16
JOHN L. NACCARATO Liggett-Stashower, Inc. KIMBERLY A. NEUENDORF Cleveland State University Content Analysis as a Predictive Methodology: Recall, Readership, and Evaiuations of Business-to-Business Print Advertising This article calls for the application of content analytic techniques to advertising as a method of predicting advertising effectiveness. A comprehensive empirical investigation examines the effect of both form variables (e.g., headline size, use of color, illustration placement) and content variables (e.g., subject matter, use of humor, use of fear appeals) on recall, readership, and evaluations in the context of business-to-business print advertising. The prediction of four different outcome variables is successful, with total variance accounted for ranging from 12 percent to 59 percent. Significant predictors vary substantially across the dependent indicators, indicating that different advertisement characteristics are likely to be needed to achieve various advertiser goals. THE ULTIMATE GOAL OF ADVERTISING is sales. As the dean of advertising David Ogilvy notes: "I do not regard advertising as entertainment or an art form, but as a medium of information. When I write an advertisement, I don't want you to tell me that you find it 'creative.' I want you to find it so interesting that you buy the product" (Ogilvy, 1983), ADVERTISING SUCCESS The direct linking of sales to advertising exposure is rarely validated in practice. Even the older, clas- sic models of advertising and marketing (e.g., "DAGMAR") have acknowledged the role of in- terrttediary processes and states (Olshavksky, 1994), including but not limited to knowledge, [positive] affect, and behavioral intention (cf., Ajzen and Fishbein, 1980). Correspondingly, and for reasons of practicality and comparability of criteria, adver- tising readership studies are viewed as the basic too! for assessing advertising effectiveness. Readership is probably the most frequently used indicator of advertising effectiveness. Unlike inquiry reports—which count how many readers request additional information and are a mainstay of business-to-business advertising—readership studies ask a representative sample of respondents whether or not they saw the advertisement, if they read it, and perhaps how much of the advertise- ment they remember seeing. Sometimes referred to as recognition or recall studies', readership studies have traditionally been considered to be a valid measure of whether or not the advertiser's message has reached the receivers. Readership studies have been conducted on a continuing basis for print media since the 1920s (Hendon, 1973). There are a number of indepen- dent organizations conducting readership studies (e.g.. Starch, Ad-Q, Ad-Chart, and Harvey), plus a host of publisher-sponsored readership services seeking to provide advertisers with information about advertising placement. However, reader- ship studies have come under certain criticisms in the past two decades (Edmonston, 1995; Johnson, 1982; Rothschild, 1987; Schaefer, 1989; Sekely and Blakney, 1994; Whipple and McManamon, 1992; Wood, 1989). Additionally, some industry observ- ers have noted the paucity of syndicated reader- ship research for industrial or business-to-business advertising (Morelli, 1986). What makes a consumer read a given advertise- ment? An early Ogilvy pronouncement declared that "Every advertisement must tell the whole sales story.., Every word in the copy must count" (Ogilvy, 1986). Images, color, and layout factors are also of great concern in the industry (Roman and Maas, 1992). A careful examination of adver- 'The terms recognition and recall are usually intended to refer to aided and umided recall, respectively: many commercial services cm>er both. In llw case of busmess-to-business adivrtising, recognition/aided recall is perhaps the more salient crilerioti. since company or product im is often a fmidamental pre-aales-catl goal. M a y . J u n e 1 9 9 8 JDUHURL OF HDUERTISinG RESEflflCtl 19

Content Analysis as a Predictive Methodology: Recall

  • Upload
    others

  • View
    2

  • Download
    0

Embed Size (px)

Citation preview

JOHN L. NACCARATO

Liggett-Stashower, Inc.

KIMBERLY A. NEUENDORF

Cleveland State

University

Content Analysis as a Predictive Methodology:

Recall, Readership, and Evaiuations of

Business-to-Business Print Advertising

This article calls for the application of content analytic techniques to advertising as amethod of predicting advertising effectiveness. A comprehensive empiricalinvestigation examines the effect of both form variables (e.g., headline size, use ofcolor, illustration placement) and content variables (e.g., subject matter, use ofhumor, use of fear appeals) on recall, readership, and evaluations in the context ofbusiness-to-business print advertising. The prediction of four different outcomevariables is successful, with total variance accounted for ranging from 12 percent to59 percent. Significant predictors vary substantially across the dependent indicators,indicating that different advertisement characteristics are likely to be needed toachieve various advertiser goals.

THE ULTIMATE GOAL OF ADVERTISING is sales. As the

dean of advertising David Ogilvy notes: "I do not

regard advertising as entertainment or an art form,

but as a medium of information. When I write an

advertisement, I don't want you to tell me that you

find it 'creative.' I want you to find it so interesting

that you buy the product" (Ogilvy, 1983),

ADVERTISING SUCCESS

The direct linking of sales to advertising exposure

is rarely validated in practice. Even the older, clas-

sic models of advertising and marketing (e.g.,

"DAGMAR") have acknowledged the role of in-

terrttediary processes and states (Olshavksky, 1994),

including but not limited to knowledge, [positive]

affect, and behavioral intention (cf., Ajzen and

Fishbein, 1980). Correspondingly, and for reasons

of practicality and comparability of criteria, adver-

tising readership studies are viewed as the basic

too! for assessing advertising effectiveness.

Readership is probably the most frequently

used indicator of advertising effectiveness. Unlike

inquiry reports—which count how many readers

request additional information and are a mainstay

of business-to-business advertising—readership

studies ask a representative sample of respondents

whether or not they saw the advertisement, if they

read it, and perhaps how much of the advertise-

ment they remember seeing. Sometimes referred

to as recognition or recall studies', readership

studies have traditionally been considered to be a

valid measure of whether or not the advertiser'smessage has reached the receivers.

Readership studies have been conducted on acontinuing basis for print media since the 1920s(Hendon, 1973). There are a number of indepen-dent organizations conducting readership studies(e.g.. Starch, Ad-Q, Ad-Chart, and Harvey), plus ahost of publisher-sponsored readership servicesseeking to provide advertisers with informationabout advertising placement. However, reader-ship studies have come under certain criticisms inthe past two decades (Edmonston, 1995; Johnson,1982; Rothschild, 1987; Schaefer, 1989; Sekely andBlakney, 1994; Whipple and McManamon, 1992;Wood, 1989). Additionally, some industry observ-ers have noted the paucity of syndicated reader-ship research for industrial or business-to-businessadvertising (Morelli, 1986).

