Creative Factors in Interactive Advertising



 

 

 

 

Hyo-Gyoo Kim

 

Doctoral Student
hgkim@mail.utexas.edu
Department of Advertising
College of Communication
The University of Texas at Austin
Austin, Texas 78712




And

 

John D. Leckenby

Everett D. Collier Centennial Chair in Communication
john.leckenby@mail.utexas.edu
Department of Advertising
College of Communication
The University of Texas at Austin
Austin, Texas 78712




A paper to be presented at the

Annual Conference of the American Academy of Advertising

Jacksonville, Florida

March 2002

 

Center for Interactive Advertising

http://ciadvertising.org

 

 

This authors wish to gratefully acknowledge the assistance of Courtney Ebeier of Nielsen/NetRatings (website: http://www.nielsen-netratings.com) for her invaluable support in the collection of the data which serve as the basis of this study.

 

CREATIVE FACTORS IN INTERACTIVE ADVERTISING

 

Hyo-Gyoo Kim, University of Texas at Austin, TX

John D. Leckenby, University of Texas at Austin, TX

 

Abstract

 

This study examines the relationship of creative factors to interactive advertising by studying their connection to banner ad clicking. A total of n=243 banner ads from the first week in July 2001 were examined. Click-through rate data and creative execution data were provided to the authors by Nielsen-NetRatings specifically for this study. Creative factors are considered in two aspects: executional factors and message factors. Findings show that message factors are more highly related to click-through rates than are executional factors although both show some relationship to click-through rates. The degree of relationship of creative factors to one measure of banner ad effectiveness (click-through rate) is consistent with previous findings of studies of creative factors and effectiveness in other media. This study once again shows similarities of this new medium to behavior of various phenomena in traditional media.

 

Introduction

 

The creation of effective advertising has long been one of the main concerns of both academicians and advertising practitioners. A considerable amount of research regarding the effects of executional factors of print advertising dates back a century. Historically, research by several psychologists and advertising practitioners was conducted to examine the effects of type of appeal, size, color, illustration, layout, copy treatment, and so on (for example, Starch, 1923). Their accumulated research results are still effectively used by contemporary practitioners, for example, the Starch score, and those results led later scholars to follow with more diverse studies of print advertising. Among them, a well-known study was done in 1968 by Diamond who analyzed more than 1,000 advertisements in Life magazine and the results of that work suggested that a sizable proportion of the observed variation with print advertisements could be accounted for by simple executional factors.

Compared to print advertisements, the study of television commercials has been somewhat limited in terms of creative analysis. The amounts of time and effort to conduct TV commercial research have prevented more diverse analysis. The most popular large-scale television commercial study was conducted by Stewart and Furse in 1984. They examined more than 1,000 TV commercials to answer the question, “What advertising executional devices influence the effectiveness of a television commercial?” While they provided insightful understanding of TV commercials in terms of execution, until now, the previous question with respect to the area of print ads, is still raised, both in academia and industry. While analysts in both areas have poured their efforts into giving more accurate answers to this question, they have approached this from slightly different angles. Academicians tend to focus more on the issue of advertising appeals or on a number of executional devices, such as humor, sex, and celebrity presence. But those research studies have not been generally useful for the advertising manager, because of their limited numbers of analyzed advertisements. Advertising agencies have to make decisions about advertisements, based on the aggregated response of a large target audience. On the other hand, as Stewart and Furse (1986) noted, industry studies have generally not been well documented and reported with respect to research design, methodology, and analysis, partly because of the intervention of the proprietary nature of these studies. Neither area has provided conclusive answers to the above question so far; nonetheless, a number of research studies do contribute to understanding and creating more effective advertising.

While much is known about the influence of executional or content factors on the performance of print and broadcasting advertising, far less is known about banner advertising. When we think about the development of the Internet, it is somewhat amazing that no major studies comparable to those for print or broadcasting advertising have been carried out for banner ads until now.

After the advent of the banner ads at Hotwired on the Internet for the first time, in 1994 (Marketing Week, 2000), this newly developed advertising format was enthusiastically welcomed by advertisers for several reasons. First, they had a totally new medium in which to place an ad, and they saw that the expansion of the Internet was faster than for any other traditional medium. Second, while the traditional media offered the ad in various forms, banner ads not only provide persuasive information but also offer the possibility of being a transaction medium themselves. Finally, advertising people can obtain considerably concrete behavior data from banner ads more easily than is the case with traditional media. In terms of the hierarchy of the advertising process, the banner click-through is considered, at least, to be in the middle of cognition or attitude change and behavior, or as a (part of) behavior.

