by explicitly using constructs and their meanings that are commonly used in the theories used for this type of research. There are no deviations from the research main stream in that respect. The dependent variable, attention, is inherently connected with memory when it is measured with for instance recognition. In fact, both attention and memory are measured at the same time when using recognition as a measure. However, in this case the focus was more on the former than on the latter since measurements were conducted immediately after exposure, making the memory construct less of an in?uencing factor. In advertising research where the focus is on memory the time between exposure and posttest is counted in weeks or months rather than seconds and minutes as it was here. In all attention research, whether it is in advertising, psychology or some other discipline, the researcher will always face the inherent problem with the attention – memory constructs when using recognition as a measure. What can be said is that the problem has been observed and was reduced to the point where no additional countermeasures could be undertaken, and still use recognition as a main measure. The reason why the experimental design was integrated into one solution was because of this issue. It was thereby possible to reduce the time between the exposure and the following recognition test to an absolute minimum. When measuring the context’s in?uence on attention, two versions of contexts are used, i.e. the internal drive and the Web environment. Both the internal drive and the Web environment are facets of the same context dimension but they are expected to impact cognitive resources in a similar way. That is, they may both use more or less cognitive resources depending upon how demanding the situation is. The results in the analysis are indeed showing that these two measures of the context dimension can use either more or less cognitive resources as expected. Using two or more dimensions that are targeting the same construct increases the likelihood of achieving adequate results, especially if the constructs are co-varying (Churchill, 1979). From the analysis we can see that the measures co-varied and that they also behaved as expected, criterions that are positive indicators of construct validity (Churchill). In the case of studying attention getting techniques four advertisements are used for the study. These advertisements are supposed to re?ect contemporary attention getting techniques on the World Wide Web. That these advertisements are good representations of and thereby also measuring various levels of
100
attention getting is merely based on the mimicking of the function and visual appearance of common advertisements used in the 100 Web sites that were studied in the ?rst pilot study. Using these types of advertisements is in line with how attention getting has been researched in other studies. External Validity External validity is focusing on whether a causal relationship can be generalized to and across different persons, settings and times (Calder, Phillips and Tybout, 1982 and Cook & Campbell, 1979). Considering that a student sample is used it is appropriate to discuss how this may impact the ?ndings. Gallagher, Parsons and Foster (2001 A, 2001 B) conducted a series of studies in laboratory environment as well as home environment and used student samples as well as random samples. Their conclusions were that there sometimes are circumstances when a student sample is particularly suitable to use. They state that “…university student samples may provide the best available indicators of where the general population will be when the Web becomes as familiar as traditional media, such as print and television.” and furthermore “In these cases student samples are superior to samples of the general population, even when researchers wish to generalize the results to the larger population.” (Gallagher, Parsons and Foster, 2001 A, p. 81). Due to the fact that university students are experienced users make them more suitable for testing than those who are not familiar with handling computers, Web browsing etc. An additional argument in favor of the appropriateness of using university students is that, in Sweden where the study took place, about 50% (SvD, 2003) of the young population is continuing with academic studies. This means that “students” is not a small minority group that may deviate signi?cantly in their behavior. On the contrary, half of the population is “represented” when using student samples. Calder, Phillips and Tybout (1982) go one step further than Gallagher et al. when arguing in favor of using student samples. Their point is that if a homogeneous group is used for the study then there will be less variation in background variables within that group. They in fact state that “Random sampling … may actually interfere with achieving a severe theory test. This is because it is likely to increase error variance” (Calder, Phillips and Tybout, 1983, p. 241). Using a more homogeneous sample will have the effect that whatever phenomena that is under study may be more protruding or at least not obscured due to increased variation within the sample. They are furthermore stating that controlling, rather than manipulating and examining the background factors should be encouraged. Some proponents of the importance of external validity have argued that research that is weak in external validity is not considered being an adequate test of theory (Lynch, 1982). Cook and Campbell are though taking a different stance and in line with Calder et al. stating that:
101