What makes a consumer read a given advertise-ment? An early Ogilvy pronouncement declaredthat "Every advertisement must tell the wholesales story.., Every word in the copy must count"(Ogilvy, 1986). Images, color, and layout factorsare also of great concern in the industry (Romanand Maas, 1992). A careful examination of adver-

'The terms recognition and recall are usually intended to refer to

aided and umided recall, respectively: many commercial services cm>er

both. In llw case of busmess-to-business adivrtising, recognition/aided

recall is perhaps the more salient crilerioti. since company or product

im is often a fmidamental pre-aales-catl goal.

M a y . June 1 9 9 8 JDUHURL OF HDUERTISinG RESEflflCtl 1 9

B-tO-B PRINT ADVERTISING

tising content may shed light on the "sales

story." The research exemplar reported

here attempts to develop a practical

schema applicable to a range of advertise-

ment types, focusing on providing a link-

ing mechanism betv 'een the production of

an advertisement and its positive recep-

tion by consumers. The chosen exemplar

examines business-to-business advertis-

ing in a trade magazine.

Advertisement characteristics and

advertising success

The question ot which advertisement char-acteristics lead to greater recall, readership,and other goals of advertising is under-studied. Against the advice of Ogilvy andothers, agencies often rely on creativecompetitions to index the content and per-suasive potential of their advertisements,but results of these competitions may bearlittle reiationship to the success of the ad-vertisement, since creative judges are pri-marily the advertisers' professional peersand not representative of the ranks of themessage targets. Nevertheless, the "con-ventional wisdom" concerning successfuladvertisement creation is a powerful andoften highly valid force (OgOvy, 1983; Ro-man and Maas, 1992; Schultz, Tannen-baum, and Allison, 1996).

While it may seem manifestly beneficialto designers of advertising to know whatgets the reader's attention, the bulk ofsuch research has been left in the hands ofacademics (e.g., Laskey, Fox, and Crask,1994; Tellis, 1994). This research frequentlytakes the form of an experiment or fieldexperiment (Gelb, Hong, and Zinkhan,1985), manipulating such variables assource credibility, the use of an appeal suchas humor, or the presence of visual imag-ery or music. Another, type of research hasbeen the emergent single-source study,which links a household's potential adver-tising exposure to actual household buy-ing behavior (Maloney, 1994; Tellis, 1994).

The typical experimental investigation

deals with one variable in isolation from

others and tests fairly abstract outcomes

(e.g., positive affect toward a spokesper-

son in the advertisement) on nonrepresen-

tative samples. The average single-source

study fails to establish audience exposure

to an advertisement and does not even

consider advertisement characteristics.

Other, complementary studies are needed

to provide practicality of prediction and

generalization. Research studies that

probe naturally occurring variations in

message characteristics include those that

content analyze.

Content analysis as a descriptive and

predictive tool

Content analysis may be defined as thesystematic, objective^, quantitative analy-sis of message characteristics. The tech-nique was initiated by communication, so-ciology, and journalism scholars some 50years ago (Berelson, 1952) and has gainedvalidation as a research tool in thousandsof studies examining messages rangingfrom television beer commercials to newsitems on the Greenhouse Effect to pub-lished Republican and Democratic Partyplatforms (Fan, 1988; Krippendorff, 1980;Weber, 1990). There has been a recogni-tion of the difference between form vari-ables, those that are linked to the formalfeatures of the medium and cannot en-dure transfer to another media modality,and content or substance variables, thosethat may exist independent of the me-dium^ (Berelson, 1952; Holbrook and Leh-mann, 1980; Huston and Wright, 1983).

^While objectivity is tlie acknowledged goat of such a social

scientific method, it is recognized that what is actually

achieved is more properly termed "intersubjectivity."

''Typical of icime publications in the advertising literature.

Rossiter <19SV termed form and content z'ariables "me-

chanical" and "message" variables, respectively.

Most prior examinations of advertisingcontent have analyzed the compositionalform variables of print advertisements,e.g., characteristics of headlines, graphics,and copy, attempting to develop formulasfor successful print advertisements.'While there is near-consensus that use ofcolor and large advertisement size arepositive contributors to readership (Hans-sens and Weitz, 1980; Marney, 1985;Standen, 1989; Twedt, 1952; VandenBerghand Reid, 1980), the evidence about otherform variables is decidedly mixed (Assael,Kofron, and Burgi, 1967; Reid, Rotfeld,and Barnes, 1984).

Various content variables such as thesubject of the advertisement or the ap-proach to the subject (e.g., use of humor,fear, puffery, celebrity endorsement, mes-sage complexity) have been analyzed, andthe conclusions presented are also quiteambiguous (Aaker and Norris, 1982;Chamblee et al., 1993; Holman andHecker, 1983). For example, the typicalstudy on the use of humor in advertisingconcludes that "sometimes it works,sometimes it doesn't" (Gelb and Pickett,1983; Madden and Weinberger, 1984;Markiewicz, 1974). When form and con-tent variables are directly compared, themore mecharucal form variables prove tobe much more important predictors ofreadership and recall (e.g., Hotbrook andLehmann, 1980). Importantly, the con-tent/style variables that have been most

''Studies of compositional form variables in print advertise-

ments include: Ne^vspaper Advertising Bureau, 1986; 1989;

Soley and Reid. 1983a: 1983b; Standen. 1989; Wesson,

7989: Wesson and Stewart. 1987. There have also been

studies of other form variables—advertisement size, posi-

tion in tbe publication, use of color, etc.. (Assael, Kofron,

and Burgi. 1967; Industrial Equipment News, 7979;

Marney. 1985; Sales & Marketing Digest, 1988; Market-

ing News, 1987: Stuhlfiiut. 1983).

2 0 JOUmiflL OF HDUERTISIOG RESEfmCH May . June 1 9 9 8

B-to-B PRINT ADVERTISING

often studied in the realm of consumer ad-vertising do not generally apply to indus-Iriti! or business-to-business advertise-ments (e.g., celebrity endorser, sex appeals).

Most extant content-analytic studieshcive neglected a comprehensive coverageof potential important predictive vari-ables, opting instead to look at just one ora handful of variable{s).' Those studiesthat have made the attempt at comprehen-siveness warrant mention.

Holbrook and Lehmann (1980) tapped48 message and mechanical variables, pre-dicting over 30 percent of the variance inStarch readership scores for Neivsweek andSports Illustrated. Their most importantpredictors included product class and thevague construct, "creativity." A majorcontribution of this study was its clearfinding that both form and content factorsare important in producing recall andleadership.

