Introduction

The Internet is "new and quite different from previous forms of mass media, exploratory studies to 'test the waters' should be done rather than using current theory more appropriate to traditional mass media studies (Wotring, Kayany and Forrest 1996 p. 273)." This early research tests the waters, investigating navigational behavior on the World Wide Web.

Several theories suggest two alternative web navigational patterns, searching and surfing. Hoffman and Novak (1996) used Hedonic Consumption (Hirschman and Holbrook 1982) and Uses and Gratification (Rubin 1984; Rubin and Perse 1987) to develop their goal-directed and experiential web navigational behavior. Visual Information (Janiszewski 1997) and the Elaboration Likelihood Method (Petty and Cacioppo 1986) also suggest searching and surfing the web.

This study investigates web page clicking behavior to support, extend and adapt these theories to the web. Clicking, jumping from one web page to another, serves as both an interactivity and revenue measure for websites. A better understanding of clicking behavior should lead to easier navigation for the user and improved site design for the website owner.

Specifically, this research will address the effect of a web page image's size and location relative to textual links. Two common web page ingredients are images and textual links. Do different web page layouts change website clicking behavior? A web page's image size and location should influence surfers differently than searchers.

Problem Statement

The web is relatively new, arriving in the early 90's. Scholars and practitioners seek practical and theoretical guidelines to harness this new communication tool's power. This clicking behavior research is important for theoretic and practical reasons.

Increasingly businesses, individuals, organizations and academicians quest better use of their World Wide Web resources. A fundamental web question, "What makes people click?", should occupy researchers well into the next millennium. Clicking brings visitors to and escorts visitors within a website. Clicking research helps website designers, owners and managers improve their website.

Literature and practitioner reports suggest that an image's size and location relative to textual links should influence clicking behavior. As the image becomes smaller, more textual links appear on the opening screen. More choices should increase clicking behavior. Additionally, some literature (Janiszewski 1990) suggests that text to the right of images should outperform text to the left of images.

Clicking behavior insights aid individuals and organizations seeking to effectively use and market their website. This research should provide clues towards the effective juxtaposition of images and text in order to improve a website's navigational flow.

Hoffman and Novak's (1996) proposed web navigational modes, goal-directed and experiential, incorporate current theory and build the foundation for this research. Hedonic Consumption's (Havlena and Holbrook 1986) expected utilitarian and experiential benefits and Uses and Gratification's (Rubin 1984) instrumental and ritualistic media orientation helped form Hoffman and Novak's web navigational styles.

In addition to revisiting Hedonic Consumption and Uses and Gratifications, the Elaboration Likelihood Method's (Petty and Cacioppo 1986) central and peripheral processing and visual information search's (Janiszewski 1997) goal directed and exploratory behavior seem applicable to the web. Hoffman and Novak (1996) together with the four other theories provide a similar view and solid foundation towards web navigational behavior. Each theory proposes similar dichotomous behaviors applicable to navigating the web, searching and surfing.

Early research (Hofacker and Murphy 1997 a,b ; Drèze and Zufryden 1997) indicates surfing elements in clicking behavior. If either the image's size or location influences clicking behavior then this research further argues the existence of a surfing element in web navigational behavior. The combined literature review synthesizes the current theories towards a parsimonious view of web navigational behavior and provides a convenient theoretical hook for communication, marketing and psychology scholars.

Background

For a quarter of a century, the Internet languished in obscurity. From the experimental beginnings in late 1969 until the early 90's, the press ignored - and the general public took little interest in - this complicated, computer driven mode of communication. Today, who has not heard of the Internet?

Close to 20 million computers, or hosts, connect to the Internet today. As many of these hosts represent a network of computers, the total number of computers connected to the Internet extends well beyond 20 million -- and keeps growing.

Recent estimates put the number of American Internet users between 31.3 million (Napoli 1997) and 50.6 million (Weise 1997). Despite the disparities, both articles estimates saw Internet users double last year and forecast similar expansion.

Like estimating Internet users, estimating web pages is a quixotic task (Murphy and Lacher 1996). Estimates on the number of web pages in August 1996 ranged from 40 to 100 million, with an annual growth factor of 10 since 1994. As the Alta Vista search engine's senior engineer Luis Monier noted, while there are roughly 100 million web pages, he could be off by a factor of ten.

