Organization site - Summer 2001
Banner Advertising: Collating with content -- Pt 1
By David R. Thompson
and Birgit Wassmuth
Editors note: A version of this article was presented at the annual Association for Education in Journalism and Mass Communication Convention, August 9-12, 2000, Phoenix, Arizona
Effective placement of online advertising is one factor in the financial success of online newspapers.
This article reports a content analysis of online newspaper practice regarding delivery of appropriate advertising messages to an audience by collating editorial content and banner ads. For example, a banner ad for ordering tickets to St. Louis Cardinals games may appear on the same "page" as a sports story about Mark McGwire's latest home run streak.
In general, findings show a low level of ads related to content found on home pages of online newspapers. An increase in the percentage of banner ads targeted to page content occurs after the first, second and third clicks. Even then, only about half the banner ads are match page content.
The purpose of this pilot study was to conduct a content analysis of the current practice of collating content of banner advertising and editorial content in online newspapers.
Personalization of online news may be enhanced by improved methods of matching, or "targeting," the content domains of ads and news items. For example, a banner ad for ordering tickets to St. Louis Cardinals games may appear on the same "page" as a sports story about Mark McGwire's latest home run streak.
For the purpose of this study, the researchers have used the term "collating" to refer to the process of matching a banner ad to page content.
Ideally, improved targeting of online ads may result in a greater likelihood that the ad will be read. Click-through should increase. Effectiveness of the ad should be enhanced. And the newspaper will have found another way to improve the self-sustainability of its online product.
For the purpose of this pilot study, the researchers concentrated on literature related to online advertising and methods of targeting ads to online readers. They believe this literature represents general principles that will transfer across content areas, particularly the relationship between advertising content and editorial or content found in online newspapers.
According to Robinson and Kaye ( 1997, p. 219), the Internet is an attractive advertising market because: Advertisers can tailor messages for certain groups of consumers by channeling ads through appropriate Web sites thus capturing a specialized viewing audience; Cyber ads are updated and changed much more quickly and easily than traditional media advertisements; and Web sites provide prestige and a competitive arena allowing many small businesses to compete with companies which have large print and television ad budgets."
This suggests that tailored messages delivered to targeted audiences may be accomplished by effective placement of ads on relevant online pages, or files. The instant publishing aspect of the Internet—with no wait for a printing deadline or for air time for television programming—provides an opportunity for advertisers to respond quickly to social or world events, modify banner ads, and replace old banners with fresh ads targeted to updated content on the Internet. Small businesses, including newspapers with low circulation, may compete in a global marketplace - or at least a regional market - with the appearance of a high-dollar advertising budget.
Four methods of targeting online ads to an audience are considered in this literature review: registration, cookies, sponsorship, and targeting banners to online content (sometimes referred to as "placement").
Registration, the first method of targeting ads to an online audience involves collecting demographic information from users. "Registration" is sometimes required before entering a site on the Internet. Often, a registration form appears on the first "page" of an online site. The form may include an option for registered users to enter a password for access to the site.
Registration forms vary widely regarding information requested from online visitors. Once the online registration form is complete, the user submits the form to the host of the site, and entry to the site is permitted.
The content provider analyzes this demographic information. Then, informed decisions may be made about how to target online advertising to this particular audience.
A second technique used to gather information about online users involves "cookies."
According to Robinson and Kaye ( 1997, p. 218), "Netscape's new 'cookies' tool is designed to download ads targeted at a specific consumer group. For instance, a woman over the age of forty can receive a different style of Levi's ad than a teenager accessing the same Web page."
Of course, Robinson and Kaye's example assumes that the two women are not using the same computer with the same browser software.
Cookies are placed on the user's computer by the information provider's server. Often this placement occurs without the user's knowledge. For this perceived invasion of privacy, cookies have come under close scrutiny.
Basically, a cookie collects information about the user, such as what sites he or she visits on the Internet. This information is relayed back to the server that placed the cookie. The information is analyzed and used to target ads to that individual user.