The most comprehensive projects todate are studies of television commercials.Stewart and Furse (1986) developed a 151-itcm content-analysis scheme, which theyrelated to measures of recall, comprehen-sion, and persuasiveness for 1,059 spots.They found both recall and persuasion tobe influenced by (a) brand performancecharacteristics (e.g., a brand-differentiatedmessage, convenience of product use),and (b) attention and memory factors(e.g., humor, mnemonic devices, front-end impact, bmnd sign-offs). Gagnardand Morris (1988) content analyzed 121CLIO award-winning commercials withcin adaptation of the Stewart and Fursescheme. They found a unique set of char-acteristics common to the award-winning

'Vor example, Rossiter 11981) examined the impact of 13

"sf/ntax" variables ort Starch readership scores Jbr adver-

li^nneuls in one issue of Newsweek. Although Iw ex-

I'tnined an impressive amount of variance, he looked exclu-

•^iivly at picture size and headline characteristics.

spots: the use of male characters and fewminorities, animals, or children; omni-present music; the use of humor; and theuse of a strong front-end impact, for ex-ample. Most of the variables employed bythese researchers are inapplicable to printadvertisements.

Only one major attempt has been madeat identifying a coniprehevsive set of printadvertisement characteristics that contrib-ute to readership. During the late 1970sand early 1980s, David P. Forsyth, vice-president of research at McGraw-HillPublicatior;s, analyzed nearly 3,600 printadvertisements covering a five-year spanof McGraw-Hill's Ad Sell PerformanceStudies (reported in Donath, 1982, andWood, 1989). His research found "signifi-cant" contributors to the advertisementbeing noticed included use of color anduse of a spread or bleed format. Contribu-tors to "creating awareness," "arousinginterest," and "building preference" werelong copy (>300 words), use of tables orcharts, and showing the product by itself.However, the McGraw-Hill project waslimited to mechanical form variables; noattempt was made to measure such con-tent characteristics as product type or per-suasive appeals. The readership studiesreported small sample sizes, and, unfortu-nately, available reports on the project failto supply sufficient detail to evaluatethe content-analysis methodology. Thisleaves hollow the claims of "significant"findings.

Indeed, very few of the studies re-viewed used all methodological standardsrecommended in the content-analysis lit-erature (Krippendorff, 1980; Riffe and Fre-itag, 1997). Flawed methodology is one

*For example, Rossiter (1981) properly reported intcrcoder

reliabilities and dropped variables that did not meet a siff

criterion (rho - .60) but luid a very poor, nonprobabiliti/

sample, making inference impossible. Holbrook und Lch-

potential reason for the wide variation in

findings across content analyses for both

content and form variables. Other possible

reasons include a failure to identify criti-

cal variables in a comprehensive fashion

and context specificity (e.g., what is im-

portant to the success of an advertisement

in a general interest magazine may not be

the same as the set of elements that lead to

success in business-to-business advertising).

This study

There is a growing recognition that therules of good quantitative methodologyought to apply to analyses of messagecontent (Krippendorff, 1980; Neuendorf,1998; Riffe and Freitag, 1997; Zollars,1994). The study described here has beenconducted with care given to content-analytic standards. Sampling was system-atic random. The sample size was ad-equate to support a large number of pre-dictor variables. Coder training waslengthy and rigorous. Variables notachieving an acceptable level of reliabilitywere dropped from final analyses.

Neuendorf (1998) proposes the integra-tive model of content analysis, whereinmessage-centered variables tapped bycontent analysis are linked with audience-centered variables or source-centeredvariables measured in additional data col-lections. The study described here followsthat model by linking content analysisto readership studies. Attempts to linkcontent and form measures to recall/readership began in the 1950s (Twedt,

mann'suseofCronbach'salphaasan indicator of reliability

(s suspect. They also used a very limited, nonprobability

sample and complained about coders becoming "exhausted"

(p. 55) after only 10 hours of coding. The first study of

advertising content as related to readership (Twedt, 1952)

gave no description of its amtent-anali/sis methodohgy at

ail.

M a y . June 1 9 9 8 JQUHOHL OF HflUERTISinG HESEHRCH 2 1

B-to-B PRINT ADVERTISING

1952) and continued intermittentlythroughout the 1970s and 1980s, as indi-cated in the above review. But, there is along gap in the literature after the mid1980s. This study updates and continuesthe quest, with a call for more rigorousresearch standards.

For logistic reasons, and in order toeliminate confounding factors and "mask-ing" effects of uncontrolled context vari-ables, this study has examined business-to-business advertisements in one particu-lar publication. We have gone for depthover breadth.

This research is guided by a pair of gen-eral research questions:

RQs: To what extent may form and con-tent attributes of print advertise-ments predict critical outcome vari-ables such as readership, recall, andperceptions of the advertisement(when limited to one particular typeof message pool and receiver type)?Are significant predictors differentacross the outcome variables?

METHODOLOGY

The publication and PARR reports

The focus of this study is on both formand content variables as applied in indus-trial or business-to-business trade publica-tion advertisements. Specifically, eight is-sues of Electric Light & Poiver {EL&P)

magazine were randomly chosen foranalysis. All advertisements in these is-sues were included in the analysis.

EL&P is published by PennWell Pub-lishing Company. It is a tabloid-sizemonthly news magazine aimed at man-agement, engineering, operating, and pur-chasing personnel in all segments of theelectric utility industry. For the advertise-ments studied here, audience data wereobtained from PennWell Advertising

Readership Research (PARR) Reports,

conducted by PennWell at no charge to

provide advertisers with a means to mea-

sure, evaluate, and compare the reader-

ship of and response to their advertising.

The PARR surveys asked the following

questions:

1. Did the reader notice the advertisement?

2. If the reader noticed the advertisement,how much of it was read?

3. What was the reaction to the advertise-ment?

a. informative

b. attractive, attention-getting

These PARR surveys were conductedby mail. Approximately three weeks afterthe regular mailing of the issue, a randomsample of readers received a duplicate is-sue. In an enclosed letter, readers wereasked to go through the issue again andanswer the questions attached to each ad-vertisement.^ The representative samplediffered by studied issue, ranging in sizefrom 200 to 700; response rates rangedfrom 10 percent to 50 percent. The publi-cation's circulation is audited by the Busi-ness Publication Audit (BPA) bureau,which indicates its readership as com-posed largely of electric utility managers,supervisors, and consultants.*^

^This classifit- I'/c I'ARR Reports as aided recall research,

in that respondents hai'e the opportunity to vie^v the adi'er-

lisemenls.