Concurrent with the development of the Internet, banner ads seemed to be developing at a similar rate of growth. But according to a recent report (Nielsen-NetRatings, 2001), the average banner ad click rate has dropped to a negligible level, less than one percent. Worried about the banner’s performance in terms of click rate, advertising people are struggling to resolve this problem.

This study is in line with current understanding of banner ads, which appears to be increasingly concerned with the creative factors in relation to the promotion of brand equity. Mainly, the focus will be on the relationships that may exist between a set of creative variables and the measure of advertising performance of banner ads.

This study sought answers to the following questions:

1. How much banner advertising performance is attributable to creative variables?

2. Where is the impact of a particular variable manifest?

Additionally, we are in the position of knowing little about the roles of executional features versus message content in determining the impact of the banner ads. It is also unclear whether or not the findings of executional research involving traditional advertising are adaptable to the Internet banner advertising. These issues will also be addressed in this study.

 

Advertising Creativity Study

 

The banner ad is a mixture of the ingredients inherent in print and TV commercials. Therefore, a review of the typologies previously applied to both media will be the starting point since there does not exist a unique typology for banner ads.

 

Typology of Advertising Creativity

The criteria for the analysis itself are the most critical issue in any kind of measurement. Several advertising researchers developed various creative coding systems to measure advertising effectiveness. As noted in the previous section, there has been a rich research tradition relating print advertisement characteristics to dependent measures, such as readership or recall. Most of these studies have used ‘Starch scores’ as the dependent variable and found that the advertising effects were dependent on the content and execution of the advertising (Diamond, 1968; Lutz and Lutz, 1977; Starch, 1966). Among them, Diamond’s study pioneered in terms of its scope and methodology. He investigated print ad considering two elements: content and format.

Several more advertising formation effects studies, in terms of ‘attitudes’ and ‘beliefs’ about the advertised brand, were conducted by Mitchell and Olson (1981) and Edell and Staelin (1983). In these studies, they found that the attitude toward the brand differed according to the picture size and pictorial information in the advertisement.

In the television commercial study area, Resnik and Stern (1977) used 14 informative dimensions and following this study, Aaker and Norris (1982) investigated the level of informativeness of commercials by product class. The most frequently cited large-scale study was Stewart and Furse’s original study (1984), and a replication of this study was done by Stewart and Koslow (1989). They used more than 150 executional variables to analyze TV commercials.

Most of the above studies focused on the role of executional variables. However, the ad is generally considered to have two elements: (1) content, the message contained in the ad, and (2) format, those ad attributes that attract the consumers’ attention. And these two elements are not totally independent of each other. Holbrook and Lehmann (1980) developed a coding system for both message and mechanical factors in print advertising.

Like Holbrook and Lehmann, the dichotomous typology is common in the analysis of advertising creativity. Popular dichotomous typology is informational and emotional appeals in advertising analysis; Vaughn (1980) considered this thinking/feeling dichotomy a situational variable and combined it with level of involvement to propose a model of four creative strategies. Similarly, Simon’s (1971) classification of creative strategies used some executional and sales promotion criteria. Laskey, Day and Crask’s (1989) typology, based on Frazer’s framework (1983), first coded an ad into the informational/rational or the image/emotional category and then into a specific strategy.

 

 

Typology Development

 

Data Used

The information regarding banner ads which were placed on the Web in the first week of July 2001 (7/1-7/7) was provided by Nielsen-NetRatings. The provided data set includes each individual banner’s creative treatment and the click-through rate for each of them. The number of the placed banner ads was a total of 267,363 in the given days, but only 807 could be reported because the rest of them did not meet the minimum sample size standards by Nielsen-NetRatings. Among the available 807 banner ads, 30% of them (243) were randomly selected with the SPSS random data selection technique and analyzed in this study.

 

Typology Development

Both of the classification systems for message and executional factors were derived by adaptation from several previous typology systems. The goal was to identify a set of mutually exclusive and exhaustive categories that reflect the nature of the banner ads. Developing the typology for investigating banner ads required several iterations before a final version of each typology was deemed acceptable. This procedure involved brainstorming to develop different sets of potential classification categories and finally applying these categories to actual banner advertisement claims.

The measure of overall and specific strategies basically adapted from the Stewart and Furse (1986) typology. Even though they did not clearly separate the notion of the ad into executional and message strategies, a similar notion already existed among 155 variables used in their study.