Theoretically, a cookie can identify a specific user and "automatically" deliver advertising that should appeal to that person. This means that a cookie cannot be effective until a user visits the same site a second time.
For this study, cookies are not a factor. Coders were instructed to use "first load" ads only. No sites were visited twice, so any cookies that may have been deposited on the coders' computers never had a chance to work.
A third form of targeting online advertising to an audience involves sponsorship of a site. For example, Air France may have chosen to sponsor the Web site for the 1998 World Cup soccer tournament. The World Cup games were played in France, so this would be an excellent opportunity to market event-related special fares on Air France.
For the purpose of this study, "sponsorship" is defined as evidence online that "[this] site is sponsored by [this] company." For this content analysis, coders were instructed to look for the words "sponsored by."
McDonald (1997, p. 27) describes possible advantages of site sponsorship: "By sponsoring a site that consumers value, the advertiser will hope to build positive associations for the brand. The communication limitations of banners [namely, bandwidth (p.25)] will be overcome by surrounding content with imagery related to the sponsoring brand."
Perhaps McDonald would agree that companies would sponsor sites that feature content that complements the products or services sold by those companies. For example, a local sporting goods store might sponsor online content related to football or baseball. Or a local sports bar might create an ad to promote its Monday Night Football specials.
The fourth type of targeting to an audience involves matching banner ads to content. Sometimes, this is referred to as "placement," meaning an ad is "placed" or put onto a content page.
Online researchers, however, sometimes refer to "placement" as "location of the ad on the page," such as top left, or bottom center. Physical location of the ad on the page is not a variable here. This study does report whether or not banner ads are targeted or matched to the content of the page on which they appear.
However, reporting the actual intent of the Web page designers is beyond the scope of this pilot study. So, assessment of a deliberate attempt to engage a process of collation is not possible here.
To define banner ads for this article, the researchers consulted the literature, where they found a variety of descriptions.
Bucy, Lang, Potter and Grabe(1998,pp.11 and 15), use the term "banner" to indicate the online equivalent of a newspaper's nameplate. To them, a "banner ad" is an ad placed within the nameplate, as opposed to an ad placed within the "body of the Web site."
Banner ads have been described as online billboards (Harvey l 997, p. l 2) and as "the interactive equivalent of the roadside sign" (McDonald 1997, p. 25).
For the purpose of this study, the researchers have adopted a definition found online: A banner ad is "a graphic or image used for advertising on the Internet." This image is usually, still, a "gif" (Graphic Interchange Format) file. But other forms of ads, such as HTML-based banners have come into use (Ban ner Design 1998).
Technically, a banner ad may be any size. At the time this study was conducted, the Internet Advertising Bureau (Internet Advertising Bureau 1998) was promoting the use of its seven standard sizes: full banner (480 x 60 pixels); half banner (234 x 60); full banner with vertical navigation bar (392 x 72); vertical banner (120 x 240); square button (125 x 125); button #1 (120 x 90); button #2 (120 x 60); micro button (88 x 31).
However, this study does not investigate file for mat or size of banner ads.
There is some evidence that special content for Web-based newspapers is being created. KRT Interactive, a service of Knight-Ridder Tribune, advertises (KRT,1998): "Interactive news packages and KRT's exclusive QuickTime animations give your site superior content on breaking news and major stories."
In the researchers' opinion, this is the ultimate in targeting to online news content. News happens, and interactive content is prepared immediately for online distribution.
However, the authors have not found an advertising service that specializes in preparing ads for use with breaking news.
For example, President Clinton's videotaped grand jury testimony about his "improper relationship" with Monica Lewinsky, among other things, was released to the public on Monday, September 21,1998. On that videotape, Clinton repeatedly reached for a can of soda. It was a Coca-Cola product. In the researchers' opinion, if advertisers were prepared to respond online with an image ad for Pepsi within 24 hours, the effect could have been remarkable.