M summary of the BPA sfatemcit lists rcatlers us: "Gen-

eral and corporate management, including financial and

lutministrativc, engineering managetncnt and supervision,

engineers, including planning, design, performance, R&D.

operations management ami supen'ision; Ofn'ratkins. in-

cluding construction, maintenance and fleet, purchasing,

commercial marketing, customer service, other qualified

functions" (SRD5, 19%, p. 505).

Coding and analysis

The codebook developed for the contentanalysis provides measures of constructsselected for their potential predictivevalue when correlated with readershipscores from the PARR Reports. This com-prehensive pool of measured variableswas generated from (a) a review of pastresearch and professional guidelines, and(b) a careful examination of idiosyncrasiesof business-to-business advertisements-The full codebook contains a detailed defi-nition of each of the 190 measured vari-ables and each category within the vari-able. The pool of variables was reduced to75 for final inclusion in analyses, via com-bining variables and eliminating variableswith low reliability'^ or extremely lowvariance.

Each construct is classified as either aform construct (pertinent to the vehicle,i.e., print magazine) or content construct(relative to the subject matter and presen-tation). A list of form and content vari-ables as used in the final analysis is pre-sented in Appendix A (including reliabil-ity figures). Coding was conducted by ateam of four trained coders. Coding as-signments were made randomly based ona total sample size of 247 readership-studied advertisements from the eight is-sues of EL&P. Average intercoder reli-abilities were calculated prior to the initia-tion of coding and again with a 10 percentsubset of the final data set.

The final analyses utilized the 54 formand 21 content independent variables

^Numerous variables were measured as they occurred in la)

the headline, Ih) the visuals, and/or (c) ttu copy. Due to lou-

frequencies of occurrence, these applications were collapsed

across the three before inclusion in the regression analyses.

Additionally, variables ivith reliability coefficients below 60

percent or r = .70 were dropped.

2 2 JQUHnHL OF HDOERTISinG HESEHHCH M a y • June

B-to-B PRINT ADVERTISING

TABLE 1Stepwise Prediction of Aided Advertisement Recall

Independent Variable

Form variables

Fractional page

Junior page

Tabloid spread

Color

Copy in bottom half

Copy in right half

Major visual chart/graph

Average size of subvisuals

Content variables

Service advertised

Total ff = .59; Adjusted Ff^ = .58

F(9.233) = 37.83; Sig. = .0001

Pearson r

-.53

-.15

.36

.48

.10

-.04

-.12

.18

.18

Reliabiiity

{% or f)

96%

96%

96%

.92(0

78%

78%

75%

85-100%

84%

Frequency

19.4%

47.4%

6.9%

NA

49.8%

14.2%

1.6%

NA

20.6%

Rnal

Beta

-.47

-.34

.31

.24

.18

-.16

-.10

.09

.12

SIg.

<.OOO1*

<.OOO1*

<,0001*

<.OOO1*

.0002*

.0005*

.0140

.0336

.0059

Niitc.- NA hiiiicates the relialnlily or frequeiny h not apylkable because vnriablea in this table have been combined, averaged, or otherit'iae manipulated from the original measurefs).

'^ig. hoiiis at p < .05 using Boriferroni test (criterion - .0007) fur the final 75 independent variables entered in the imtUiple regression.

listed in Appendix A " and four depen-dent variables taken from the PARR Re-ports: Aided Advertisement Recall, Ad-vertisement Readership, Informativenessof the Advertisement, and Attractivenessof the Advertisement. A stepwise-mul tip le-regression model was developedfnr each dependent variable. Categoricalindependent variables were included viastandard procedures for dummy and

'"Wi? chose not to factor analyze the predictor set, a tech-

nique used fiy Twedt (1952) and Holbrook and Irhninnn

(1980). While a reduction in the predictor set is beneficial to

degrees of freedom and power, the collapsing of variables

also washes out individual variances and potential predic-

tive ability. Instead, zt-e included individual variables

iind employed the Boiiferroni adjustment for nniltiple

effect coding {Cohen and Cohen, 1983). In-

spection of interitem correlations for the

predictor variables and condition in-

dex/VIF coefficients revealed no signifi-

cant multicoUinearity problems.

RESULTS

Advertisement aided recall

In the prediction of Aided AdvertisementRecall, the step wise-multiple-regressionanalysis yields a total of nine predictorsfrom the list of seventy-five variables—eight form variables and one content vari-able. Table 1 displays a summary of thezero-order correlations, reliabilities, fre-quencies, final betas, and levels of signifi-cance for Aided Recall. The total R^ of .59indicates a High level of variance ex-plained by the nine predictors.

Final regression coefficients for the pre-

dictor variables for Aided Recall showfour negative predictors—fractional page(p = -.47), junior page (-.34), copy in theright half of the advertisement (-.16), anduse of a chart or graph in the major visual(-.10). Positive contributions to Aided Re-call are indicated for tabloid spread O =.31), color (.24), copy in the bottom half ofthe advertisement (.18), service as the sub-ject of the advertisement (.12), and the av-erage size of secondary visuals (.09).

Predictors relating to the size of the ad-vertisement—fractional page, junior page,Lind tabloid spread—thus provide someinteresting comparisons when all predic-tors are submitted in a regression. The fre-quencies indicate that junior pages are themost-often-used page size (47.4 percent)followed by fractional pages (19.4 per-cent). Yet, the final standardized regres-sion coefficients (betas) indicate that both

M a y . June 1998 JOUBOflL OF RDUEflTISinil Mf l f lCH 2 3

B-to-B PRINT ADVERTISING

TABLE 2Stepwise Prediction of Advertisement Readership

Independent Variable

Form variables

Subject apparent in visuals

Tabloid page

Headline in bottom left section

Content variables

Logical argument used

Fear appeal used

Total ff = .12: Adjusted ff = .10

f [5,242) = 6.37; Sig. = < .0001

Pearson r

.19

-.18

-.08

-.10

.11

Reliability

(% or r)

75%

96%

76%

60-93%

89-95%

Frequency

(%)

60.7%

17.8%

1.2%

26.3%

11.3%

Rnal

Beta

.20

-.22

-.13

-.16

.12

Sig.

.0016

.0008

.0365

.0127

.0448

have rather strong negative partial rela-tionships to Aided Recall. Both predictorsare also highly significant {p < .0001),which meets the ;) < .05 criterion and theystricter Bonferroni test (p < .0007; Hair,Anderson, Tatham, and Black, 1995) em-ployed throughout the analyses.