“The priority among validity types varies with the kind of research being conducted. For persons interested in theory testing it is almost as important to show that the variables involved in the research are constructs A and B (construct validity) as it is to show that the relationship is causal and goes from one variable to the other (internal validity). Few theories specify crucial target settings, populations, or times to or across which generalization is desired. Consequently, external validity is of relatively little importance. In practice, it is often sacri?ced for the greater statistical power that comes through having isolated settings, standardized procedures, and homogeneous respondent populations. For investigators with theoretical interests our estimate is that the types of validity, in order of importance, are probably internal, construct, statistical conclusion, and external validity.” (Cook and Campbell, 1979, p. 83):
Considering these convincing lines of argumentation a student sample was employed for the studies conducted in this research. Transferability Even though external validity was considered the least important to Cook and Campbell this did not discourage the author to put in considerable effort to reassure the transferability aspect. It was considered important that the experiments would be good representations of hands-on experiences in the out side world (outside of the lab). This meant that the experiments were constructed to be as realistic as possible. As a matter of fact, many of the respondents asked informally if it was a new news Web site that was used for the experiments and where they could ?nd and use it at a later point in time. This kind of inquiry could perhaps be interpreted as a quality mark for the experiments’ realistic design. Czarniawska (1999, p 7) argues that the transferability of an experiment to any other kind of social reality is a matter of whether the reality of the experiment have traits in common with these other realities. Hence, the issue is not whether the experiment is considered “unreal” since the experiment as a matter of fact constitutes a reality in itself, and thereby it is real. Instead it is a question of how many real situations that this tampered reality is a good representation of. That is though a question that can best be answered in relation to the properties of similar or dissimilar realities. In order to get an understanding for the phenomena in question it was decided to include a measure of realism. The respondents were asked to give their subjective view of how realistic they subjectively perceived the experiment to be, as a representation of a real world Web site experience. Their responses to this measure was that 94.5% of the respondents thought that the Web sites used in the experiments had a high or very high resemblance of real Web sites. Another 4.8% thought that the resemblance was neutral or slightly dissimilar whereas merely 0.6% or 4 respondents thought that the experiment Web sites
102
had little resemblance with real Web sites. Not one single respondent, of the 702 that was included, ticked the alternative that the experiment Web sites had very little resemblance with their out side world counterparts. These ?gures are assuring a high level of transferability as discussed in the above. Reliability Reliability refers to how stable results are from one test to another. If results would differ between tests conducted at different times then the measures are unreliable or suffer from low reliability. Low reliability in?ates standard errors and reduces statistical power, such as ?nding differences between means of different treatment groups (Cook & Campbell, p. 43). Often it is dif?cult, when conducting research, to point at how reliable a certain study is or how stable it has been over a period of time. This issue can sometimes not be addressed since the appropriate measures are absent. Though, in the present study it was decided to run a second round of experiments, which gave an opportunity for a test – retest measurement. Instead of merely enlarging the study with the new parts, it was decided to include the old treatments as well. This made it possible to see whether the results were stable. In the ?rst round treatments 1 through 5 were conducted and in the second round treatments 1, 3, 5 and 6 were conducted. Thereby the new results from treatment 1, 3 and 5 could be compared with the older results. The ?ndings from the analysis of reliability revealed that the measurements were indeed very stable since the differences within treatments were low. The difference in treatment 1 was 2.4% between test and retest, the difference for treatment 3 was 2.9% and the difference for treatment 5 was as low as 0.3% when measuring the dependent variable. These ?gures give an indication of high reliability and ensure stability over time. This also made it possible to merge the data into one data set in the analysis. The time between the test and the retest was 1 week.
Analysis
A major part of the work that was conducted in this study was the statistical analysis that took place after the data was collected. Data was collected through the on-line questionnaire and stored in a database. Data was then compiled into data ?les to be analyzed with the software package SPSS version 13.0. The software package is a sophisticated tool for statistical analysis in general and multivariate techniques in speci?c. Three different means for statistical analysis were used; Chi-square, Logistic Regression and ANOVA. ANOVA and Logistic regression will be described in the following.