 

Below are the variables for both strategies used in this study:

 1. Modified Executional Factors for Banner Ads:

1) Size of the banner, 2) Shape of the banner, 3) Animated or static banner, 4) Looped or not, 5) Number of words in banner, 6) Number of images in banner, 7) Setting is related to product or not, 8) Sentence type, 9) Computer form used, 10) Any number existence, 11) Logo existence, 12) Web address existence, and 13) Button existence.

 

2. Message Factors for Banner Ads:

1) Emotional approach or not, 2) Product benefit or not, 3) Psychological benefit or not, 4) Product identification is possible, 5) Sexual appeal, 6) Welfare appeal, 7) Safety appeal, 8) Curiosity appeal, 9) Emergency appeal, and 10) Negative appeal.

 

The study reported here used the method of content analysis. All banners were coded for executional and message factor devices. Two coders were recruited to code the selected banner ads. They had a one hour training session before starting to code and intercoder reliability tests using Holsti’s formula were done on a randomly selected sub-sample of 10 percent. The level of agreement did not satisfy the suggested criterion after the first session: 84 percent for execution and 78 percent for message factors, respectively. Therefore, we had one more session to increase the agreement level and the sample was randomly split into halves and coded.

In order to assess the ability of executional and message variables, multiple regression will be performed. Then follows the factor analysis because of the concern of the multicollinearity within the executional and message variables, and the factor scores will be used as an independent variable in another set of regression to predict banner ad click-through rate.

 

Research Questions

 

As a part of the exploration of the nature of the banner ad, this study will ask: “How much will the banner ads’ creativity explain the banner ads’ click-through rate?”

However, in terms of the ad formation perspective, the banner ads are generally considered to be the combination of print and TV commercials. Further, the banner ad has relatively small property in itself, both in space and time, compared to the traditional advertising. This is a weakness for the basic advertising purpose--the providing information or message. Therefore, the execution strategy might be simpler than the older ones.

RQ 1a: In terms of executional format, the explanation power will be limited, compared to the traditional ads.

However, there is no reason to believe that the adaptation of message strategy will be limited in banner ads due to the spatial and time limitations.

RQ 1b: In terms of the message factors, the banner has the same amount of explanation power to predict the ad performance as the traditional ads.

Overall, RQ 1c: The message factors will perform to predict the banner ads’ click-through rate better than will the executional factors.

 

Results

 

Frequency Distribution

Almost two-thirds of the displayed banner ads are animated full banners (468 x 60 IMU, Internet Advertising Bureau) and relatively more words are used, compared to street banner ads off-line. More than half included non-declarative sentence types in their ads and displayed a ‘clickable button’ type to lead users to act. The advertisers also tried to relate the advertising setting and the product or service in the ads.

In terms of message variables, most of the banners are clear about which product or service they are selling--at least product type--and they used more ‘product benefit approaches’ than ‘psychological benefit approaches.’ ‘Sexual appeal’ and ‘negative appeal’ were rarely used in Internet banner ads. Table 1 and Table 2 show frequency distribution of analyzed banner ads in two dimensions; execution and message.

 

Table 1. Frequency Distribution: Executional Variables   

 

Level (Frequency / %), N=243

 

0

1

2

3

 1. Size of Banner

N/A

Smaller (25/10.3)

Full (185/76.1)

Larger (33/13.6)

 2. Shape of Banner

Full (190/78.2)

Others (53/21.8)

N/A

N/A

 3. Animated Banner

No (75/30.9)

Yes (168/69.1)

N/A

N/A

 4. Looped or not

No (125/51.4)

Yes (118/48.6)

N/A

N/A

 5. Number of Words

N/A

Less than 8 (46/18.9)

8 to 14 (106/43.6)

More than 14 (91/37.4)

 6. Number of Images

N/A

One image (164/67.5)

More than one (79/32.5)

N/A

 7. Relevant Setting

N/A

Not Relevant (84/34.6)

Somehow (91/37.4)

Directly (68/28.0)

 8. Sentence Type

Declarative (92/37.9)

Other Types (151/62.1)

N/A

N/A

 9. Computer Form

No (188/77.4)

Yes (55/22.6)

N/A

N/A

10. Any Number

No (155/63.8)

Yes (88/36.2)

N/A

N/A

11. Logo

No (72/29.6)

Yes (171/70.4)

N/A

N/A

12. Web Address

No (179/73.7)

Yes (64/26.3)

N/A

N/A

13. Clickable Button

No (87/35.8)

Yes (156/64.2)

N/A

N/A

 

 

            Table 2. Frequency Distribution: Message Variables   

 

Frequency (%), N=243

 