On the other hand, tabloid spreads havea very low frequency in this study (6.9percent) but hold the strongest positive re-lationship to Aided Recall, with a finalbeta of .31 (p < .0001). Taken as a whole,these findings indicate that large, tabloidspread advertising units are best remem-bered. Unfortunately, it also seems thatthe unit favored by advertisers in thisstudy, the junior page, is poorly recalled.And, it is only marginally better remem-bered than the much smaller, less expen-sive fractional page unit.

Not surprisingly, color is a significant(p < .0001) positive predictor of Aided Re-call." Frequencies indicate the abundantuse of color—frequency for two-color is10.1 percent, three-color 3.2 percent, and

"Color was entered in the regression as black and white =

J, two-color = 2, three-color - 3. and ftiur-color = 4,

four-color 64.0 percent, versus black andwhite at 21.9 percent.

Having copy in the bottom half of anadvertisement and not having copy in theright half of the advertisement relate togreater Recall. On the other hand, havinga major visual chart or graph, and a largeaverage size of secondary visuals areweaker predictors (negative and positive,respectively), not meeting the Bonferronicriterion.

Another notable point in Table 1 is theperformance of the only significant con-tent variable—subject of the advertise-ment as service (versus product, institu-tional, etc.). Service's reliability (84.0 per-cent) is good, its frequency (20.6 percent)is second only to product advertisements(67.6 percent), and its final beta is positive(.12).

Advertisement readership

The stepwise-multiple-regression analysisfor Advertisement Readership producestwo positive and three negative predictorvariables: subject apparent in the visuals(3 = .20), fear appeal used (.12), tabloidpage used (-.22), logical argument as anapproach/appeal to the subject (-.16), andheadline in the bottom left half of the ad-vertisement (-.13). Three are form vari-ables, while two are content variabies. Allof the Readership independent variablesmeet the p < .05 level of significance, butnone meets the stricter p = .0007 Bonfer-roni level.

Tbe summary statistics for Advertise-ment Readership are shown in Table 2. Thetotal R^ is .12, and while it is not as massiveas that for Aided Recall, it does achieve ahigh level of statistical significance.

. . these findings indicate that iarge, tabioid spread ad-

vertising units are best remembered. Unfortunateiy, it

aiso seems that the unit favored by advertisers in this

study, the junior page, is poorly recaiied.

2 4 JDURimL DP BDUEBTISinG RESEflRCH May . June 1 9 9 8

&-to-B PRINT ADVERTISING

Once again, size of the advertisementseems to be a significant predictor in theregression, with tabloid page {p = .0008)just shy of the Bonferroni criterion forReadership. Its frequency (17.8 percent)places it third behind junior page (47.4percent) and fractional page (19.4 per-cent). Readership for the tabloid pageshows a negative relationship (P - -.22)which could indicate the larger format is adetriment to readability.

The strongest positive relationship forReadership is having the subject apparentin the visuals, with a final beta of .20. Thereliability of the predictor is a respectable75 percent, and its frequency (60.7 per-cent) shows that more than half the adver-tisements depict the subject in the visuals.Placing the headline in the bottom leftportion of the advertisement shows an in-verse relationship (P = -.13) to Reader-ship. Overall, being able to divine the sub-ject of the advertisement by looking at thevisuals appears to aid readership.

Two content variables are expressed inthe Readership regression. The first, logi-cal argument as an approach/appeal (p =

-.16) has a frequency (26.3 percent) thatplaces it near the middle of the other ap-proach/appeal variables. As a predictor,logical argument is negatively related toAdvertisement Readership—even with anaudience of engineers, logical argumentdiscourages readership.

The second content variable in theReadership regression is the use of a fearappeal. Fear appeals are infrequently usedin these advertisements—at 11.3 percent itis third from the bottom among the 13 ap-proach/appeal variables submitted to theregression. Interestingly, the final beta forfear (.12) indicates it is positively relatedto Readership. Therefore, inducing fear inreaders cannot be discounted as a methodof getting them to read advertisements.

Informativeness of the advertisement

Stepwise-multiple-regression analysis forthe third dependent variable, perceivedInformativeness of the advertisement, re-sults in two negative predictors and fourpositive predictors; four are form vari-ables, and two are content variables. As inthe case of Readership, all predictors of

Informativeness meet the criterion p < .05,but none meets the stringent Bonferronitest (;; = .0007).

Table 3 shows the summary statisticsfor Informativeness of the advertisement.The total K of .20 is highly statisticallysignificant {p < .0001). Once again, adver-tisement size variables have run thegauntlet of the stepwise multiple regres-sion, this time to emerge as significantpredictors of Informativeness. Interest-ingly, the tabloid page, a large and fre-quently used format, is the strongestnegative predictor (p = -.19). The frac-tional page, a small format and frequentlyused unit, has the second highest positiverelationship (p = .16) for Informativeness.It seems these readers consider little ad-vertisements more informative than bigones.

The use of subheads and placement ofthe headline in the top half of the adver-tisement also appear to result in more in-formative advertisements. Frequencies fornumber of subheads vary from 0 to 12with a mean of just under 1 per advertise-ment. Headline in the top half of the ad-

TABLE 3Stepwise Prediction of Advertisement Informativeness

Independent Variable

Form variables

Tabloid page

Fractional page

Headline in top half

Number of subheads

Subject apparent in visuals

Content variables

Altruism appeal used

Total f^ = .20; Adjusted Ff^ = .18

F(6,24i) = 10.10; Sig. = < .0001

Pearson r

-.25

.21

.19

.24

.22

-.13

Reliability

(%or r)

96%

96%

96%

.81(0

75%

90-96%

Frequency

(%)

17.8%

19.4%

64.4%

NA

60.7%

14.6%

Rnal

Beta

-.19

.16

.15

.19

.14

-.14

Sig.

.0021

.0091

.0119

.0025

.0265

.0236

Note: NA indicates the reliability or frequency is not applicable because variables in this table have been combined, overa^^ed. or othenm-ie manipulated from the nrif^inal measiirfdi).

M a y . June 1 9 9 8 JOUROHL OF RDUeRTiSldl) flESEHHCH 2 5

8-to-B PRINT ADVERTISING

vertisement has the highest frequency of

all positions (64.4 percent). Both are posi-

tively related according to the final betas:

number of subheads (P = .19) and head-

line in the top of the advertisement

{p = .15). Having the subject apparent in

the visuals is a positive predictor (p = .14)

of Informativeness but is the least signifi-

cant of the regression predictors {p =

.0265).

Altruism is the only content variable in

the Informativeness regression. Its fre-

quency is relatively low (14.6 percent)

compared to the other 12 approach/ap-

peal constructs. It has the weakest signifi-

cance of the Informativeness predictors.