The researchers chose to conduct a content analysis that coded for content beyond the home page. An information location task (Guthrie 1988, Guthrie and Kirsch 1987, Guthrie, Britten and Barker 1991; Thompson 1993) was assigned that required coders to code every banner ad on each page encountered during a search for specific content. For the purpose of this study, a search for classified ad for a used BMW as used.
Because this study examines the opportunity of targeting within an information location process (i.e., a search for a classified ad for a specific product), the researchers examined literature concerning banner advertising in classified sites. One item was found.
In her comparison of display ads in print to online classifieds, Lainson (1998) states: "I definitely think there is something to be said for local display ads being placed adjacent to news to capture the attention of the page flipper. But I don't think the strategy will work for classifieds."
Apparently, Lainson has not considered the possible advantages of finding a banner ad that is targeted to the focus of her classifieds search. The authors suspect that a robust effect may be found by showing her a banner ad that "matches" the goal or destination of her search. For example, the most effective placement of a banner ad for a local Ford dealership may be the point at which classified ads for used Fords appear.
Related to task completion, such as a search for a specific classified ad, Hays (Media Buying Tips 1998) advises: "Avoid having your banner included on web pages that lead a user through a process. If a user is in the middle of filling out forms to finish a task, they are not likely to click on an ad banner."
This would suggest that no banners should be placed between the classifieds index and the final page of the search. But how would the computer know when the search is complete?
In general, prior research supports matching content of a banner ad with content of the page on which the ad appears.
"Banner ads for product categories that are highly relevant to the content of the Web site where they appear generate more clicking," Cho (1998, p. 303) states.
Andrews (1996, online) advises: "Keep ads in context. Always try to make the ad match the content of the page it will display on."
Angwin (1996, online) states: "Experts agree that the most successful banners are related to the site where they appear - such as an ad for bonsai trees on a gardening Web page."
This relates to literature that argues for collating content of a banner ad to content on the page that has "immediate relevance of the message to the audience" (Briggs and Hollis, 1997, p. 43).
This study codes for "banner targeted to content." In addition to checking for registration procedures and site sponsorship, the following tour research questions were formed, based on this literature review:
R1. How many newspapers used banner ads at all?
R2. Of all the "pages" viewed, how many included at least one banner ad?
R3. How many of the banner ads seen were targeted to the content of the page on which they appear?
R4. What is the breakdown of banners targeted to content by level of information hierarchy (location on the site, such as the home page, after the first click, after the second click, etc.)
This content analysis collected data about targeting banner ads to content of the "pages" on which they appear. Coders were instructed to code for: registration for entry into the site (yes; no); evidence of sponsorship (yes; no); banner targeted to content (yes; no; can't tell); banner ad(s) on home page (yes; no); site has classified ads (yes; no); banner ads placed in classifieds section (yes; no); and host of classifieds (in-house; third party; can't tell).
Content analysis was chosen because it is an appropriate methodology for collecting baseline data that demonstrate a specific advertising practice. Active online newspaper sites from across the United States were used for data collection.
Two coders were assigned an information location task: Find a classified ad for a BMW in the online newspapers selected for this content analysis.
An information location task that involves classified advertising is used because it allows the researchers to examine banner use on home pages, on classifieds index pages, and within all levels of the classifieds site.
Performing a search that progresses from general content (the home page) to more and more specific information (classifieds index, to automotive classifications, to classified ads for a specific product) provides a systematic method for examining targeting practices.
As a user navigates deeper into the site and moves closer to the goal of the information location task, an increasing percentage of banner ads targeted to content may be seen.
In other words, on the home page of an online daily newspaper, banner ads may not be targeted to "today's top stories." But when the automotive classifieds index is displayed, users should see a banner with content targeted to transportation (in general), for example. In this study, the task is: Find a classified ad for a BMW. Therefore, at the point when actual classified ads for BMWs appear, users should expect to see a banner targeted to that specific product, such as the name and phone number of the local BMW dealer.