Perhaps the most interesting Internet growth statistic is the Internet's influence on the US economy. The US economy's growth in 1996 can be attributed solely to the Internet (Amano and Blohm 1996).

Regardless of the growth metric's accuracy, the web helps fuel the Internet's astounding growth. Companies, organizations, universities, students, societies, newspapers, individuals and miscellaneous others flock to the web. If the sanguine World Wide Web predictions coursing through the world press -- and the World Wide Web -- are believable, World Wide Web users should reach the hundred million mark by the end of this millennium (Meeker 1996). Web indices such as registered domain names, web pages and hosts mushroom similarly (Meeker 1996; Murphy and Lacher 1996; Murphy and Massey 1997).

How do the many and diverse users use this new medium -- the web? Individuals, businesses, educational institutions and others grapple this conundrum. Confounding the conundrum is the web itself, continually evolving. The web is a more than a moving target; this mercurial bulls-eye mutates as well as moves.

A server's log files help track a web site's evolution and usage. Web site metrics such as clicksteams, click-throughs, hits and visits center around the question, "What Makes People Click?"

Clicking

Clicking plays a critical role in website models. Clicking brings visitors to and keeps visitors at a website. The visitor clicks on an external web page's link to arrive and clicks on a web pages internal links to explore the website.

An earlier experiment by Hofacker and Murphy (1997a) increased an advertising banner's click rate three fold by changing the copy. Since advertising banners are often sold by the click, this could multiply the website's advertising revenue comparatively. This research should provide clues towards improved advertising banner layout and placement.

Advertising banners are also sold by the impression, costing from next to nothing to a couple hundred dollars per thousand impressions. For this model, keeping the visitor on the website the longest possible works best. This research should provide clues towards the effective juxtaposition of images and text to increase a website's payable impressions.

For all website managers, even those not tied to an advertising model, improving the website is important. This research hopes to encourage further use of web-based research techniques to improve websites.

Central Thesis

Several theories help make sense of what makes people click on a link, each offering a similar dichotomy on user behavior. Is the user searching or surfing the web?

A requisite first step for tackling if searching and surfing apply to the web, is tackling how the web communicates. Comparisons to other media so far, include print, television, telephone, radio and video. Furthermore, Turkle (1995) posits that Internet communication has strong virtual characteristics.

The web is a new communication medium -- unlike existing media and difficult to classify. With other media, communication is interpersonal, group or mass. Morris and Ogan (1996) suggest a mass media comparison while for Hoffman and Novak (1996) the web weaves intricate twists on a many-to-many communication model.

Interactivity distances the web from traditional mass media (Hoffman and Novak 1996). Normally associated with interpersonal (one-to-one) communication, interactivity as defined by Rafaeli (1988) applies to new media. He defines three levels of interactivity: "two-way (non-interactive) communication, reactive (or quasi-interactive) communication, and fully interactive communication" (1988, p. 119).

Conventional mass media broadcast preformatted content in a passive one-to-many communication model (e.g., Laswell 1948; Katz and Lazarsfeld 1955); the media mandate what the masses may consume. Clicking on a link -- be it an image map, button, picture or text -- and changing web pages differs from changing TV, radio stations or newspapers. On the web, the user selects preformatted or customized content; the surfer tells the browser where to go.

Newhagen and Rafaeli (1996) suggest communication's 'Uses and Gratification' approach for the web. Rather than focus on the message, uses-and-gratifications focuses on the receiver -- an active, goal-directed media consumer. Expected gratifications drive media usage.

Uses-and-gratification's distinction between instrumental and ritualistic media orientation fit navigating the web. Instrumental users drill down and work a site, seeking information (searchers). Ritualistic users wander all over the web, seeking entertainment (surfers).

Marketing's 'Hedonic Consumption' (Hirschman & Holbrook 1982) categorizes a product's benefits as either utilitarian (search) or hedonic/experiential (surf). A consumption experience has afferent physical attributes such as taste or color and efferent experiences such as fun, fantasy and feeling.

Experiential products like going to a movie, theater or football game are consumed diffeerently than utilitarian package goods or major durables. Hedonic Consumption's utilitarian web user, like Uses and Gratification's instrumental web user, consumes purposefully.

Hoffman and Novak (1996) used hedonic consumption and uses and gratifications to introduce their web navigational modes, goal-directed and experiential. Two other approaches hold promise for extending Hoffman and Novak's navigational modes.