And, with a final beta of -.14, its reverse

relationship to Informativeness indicates

that appeals to altruism in an advertise-

ment are not viewed as informative by the

sample of readers.

Attractiveness of the advertisement

Table 4 summarizes the multiple regres-

sion for Attractiveness of the advertise-

ment. The overall R^ is a substantial .42,

once again highly statistically significant.

In the stepwise multiple regression for

Attractiveness, size of the advertisement

is again represented by three significant

predictors, all bearing a negative relation-

ship to Attractiveness: fractional page at

3 = -.31, followed by junior page (P =

-.28), and tabloid page (p = -.14).

Two other predictors demonstrate sig-

nificance that meets the Bonferroni test:

color ip < .0001) and copy in the bottom

half of the advertisement (p = .0005). The

frequency for copy in the bottom half of

the advertisement (49.8 percent) is the

highest for all the copy position variables.

These two predictors also show the

strongest positive relationships: color (p =

.41) and copy in the bottom half of the

advertisement (p = .23). Both having copy

in the right half of the advertisement and

the number of subheads demonstrate a

moderate negative contribution to Attrac-

tiveness (P = -.17 and p = -.15, respec-

tively). Attractiveness is most positively

predicted by color and copy placement.

Two content predictors are present for

Attractiveness of the advertisement: fear

and logical argument as approaches/ap-

peals. The frequency for logical argument

(26.3 percent) is in the mid-range while

fear (11.3 percent) is quite low. What is

interesting is that fear (p = ,16) is posi-

tively related to Attractiveness. The final

beta for logical argument (p = -.15) shows

it to be negatively related to Attractive-

ness. What makes fear a positive attribute

for Attractiveness and logical argument

negative is open to speculation. Per-

TABLE 4Stepwise Prediction of Ad Attractiveness

independent Variabie

Form variables

Fractional page

Junior page

Tabloid page

Copy in bottom half

Copy in right half

Color

Number of subheads

Content variables

Fear appeal used

Logical argument used

Total f^ = .42; Adjusted i

F,9.238) = 18.44; Sig. = <

Pearson r

- .41

-.02

.11

.21

.03

.50

-.16

.08

-.08

R^= .40

.0001

Reiiabiiity

(% or 1)

96%

96%

96%

78%

78%

.92(r)

.81(0

89-95%

60-93%

Frequency

19.4%

47.4%

17.8%

49.8%

14.2%

NA

NA

11.3%

26.3%

Rnai

Beta

- .31

-.28

-.14

.23

-.17

.41

-.15

.16

.15

Sig.

<.OOO1*

<.OOO1*

.0189

.0005*

.0106

<.OOO1*

.0070

.0024

.0049

Note: NA indicates the reliability or frequency is not applicable because variables in this table have been combined, averaged, or othenvise manipulated from the original measureis).

*Sig. holds at p < .05 using Bonferroni test (criterion = .0007) for the final 75 independent variables entered in the multiple regression.

26 OF flDUEHTISinG RESEeRCH May . June 1 9 9 8

B-to-B PRINT ADVERTISING

haps advertisers who use fear as an ap-

proach/appeal present fear in a dynamic

way to draw the reader's attention to the

advertisement. And, perhaps it is difficult

to devise an attractive method of express-

ing logic in an advertisement.

DISCUSSION

The utility of content analysis

This research has demonstrated the mani-fest value of content analysis as a vitalpredictive tool in the process of assessingadvertisement success. Our research ex-tends the earlier efforts of researchers(e.g., Chamblee et al., 1993; Donath, 1982;Gronhaug, Kvitastein, and Gronmo, 1991;Soley, 1986) and provides strong evidencefor the efficacy of content analyzing rel-evant variables for prediction of advertis-ing success. The variance accounted forboth for recall and for advertisementevaluations exceeds that achieved byZinkhan's (1984) innovative effort to pre-dict buying intention from five factors ofimmediate audience reactions (15 per-cent).

In all four regressions, we successfullypredict an important component of vari-ance in the dependent variables from care-fully measured content and form vari-ables. For Aided Recall, the figure is .59.The prediction of Attractiveness is alsoquite successful, with 42 percent of thevariance explained. Even the lowest R^,.12 for Readership, is highly statisticallysignificant. These findings point to thevalue of this methodology for the buildingof grounded theory and for application incommercial settings. The nearly 60 per-cent variance explained for Aided Recallis certainly worth even the considerableeffort of a comprehensive content analy-sis. We propose that content analysis be con-

sidered as an itUegral part of publisher and

advertiser research agendas.

. . . "design" variables may get noticed but it taices both

"design" and "substance" or "style" variables to get an

advertisement read and taken seriously.

Our content analysis used proper meth-ods. Other fledgling attempts, includingthe most comprehensive ones (Donath,1982), fail to report such essentials as reli-abilities and sampling methodologies(Krippendorff, 1980; Riffe and Freitag,1997). Thus, it's difficult to compare ourresults to others, and we therefore tend toview our own attempt as benchmark.

Can we identify standard or universalvariables to content analyze in every case?With the evidence to date, clearly the mostuniversally significant variables are use ofcolor and large advertisement size. Thisstudy provides further confirmation ofthese two "standards." But beyond this,our current content-analysis applicationprovides results specific to a technical au-dience for a trade publication. We do notbelieve that the aggregate approach usedby the McGraw-Hill group (as reported inDonath, 1982) is optimal, leaving vari-ances untapped, and resulting in de-pressed predictive ability, "masked" ef-fects and patterns. Thus, we call for repli-cations and extensions across publicationsand audiences, eventually allowing for ameta-analysis (Rosenthal, 1991). This willbe our best shot at bringing resolution todivergent results and charting useful pre-dictive models for print advertisement de-velopment. Meta-anaiysis will allow the sta-tistical tracking of the interaction of relevantcontent and form variables zuith audience types.

Diverse advertiser goals

It is apparent that the predictors emergingfor the four dependent variables are notcongruent across regressions. Simplystated, variables that predict readership

are not the same as those that predictaided recall, informativeness, or attrac-tiveness. The disagreement among predic-tors across the regressions is an importantfinding.

Although the content variables as aw hole do not perform nearly as well as theform variables (consistent with much pastresearch, e.g., Hotbrook and Lehmann,1980), they do much better for Readershipand Informativeness than for Aided Recalland Attractiveness. This suggests "de-sign" variables may get noticed but ittakes both "design" and "substance" or"style" variables to get an advertisementread and taken seriously.