Coders were instructed to code evidence of registration (yes; no) and site sponsorship (yes; no). Each banner ad, each "impression," encountered during this search task was coded for "banner targeted to content" (yes; no; can't tell).
For example, an ad for a Pontiac dealership would be coded as "no" on the home page, unless a headline for one of the top stories mentioned Pontiac (plant closing, new models released, etc.). The same Pontiac dealership ad would be coded "yes, targeted to page content" if it appeared in the automotive index of the classifieds site. And the same ad would be coded "no" if it appeared at the point BMW ads were listed.
Ads with ambiguous messages like "free stuff" or "loans" were coded as "no" on home pages and "can't tell" within the classifieds portion of the site.
After pre-testing four sites, the coding instrument was fine-tuned and finalized. Pre-test data were not included in the data analysis.
All data were collected in one week during September 1998.
The coders used identical computers, browser software, and Internet connection: an IntergraphTD22 computer (an IBM clone) with 48-MB RAM and Pentium II processor; an Intergraph 1 7sd69 monitor ( 1 7-inch color); Microsoft Internet Explorer 4 browser software; and a direct connection to a T- 1 fiber optic line.
A systematic sample was obtained from Editor & Publisher's online listings of newspapers in the United States (http://www.mediainfo.com/emediajs). The site included 790 daily, general interest newspapers published in the United States. Weekly, business, and other special-interest newspapers were excluded from the sample.
Fifty sites were selected for coding by dividing 790 by 50, which is 16. A random numbers table was used to establish a starting point. From there, every 1 6th daily, general interest paper was selected for coding. Anticipating the possibility that some sites may not load properly, the researchers selected four back-up sites.
Time allowed more sites to be coded, after coding the original 54. So another 20 sites were selected using the same procedure. A random numbers table was used to select a new starting point. This time, every 39th site was selected.
For intercoder reliability, the researchers selected an additional seven sites, about 10 percent of the sample of 74 online newspapers (considered accept able for the calculation of Scott's Pi). The same procedure was used to select these sites. For this sample of seven, every l13th daily, general interest newspaper site was selected.
This brings the total to 8 l sites selected for coding: 54 from the first selection, 20 more sites, and seven additional sites used for both the data set and intercoder reliability.
Each researcher coded sites independently. One researcher looked at 35 sites. The other researcher looked at 39 sites.
To establish an intercoder reliability score, both researchers coded the same seven sites on the same day at approximately the same time (the same ads in the same issue of the same online newspaper).
Exclusions: Three newspaper sites were no longer online and could not be coded. One newspaper required paid subscription for entry into the site. It was not coded. One site was not coded because it was "under construction." One other site was excluded from the data set because of bad data caused by failure of large portions of the site to load. The site loaded text, but all the images were seen as "broken" gifs. Therefore, it could not be determined whether or not the site included banner ads.
Usable data were gathered from 75 online newspapers.
Results and Findings
This pilot study attempted to assess the practice of targeting banner ads in online newspapers and provides baseline data for future studies on Internet advertising. Results have been reported with descriptive statistics.
This study coded 75 online newspapers using content analysis. The total number of "pages," or files accessed, was 280. An average of 3.73 pages per online newspaper was found. The total number of banner ads coded was 449.
To establish intercoder reliability, the same ads on the same edition of seven online newspapers were coded independently. Scott's Pi was used because it assesses both agreement and probability of disagreement (Scott l 955). The minimum level of acceptability for Scott's Pi is 6=.75.An independent researcher calculated the Scott's Pi Index. With the exception of "banners found in classifieds" (•=.70, N=7, disagree=l), all variables fell within the acceptable range.
R1. How many newspapers used banner ads at all?
Of the 75 online newspapers coded: 65 (87 percent) used banner ads; no classified ads were found on five (7 percent); and two (3 percent) had neither banner ads nor classifieds.
R2. Of all the "pages" viewed, how many included at least one banner ad?
Of the 280 pages coded: 166 (59 percent) included at least one banner ad; 112 (40 percent) had no banner ads; and 2 (1 percent) ads failed to load onto the page.