Logistic Regression Analysis
The classical linear regression model can be used as a starting point in deriving the logistic regression model. As we know the linear regression analysis is
103
aiming at predicting the dependent variable with one or more independent variables as predictors. In traditional linear regression the dependent variable is continuous in nature and the independent variables can be both continuous and dichotomous in nature. The ordinary linear regression model is typically written in the following way: y= ?+?*X+e where y is the dependent variable to be predicted, alpha is the intercept, beta is the regression coef?cient and X is the independent variable. The term e is a random disturbance term with mean zero and variance ?2. This model can easily be extended to a multiple regression with the following formulae: y=?+?1X1+ ?2X2+... ?kXk+e The multiple regression is adding additional independent variables and their respective betas. However, both the simple as well as the multiple regression are adequate when using continuous or metric variables, but they are less appropriate when the dependent variable only can assume either of two values like categorical variables. In a logistic regression the dependent variable does not have to be continuous. Due to the underlying formulae the logistic regression is appropriate in particular when the dependent variable is dichotomous. For instance whether an individual owns stocks or not is a dichotomous variable. This can for instance be coded as 1 for individuals who own stocks and 0 for individuals who do not own stocks. Now, the dichotomous property is making it somewhat problematic to use either of the linear regression models above. The inherent problem with the dichotomous dependent variable, is that it can only vary between 1 and 0 when the classical linear regression model assumes a dependent variable that can vary between -? and +?. To solve this problem a logarithmic link is used. This link is called the logit function and it is the most popular linking function (Sharma, 1996). The logit function is a nice solution where log of odds are used since log of odds can vary between -? and +? and are symmetric, contrary to odds that can only range from 0 to +? or probabilities that range from 0 to 1. By using log of odds this link is added thereby forming a logistic regression. The logistic regression expression takes the following form: P(y=1)=1/e –(?0 + ?1X1 +
?2X2 + ... ?kXk)
Another way to formulate the model is: log [P(1)/P(0)] = ?0 +?1X1 +?2X2 + ...?kXk
104
When running a logistic regression in SPSS a considerable amount of output can be produced but for the purposes in this study the main interest is the calculated odds for the events that have been studied. Since odds and also log of odds are the measures being used in logistic regression these measures will be used in the analysis and presentation of results. A small recap of odds and their relation to probabilities is perhaps in place since odds are not as frequently used in Sweden as for instance in the U.S. Probabilities can range from 0 to 1. When stating that there is a “?fty-?fty” probability for the event Q to occur, then this means that the probability that the event Q will occur is 0.5 or 50% and the probability of non-occurrence is 0.5 or 50%. If this is transformed into odds it is adequate to say that the odds are 1:1 for the event Q to occur. This implies that the odds that the event Q will not occur are 1:1. A different probability, for example 0.66 or 66% for the occurrence of Q, will produce a different odds. Then the odds for the occurrence of Q are 2:1. A probability of 0.75 or 75% gives the odds 3:1 and so forth. A straightforward way to calculate the relationship between probabilities and odds can now be formulated: Oi=Pi/(1-Pi) When reading the diagrams, which essentially is output material produced by SPSS, in the analysis it is advisable to keep in mind that the logistic regressions are not primarily used for prediction. Instead, what are being aimed at are the differences in odds between treatment groups and whether this difference is signi?cant or not.
Analysis of Variance - ANOVA
A second tool for analysis that was used in this study was analysis of variance. It is often desirable to measure differences between treatment groups. The researcher may for instance be interested to see if there is an effect of a certain treatment. One way to go about is to measure the difference in means between the treatment group and the control group. This can for instance be done with an independent samples t-test. A problem that will surface pretty soon is though if there are many groups to be compared. The independent samples ttest is suitable for situations where the researcher is interested in whether two means differ. If there are more than two groups the researcher would have to conduct multiple pair-wise tests since comparing three or more groups at the same time is not possible with the t-test (Howell, 1997, p. 299). At the same time, conducting multiple t-tests may result in an increased risk of incorrectly rejecting a null hypothesis of no difference among groups. The problem with multiple t-tests is that as the number of groups increases, so will the likelihood of ?nding a difference between any pair of groups, simply by chance, i.e. when no real difference exists. Hence, the risk for making this error (a type I error or ?) increases as the number of comparisons made increases (Chu, 1999). One possible solution would be to use a Bonferroni correction, which means that the signi?cance level is divided with the number of tests conducted (Chu).