The absence of common predictorsstrongly suggests readership, aided recall,and advertisement evaluations are fairlymutually exclusive processes. The impli-cation for advertisers is that they shouldset their objectives accordingly. If theywish simply to have the advertisement(and their product, service, or company)remembered, it should be jiesigned forthat purpose. Conversely, if the advertise-ment is to be carefully read, the designshould reflect that goal. Informative-ness should be approached differentlythan Attractiveness,^^

'-T/fese differential patterns may be seen quite clearlt/ in

Table 5. And, zi'e may aho note the variables tfial did not

contribute significantly to any of Ihe four outcnme vari-

abk's: headline size, tnajor visual size and placement, type

and location of subvisuals, copy length, advertisement lo-

cation, all five different advertisement approaches (techni-

cal, analogy/allegory, aisi- history, spokesperson/expert use.

May . June 1 9 9 8 JOURURL OF flQUERTlSIRG RESERRCH 2 7

B-to-B PRINT ADVERTISING

TABLE 5Summary of Significant Results, in Light of Practitioners' "Conventional Wisdom"

Practitioner

Recommendation?

Characteristics of

Advertisement

Form Variables:

Significant Predictor of:Recaii

Readershipinformativeness

ft AttractivenessT/ Larger size

/ Subject apparent:

In headline

In visuals

/ Copy length

</ Color(s)

/ Location in publication

Headline placement:

Top

Bottom left

Number of subheads

Major visual—chart or graph

Larger size of subvisuals

Copy placement:

Bottom

Right

Content Variabies:

/ Technical approach

/ Case history approach

/ Spokesperson approach

/ Competitive comparison

v' Question appeal

/ Humor appeal

/ Status appeal

/ Learned motive appeal

/ Logical argument appeal

/ Problem/solution appeal

/ Calls to action

Fear appeal

Altruism appeal

Adv. ^pe—service

.f

0

0

0

+

0

0

0

0

-

+

-

0

0

0

0

0

0

0

0

0

0

0

0

0

+

-

0

+

0

0

0

0

-

0

0

0

0

0

0

0

0

0

0

0

0

0

-

0

0

+

0

0

-

0

+

0

0

0

+

0

+

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

-

0

Mixed

0

0

0

+

0

0

0

-

0

0

+

-

0

0

0

0

0

0

0

0

+

0

0

+

0

0

NOTE: The variables abmv are limited to those that either (a) are consistently recommended in practitioners' texts or (b) prove to be significant predictors in at least one of this study's four

regression equalions. A list of nil variables m the study appears in Appendix A.

2 8 JOURnflL Of lllEflTISIHG IIESEflRCH May . June 1 9 9 8

B-to-B PRINT ADVERTISING

Conventional wisdom and the findings

The experiential Hndings of practitionerswere a strong motivating force behind thisresearch, and many of the form and con-tent variables were derived from industrytenets. Table 5 summarizes the findingsfor the four regressions in light of practi-tioner recommendations.

"Conventional wisdom" from the ad-vertising industry tells us to create simple,orderly print advertisements,'"' withreadily apparent subjects, and attractivevisuals that are more important thanheadlines (Roman and Maas, 1992). Wehave been urged to use color when pos-sible (Sandage, Fryburger, and Rotzoll,I' 89) in relatively large advertisements(Dunn and Barban, 1986). This researchconfirms all those recommendations. Ro-man and Maas also encourage, "don't beafraid of long copy," and the nonsignifi-cant contribution of copy length in ourstudy supports the notion that long copywill not decrease readership or other posi-tive outcomes.

However, other "old salts" from the an-nals of advertising are not confirmed. Of-fering a benefit such as status, a learnedmotive (e.g., patriotism, friendship), or asolution to the reader's problem, is not

lompetitive comparison), five persuasive appeals (question,

iiumor. ghilus, Warm-d motives, problem/solution), calls to

• lition. reader orientation, and compmiy name in the Iteatt-

lini: The fad Ihnt many of these variables have been finind

to be significant predidors in other studies indicates that

either (a) in the presence of other, stronger predictors, tiieir

impact is superceded, or (b) this specific application in a

business-tO'business context shows different results for a

technical audience.

'^Htniwer, this study found the number of subheads in-

cluded in an advertisement to be positively related to ad-

Kertisement informativeness (while at the same time being

negatively related to advertisement attractiveness).

found to contribute to positive advertise-

ment outcomes, as some practitioners

would suggest (Roman and Maas, 1992).

Other "recommended" approaches and

styles that do not pan out include: testi-

monials, use of technical evidence, and

competitive comparisons (Schultz, Tan-

nenbaum, and Allison, 1996), use of ques-

tions, case histories, calls to action, and

humorous copy (Dunn and Barban, 1986;

Ogilvy, 1983), and placement of the adver-

tisement within the publication.

Instead, fear, altruism, and logical argu-

ments emerge as important approaches to

consider. Fear and altruism are not usu-

ally mentioned in "how-to" lists of recom-

mendations for print advertisements yet

are found to have positive impacts in this

study. The use of logical/rational argu-

ments—^advocated by Ogilvy & Mather

(Dunn and Barban, 1986)—has mixed re-

sults. Use of such arguments relates posi-

tively to attractiveness but negatively to

readership.

CONCLUSION

This study has renewed the scrutiny ofmessage variables for clues in the predic-tion of advertising success. We have dem-onstrated, in a business-to-business con-text, the utility of conducting valid andmethodologically rigorous content analy-ses as an integral part of an applied re-search plan.

Rather than attempting to identify ahost of universally predictive messagevariables, we instead acknowledge the id-iosyncracies of (a) varying desired out-come variables, and (b) specialized publi-cations and audiences. We propose the es-tablishment of a line of research usingcomprehensive content-analysis tech-niques with diverse dependent variablesin a variety of contexts. Meta-analysisseems ideally suited to the task of statisti-cally profiling successful message vari-

ables for divergent publications and audi-ence types.

According to Schultz, Tannenbaum,and Allison (1996), "advertising [is] justlike the personal salesperson, that is, it de-livers or should deliver a sales messagefor the product or service being adver-tised." It is through the selection of con-tent and form characteristics that this sell-ing goal is variously achieved via printadvertising.