Fifty-seven (76 percent) of the home pages coded did include at least one banner ad.
Four targeting strategies were covered in the literature review. The use of "cookies" was excluded from this study. Registration for entry to the site, sponsorship of the site, and targeting banner ads to the pages on which they appear were coded.
Only one of the 75 online newspapers required registration for access to content: The New York Times (www.nyt.com). This site does place a cookie. So on subsequent visits to the site, the user should be "recognized" and allowed full access.
Evidence of sponsorship was found on two of the 75 sites.
R3. How many of the banner ads seen were targeted to the content of the page on which they appear?
Of the 449 banner ads, ] 22 (27.2 percent) were targeted to content; 314 (69.9 percent were not targeted to content; and 13 (2.9 percent) were coded as "can't tell." See Figure 1.
Two hundred thirty-seven (52.8 percent) banner ads were seen on home pages, one of which did not load. Of the 237 banner ads found on home pages, 214 (90 percent) were not targeted to page content.
The researchers believe this is an important finding. Advertising departments at these online newspapers are not matching banner ad content to page content in an effective manner.
R4. What is the breakdown of banners targeted to content by level of information hierarchy (location on the site, such as the home page, after the first click, after the second click, etc.)?
Table 1 and Figure 2 show data related to this research question.
The evidence shows an extremely low level of banner ads targeted to content on home pages and an overall increase in the percentage of banners targeted to page content after the first, second and third clicks. Targeting peaks at the third click.
The third click is the point at which many of searches for BMW classified ads were completed.
Although not defined as "the 3-click rule," common practice and one study in particular mention this principle of effective site mapping, or "story-boarding" the information hierarchy of a well-structured Web site.
Robinson and Kaye (1997, p. 221) state: "Users should be able to easily navigate a Web site. All pages should link back to the home page and content should never be more than three clicks away."
Effective targeting practice for this information location task (finding a classified ad for a BMW in an online newspaper) would suggest that banner ads become increasingly targeted to the content of the pages on which they appear as the search progresses. The closer to the goal of the search for the ad, the more narrowed the search becomes. Banner ads may be expected to do the same: The "deeper" into the site they appear, the higher the percentage of banner ads targeted to the content of the page.
However, results reveal a low level of collating, or matching, banner ads to content with home pages and an overall increase in the percentage of banners targeted to page content after the first, second and third clicks.
In fact, only one banner ad related to BMW (a BMW dealership) was found at the point where classified ads for BMWs were displayed. It was found after the third click.
The fact that targeting banner ad content to page content dramatically drops after the third click is not surprising. According to Robinson and Kaye's (1997, p. 221) recommendation, sites that force readers past the third-click level are more difficult to navigate. In general, these sites are poorly organized.
One example that stands out was an ad for a Pontiac dealership that appeared after the sixth click - the point at which the search task was complete. In the authors' opinion, someone who is so highly motivated to find BMWs in online classified ads is not like]y to be persuaded to buy a new Pontiac, no matter how exciting the ad.
Is this poor targeting or good guerrilla placement by Pontiac? Future experimental studies will examine this issue.
Aside from apathy and ignorance, one possible reason for designing a site that ignores the 3-click rule involves click-through measurement. In an information location task, such as the search for a BMW classified ad, users are motivated to find their goals. In principle, they should continue to click-through until the goal has been reached. Poor site design is one way to inflate click-through rates.
Also, adding "extraneous" levels to the information hierarchy to boost click-through provides the opportunity to place more ads.
According to the trend reported in this pilot study, the percentage of banner ads collated with page content should continue to rise. Few banner ads were found after the fourth, fifth and sixth clicks. Had they been targeted effectively to page content, a 100 percent of ads targeted to content could have been achieved with relative ease. Therefore, online newspapers that designed their classifieds sites to force users beyond three clicks were ineffective at both taking advantage of the opportunity to place a greater number of ads and targeting banner ad content to page content.