A persuasion approach the 'Elaboration Likelihood Model'(Petty & Cacioppo 1986), posits that individuals process information based on central (search) or peripheral (surf) cues. Janiszewski (1997) views visual information search as a combination of two distinct types of behavior, goal-directed (search) and exploratory (surf).

Searching and Surfing

Although the above cases point towards dichotomous behaviors, the distinction is a continuum rather than a dichotomy. Individual differences drive a person's information and entertainment consumption processes. Additionally, individuals shift from one mode to the other. If these theories apply to the web, users may start surfing a web site, then shift to searching if the information interests them, or vice versa.

Furthermore, Janisewski (1997) demonstrates that in spite of an individual's search mode, goal-directed or exploratory, better layouts increase attention for both visual information search types. Layout is important, no matter what the motivation.

Lord, Lee and Sauer (1995) show a similar effect, using the Elaboration Likelihood Model. They "offer evidence for a 'combined influence hypotheses,' viewing the attitude towards an ad as a function of both central message arguments and peripheral cues (p. 74)."

Do the above implications apply to the web? These theories suggest that subtle web page layouts will appeal to the web surfer in all of us, increasing the clicking on a web page.

Early Results

DoubleClick, a New York city media company and recognized leader in the emergent field of delivering ad banners on the web continually tests ad banners. Presentations that increase clicks include animation (25%), cryptic messages (18%), posing questions (16%), call to action phrases (15%), color and animation; a sense of urgency tends to decrease response rates (DoubleClick and I/PRO 1996).

Hofacker and Murphy (1997a) found that offering less copy on an advertising banner increased clicking. They also found (1997?) that the adding more banners on a web page influences total banner clicking.

Hofacker and Murphy (1997b) found that textual link location influenced click rates. The top link on a list of links, for example, garners more hits. Finally, Hofacker and Murphy (1997 ?) found evidence, albeit conflicting, that image location relative to textual links influences click rates.

Drèze and Zufryden (1997) experimented with a web page's background, image size, sound file displays, celebrity endorsements, Java Scripts, frames, and operating systems. Rather than measure clicks, their metrics were time on site and pages accessed. They found that the 'Number of Pages Accessed' and the 'Time Spent' could be explained by the alternative attribute designs of these independent variables.

These early web experiments tend to support the subconscious aspects of clicking and a surfing element in web navigation. A web page's image size and location should influence clicking behavior.

Methodological Issues

Although the web suffers from a lack of existing research, advertising offers one possible methodological avenue. Rice's (1992) review of computer-mediated-communication found little theoretical or empirical research in this area. The web surfaced after Rice's review, making his observation even more poignant. "In theory," say Novak and Hoffman (1996), "institutional advertising practices and metaphors can be borrowed from traditional media environments..."

Novak and Hoffman (1996) operationalize web interactivity with traditional mass media's effective reach and frequency metric and new web metrics such as click through, duration time and site depth. How deep a user explores or how long user remains on a web site, depends upon their clicking behavior. The click -- on an advertising banner, image map or hypertexted link -- determines how a user navigates or exits a web site.

Clicks on a web page's links will be used to determine the efficacy of different web page layouts. This study uses a three by four factorial design to measure clicking behavior, using three image map sizes and four image map locations (top, bottom, left and right) relative to the same textual links. The textual links remain constant, but the image maps range in size and vary in location relative to the textual links.

Do certain layouts have higher click rates, as suggested by these five theories?

This is a field experiment, using the opening page of a live website. A cgi program will randomly assign one of the dozen possible web pages to each web visitor. The website's log files contain each visitor's activity, noting which web page a visitor saw and what links if any that same visitor clicked.

Click through rates will be compared across the twelve different web page layouts to determine if image size and location influence web page clicking behavior.

Organization of Dissertation

The dissertation has five chapters. Chapter two reviews the literature in this area, including the Internet's history and diffusion, the web as a medium, communication, communication effects, and five theories applicable to web navigation. Chapter two ends with the rationale leading to four specific hypotheses concerning the effect of image size and location on a web page's click rates.

Chapter three discusses this unique methodology praised for its unobtrusiveness, external validity, large sample sizes, timeliness and low implementation cost (Dréze and Zufryden 1997). The methods chapter also discusses the special data cleaning steps and statistics used for this nascent data source, log files.

Chapter four will present the results, and chapter five will provide conclusions, implications and suggestions for future research.

Return to Proposals     Return to Course Notes    Return to Syllabus