105
An additional and more appropriate way to go about is to use analysis of variance, ANOVA. The Anova analysis is widely used and allows to measure whether three, four or k means differ (Howell, 1997). This kind of analysis makes it redundant to conduct multiple pair wise comparisons and is thereby reducing the risks for errors related to the increased numbers of comparisons. When using the Anova to compare multiple means we are actually investigating whether there are no differences between the treatments or if there indeed are differences. If there is a difference in one or more means then we can state, with some degree of certainty, that this is a treatment effect. That is, an effect related to the various treatments distributed to the various groups. This is the exact way that this analysis technique is being used in the analysis section.
Operationalization of Hypotheses
In order to appropriately answer the research questions posed, the hypotheses generated in the previous chapter have to be examined in some way. The hypotheses do not say much about how they are going to be tested at an operational level. They are merely statements that are pointing in a certain direction. These statements have to be tied to speci?c measures and in particular to parameters that can be tested statistically. The hypotheses are operationalized by transforming the statements into parameters that can be tested. In this case it is the exponential betas of the logistic regression that are being tested, Exp B. Recall the expression stated for logistic regression: 1/e –(?0 + ?1X1 + ?2X2 + ... ?kXk). In the following the hypotheses H1 and H8 are operationalized and formulated in the way that they will be tested. Hypotheses H2 through H7 are operationalized following the same pattern as H1. The display of H2 – H7 is therefore redundant. In the hypotheses chapter hypothesis H1 together with the null hypothesis for H1 was formulated in the following way: H1: A situation where respondents are exhibiting a less goal oriented sur?ng behavior will elicit more attention to an advertisement than what a more goal oriented sur?ng behavior will. To be rejected; [H0H1 A situation where respondents are exhibiting a less goal oriented sur?ng behavior will elicit equal or less attention to an advertisement than what a more goal oriented sur?ng behavior will.] In the statistical model to be tested in connection to hypothesis 1, attention is the dependent variable, the dichotomous y-variable. The independent xvariable is the treatments which are treatment 3 and 4 for hypothesis H1. In the statement in H1 it is expressed that in a “situation where respondents are
106
exhibiting a less goal oriented sur?ng behavior”, which they are expected to show in treatment 3, “will elicit more attention to an advertisement than what a more goal oriented sur?ng behavior will”, which they are expected to show in treatment 4. The Exp B - exponential beta - is expected to be greater than 1 if there is a larger effect in treatment 3 compared to treatment 4. The input data in the model is such that; the y-variable is either 0 (for non attention) or 1 (for attention) and the x-variable is 0 (for treatment 3) and 1 (for treatment 4). The model used in the statistical test is then; Logit(P1/P0)=a+BX. When hypothesis H1 is operationalized it is expressed in the following way: H1: Exp B > 1 H0H1 Exp B = 1 The reason why H0H1 is not expressed as Exp B ? 1 is because when conducting the statistical test then H1 is in fact compared with a point in H0 where no difference between treatments is assumed, i.e. Exp B for the two treatments are equal15. The test of hypothesis 8 is a test of multiple treatments means using analysis of variance. The variable to be tested is “annoyance” that is expected to increase as attention getting increases. It is this direction of the relationship that is expressed in the hypothesis. H8: The risk for an advertisement to be perceived as annoying will be greater when more attention getting techniques are employed in the advertisement. This would mean that a static advertisement will be the least annoying, an animated advertisement will evoke intermediate annoyance and a pop-up advertisement will be the most annoying. It would be preferable if hypothesis H8 could be operationalized in the following way: [H8: µs