No

Yes

 1. Emotional Approach

         157 (64.6)

          86 (35.4)

 2. Product Benefit Approach

           93 (38.3)

        150 (61.7)

 3. Psychological Benefit Approach

         168 (69.1)

          75 (30.9)

 4. Product Identification

           19   (7.8)

        224 (92.2)

 5. Sexual Appeal

         233 (95.9)

          10   (4.1)

 6. Welfare Appeal

         167 (68.7)

          76 (31.3)

 7. Safety Appeal

         224 (92.2)

          19   (7.8)

 8. Curiosity Appeal

         186 (76.5)

          57 (23.5)

 9. Emergency Appeal

         196 (80.7)

          47 (19.3)

10. Negative Approach

         237 (97.5)

            6  (2.5)

 

Regression Analysis

The first phase of analysis consisted of analyzing each of the two sets of creative factors—message and execution—against the banner click-through rate. Because of the comparatively small number of observations and the correlations among the predictor variables, the precise estimation of the effects of these variables on banner click-through rate will be limited. Since the major focus of this research is to establish whether message and executional variables significantly relate to banner click rate, the interpretation of each estimate of the effects of individual variables requires caution.

The result of regression analysis shows that the executional variables predicted the banner click-through rate at a minimal level (adjusted R2 = .04). Compared to the previously reported studies, this explanation level is almost the same as Stewart and Furse (1984), but far below Holbrook and Lehmann (1980). However, the interpretation of this number requires caution because those studies measured different things in a different medium. Among thirteen inserted variables, four are statistically significant at a probability level of .05: ‘shape of the banner (standardized coefficient: -.17),’ ‘degree of advertisement setting relevance with the product or service (standardized coefficient: .13),’ ‘number existence (standardized coefficient: .13)’ and ‘logo existence (standardized coefficient: -.16).’ 

Sequential regression for the message variables predicted the banner click-through rate fairly well, compared to the executional variables (adjusted R2 = .17). This figure is better than in Stewart and Furse’s study (1984) and comparable to Holbrook and Lehmann’s study (1980). Here, we can see that the ‘emotional approach (standardized coefficient: .22)’ performs better than the ‘rational approach’ in message strategy in a banner ad. This result may require some re-thinking of the Internet medium. Generally, the Internet is considered to be an information-oriented medium, but this result shows the opposite, even though it was found from banner ads, not from the whole Internet. Still, ‘sexual appeal (standardized coefficient: .14)’ was an effective way to set the message strategy on the Internet, even though the usage is not easily found. And the ‘welfare or free (standardized coefficient: .14)’ or ‘emergency or now (standardized coefficient: .16)’ notion was effective and the ‘problem-solving or safety (standardized coefficient:  .16)’ approach also had some effects on banner ads’ click rate.

 Because the two sets of predictor variables (execution and message) are potentially interrelated, this study performed a step-wise regression with both types of predictor variables included together. The results indicate that only one variable from the executional and five variables from message variables passed the statistical significance test. Comparing the beta coefficients, the ‘Emotional Appeal’ held the greatest power to predict the banner click-through rate; this was followed by the ‘Ad Shape’ and ‘Emergency Appeal’ in terms of weight. The overall predictive power of the equation increased slightly (adjusted R2 = .20), compared to when the two sets of variables were used separately.

  

Factor Analysis

At this point, concern about the possible effects of multicollinearity within the two variable sets on the individual coefficients required investigating the interrelations among the message and executional variables. Separate principal component factor analyses of the executional and message variables with the Varimax-rotated method were performed. The six executional and four message factors with eigenvalues greater than 1.0 accounted for 68.7 percent and 57.1 percent of the variance, respectively, and the judgmentally given factor names are listed in Table 3. This result also indicates that the developed and analyzed variables in this study can be reduced into reduced number. Each executional and message dimensions can be considered as the composition of six and four dimensions instead of thirteen and ten variables, respectively. 

 

Table 3. Subtracted Factors

 

Factor

Executional factors

Message Factors

I

Number of Elements

Emotional

II

Animation

Curiosity

III

Computer Form

Arousal

IV

Identification

Free

V

Product Relevance

N/A

VI

Sentence Type

N/A

 

When the banner click-through rates were regressed with subtracted executional factors, significant contributions appeared for ‘identification (Web address or product logo appeared in banner ads)’ in predicting click-through rates. Interestingly, when consumers identified the advertised product or Web address, they did not click those banners. The reason why is not clear at this moment, but the following assumption is possible: People already know what will be going on after clicking the banner ads or they have enough information about the advertised product; they do not have to click, because of their diminished curiosity. But the overall predictive power with these factors was very low (adjusted R2 = .02).

With the message factors, the results indicate significant effects of the ‘emotional approach,’ ‘curious approach,’ and ‘free offer or gift giving’ on banner click-through rates. Overall, the message factors performed far better than the executional factors in predicting click-through rate (adjusted R2 = .14 versus .02). Here, the initial question as to whether message factors can predict banner ad click-through rate over executional variables seems verified.

Additionally, when all the subtracted factors are input together in the regression model to predict the click-through rate, as shown in Table 4, one factor of execution and three factors of message account for 15 percent of the click-through rate variance. Between the two, message strategy performs better than executional strategy in terms of their relative power to predict banner click rate. The two most useful predictors were ‘curiosity’ and ‘emotional message approaches.’

 

Table 4. Unstandardized and Standardized Coefficients of the Executional and Message Factors. (Displayed for statistically significant variables only)

 

 

Banner Click - Through Rate

 

Unstandardized

Coefficients

Standardized

Coefficients

t

Execution: (IV) Identification

            -.98

          -.13

         -2.13*

Message:   (I) Emotional

            2.61

           .20

          3.10**

                  (II) Curiosity

            3.41

           .30

          4.96**

                  (IV) Free

            1.00

           .17

          2.69**

[* p< .05, ** p< .01; R2 (adjusted) = .15]

 

Summary and Discussion

 

This study has identified major differences and similarities at the same time among advertisements which are placed in the Internet medium and in other media, with respect to the quantity of used cues, appeals and so on.

The results of this study are not entirely comparable with those from previous work of a similar nature in traditional media, even though one of the main purposes of this study is comparison with them. For example, most of the creative element studies (Diamond, 1968; Holbrook and Lehmann, 1980) measured a different thing (i.e., Starch score or intention) in a different advertising format, e.g., print ads. Stewart and Furse (1984) measured recall, comprehension and persuasion in the TV commercials situation. Nonetheless, some tentative comparisons will be possible. In general, the coefficient of determination fell in the middle of the previously reported studies (adjusted R2 = .15 with all subtracted factors). It was low compared to some print ad studies (for example, in Holbrook and Lehmann’s study, the reported R2 ranged from .10 to .43), but it is still high compared to TV commercials (e.g., in Stewart and Furse’s case, the adjusted R2 ranged from .06 to .12). While few studies of advertising are carried out using actual ads and behavior data for existing products, and they have been criticized for the artificiality of the advertising stimuli used, this analyzed data, which was collected in the natural environment, reveals somewhat amazing effectiveness.

As revealed in Holbrook and Lehmann’s study (1980), this study shows a similar finding: The message factors perform better than the executional factors, regardless of the measured variables in both studies. This was not found in TV commercial studies but this might be due to the simplicity of the executional devices of both print and banner ads compared to TV commercials. Another explanation of why there is greater explained variance in print or on the Web than in television could relate to the very nature of the medium. At this moment, it can be said that the banner ads resemble print more than they do TV in executional terms. Another interesting interpretation is that this could provide indirect evidence to show people do not just look at but they read banner ads on the Web. The decision of whether they will click or not is determined within a very short time limit. They clicked banner ads not because of mechanical elements but because of message appeal.

This study has a number of limitations. This research examined only a limited array of potential executional and message factors, and it examined only one measure of advertising effectiveness: click-through rate. Moreover, this study cannot address why and how particular factors have their effect. Another concern is the lack of relationship between the placed ads and the contents of Web sites. People do not go to see the banner, but to see the contents on the Web. Content-dependent ad placement should be analyzed.

 

References

 

1.                  Aaker, David A. and Donald Norris (1982), “Characteristics of TV Commercials Perceived as Informative,” Journal of Advertising Research, v22, n2, p61-70.

2.                  Diamond, Daniel S. (1968), “A Quantitative Approach to Magazine Advertisement Format Selection,” Journal of Advertising Research, v5, p.376-86.

3.                  Edell, Julie A. and Richard Staelin (1983), “The Information Processing of Pictures in Print Advertisements,” Journal of Consumer Research, v10, p.45-61.

4.                  Frazer, Charles F. (1983), “Creative Strategy: A Management Perspective,” Journal of Advertising, v21, n4, p.36-41.

5.                  Holbrook, Morris D., Donald R. Lehmann (1980), “Forms vs. Content in Predicting Starch Scores,” Journal of Advertising Research, v20, i4, p.53-62.

6.                  IAB (Internet Adv