This research has developed a practical,widely applicable scheme for tappingprint advertisement characteristics thatmay predict important goals of advertis-ing. An initial database has been con-structed, which could be further devel-oped with the addition of data for otherpublications. The establishment of abroad-reaching database would be cost-effective; coding could be completed byonly one or two coders, and programmingis simple. And such a data service, if pro-vided commercially, would be an ex-tremely economical addition to currentaudience research services. With this, ad-vertisers and their agencies could utilizethe coding results as part of their market-ing analysis to help answer the question of"why" readers pay attention to and liketheir advertisements. lEID

JOHN L. NACCARATO is vice president, general

manager of Liggett-Stashower Interactive in Cleveland,

Ohio, and Instructor of public relations and advertising

at Cieveland State University. He received his B.A.

from Kent State University and his M.A. from

Cleveland State University. His 26 years in

advertising, public relations, saies promotion,

research, and media have Included work with

regional and national clients in the fields of power

generation, steei, construction and mining equipment,

medicai equipment and hospitals.

KIMBERLY A. NEUENDORF is associate professor of

communication at Cleveland State University. She

received her Ph.D. from Michigan State University. Her

May • June 1 9 9 8 JDUHnHl OP HBOEHTISinG RESEflHCH 2 9

B-to-B PRINT ADVERTISING

teaching and research interests include media use APPPiMniY A

and ethnic identity, the sociai impact of advertising, _ ,

Content Analytic Vanablesand research methodologies. She has served asprincipal investigator, advisor, or researcher on nearly , Frequency Reiiabiiitv100 content analyses. Her work has appeared in such

Form Variabies (% or mean) (% or r)publications as Jourrjal of Broadcasting and Eiectronic„ ^. , ,. ^ , , , , 1. Tabloid spread 6.9% 96%Media. Journalism Quarterly. Journal of „„......

Communication. Communication Monographs, and 2 . TablOid page 17 .8% 96%

Communication Yearbook. 3. Junior Spread 6 . 1 % 96%

4. Junior page 47.4% 96%

5. Baby spread 1.6% 96%REFERENCES

6. Fractional page 19.4% 96%

AAKFR, D. A., and D. NORRIS. "CharacterisHcs J- Headline in top half of ad 64.4% 76%

of TV Commercials Perceived as Informa- 8. Headline in bottom half of ad 11.3% 76%

Hve." Jounml of Advertising Research 22, 2 9rHeadline in'left half Of ad 1^2%' 76%"{1982); 61-70. •

10. Headline in right half of ad 0.8% 76%

AJZEN. I., and M. FISHBHIN. Understanding Atti- ^L^^^^:^.^''.}^}'.}^.^^?''.^.^!^. . .9:?. !?.^..,

tudes ami Predicting Social Bi'hnvior. Englewood 12. Headline in top right section of ad 4.5% 76%

Cliffs, NJi Prentice-Hall, inc., 1980, 13. Headline in bottom left section of ad 1.2% 76%

14. Headline in bottom right section of ad 1.2% 76%ASSAEL, R , J. H. KOFRON, and W. BURGT. "Ad-

,. . „ , ^ ,. , „ . , 15. Headline size (>.25") 48.2% 84%vertising Performance as a Function of Prmt

Ad CUaracteristics." lourml of Advertising Re- ^9.:..^^.^^.^}'}^..}^!^^^^.}^..'^°^.^^ ^,[^^, ; 7 5

search 7, 2 (1967): 20-26. 17. Subject apparent in headline 52.2% 83%

18. Number of subheads 0.98 .81BERFLSON, B. Content Anali/sis in Coinrmiinca-

19. Major visual—full ad 25.5% 65%tion Research. New York: Hafner Press, 1952.

20. Major visual in top half of ad 34.0% 65%

CHAMBLEE, R., R. GILMORE, G. THOMAS, and G. 2 1 . Major visual in bottom half of ad 10.9% 65%

Soiixjw. "When Copy Complexity Can Help 22. Major visual in left half Of ad 8.1% 65%

Ad Readership." Journal of Advertising Re- 23" M^or visualin'righi^ half of'ad 5 7% 65%'"sennrh 33, 3 (1993); 23-28.

24. Major visual in top left section of ad 2.0% 65%

COHEN, J., and P. COHEN. Applied Multiple Re- ^^.:..^.^}?L.'f!.^.^.^[!^.^'!^.^ 2 :8% 6 5 %

gression/Correlation Analysis for the Behavioral 26 . Major visual in bottom left section of ad 0.8% 65%

Sciences, 2nd ed. Hillsdale, NJ: Uwrence ErI- 27. Major visual in bottom right section of ad 0.8% 65%

baum, 1983. 00 .7 • • ', I i28. Major visual a photograph 64.8% 75%

r^-,K,,,-., u -^Ar^ r-1- • r .AH .A* 1 ^ 9 . Majorvisual an illustration 25.5% 75%DoNAiH, B. Ad Copy Clinic: Q: Wliat Makes

the Perfect Ad? A; It Depends." Industrial ?.?:..f^.^J.°.'!.!^i^H^I..^..P'^^!! ^L^'^^.P.';'. h^^i. Z?. .Marketing 67, 8 (1982): 89-92. 31. Size of major visual (> half of ad) 41.3% 83%

32. Subject apparent in visual(s) 60.7% 75%DUNN, S. W., and A. M. BARBAN. Advertisins: ^.^ „

33. Proportion of subvisuals that are photographs .72 84-100%Its Role in Modern Marketing. 6th ed. Chicago:

The Dryden Press, 1986. ?.l

3 0 JDUIinflL OF flDUERTISIflG HESEflRCH May . June 1 9 9 8

B-to-B PRINT ADVERTISING

APPENDIX A

Continued

Form Variables

35. Proportion of subvisuals that are

charts or graphs

36. Average color (1-4) of subvisuals

37, Average size of subvisuals in columns

38. Proportion of subvisuals in top left of ad

39. Proportion of subvisuals in top right of ad

40. Proportion of subvisuals in bottom left of ad

41, Proportion of subvisuals in bottom right of ad

42. Copy in top half of ad

43. Copy in bottom half of ad

44. Copy in left half of ad

45. Copy in right half of ad

46. Copy in top left section of ad

47. Copy in top right section of ad

48. Copy in bottom left section of ad

49. Copy in bottom right section of ad

50, Number of paragraphs of copy

51. Ad located before center spread

52. Ad located after center spread

53, Ad located in premium position

54. Color(s) used in ad (1-4)

Content Variables

1. Ad for product

2. Ad for service

3. Ad for process

4. Corporate ad

5. Institutional ad

6. Technical approach

7. Analogy/allegorical approach

8. Case history approach

9, Spokes person/expert approach

10. Competitive comparison approach

11. Question appeal

12, Humor appeal

Frequency

(% or mean)

,07

2.89

1.17

,09

.20

,18

.30

9.7%

49,8%

3.2%

14.2%

2.4%

1,6%

2,4%

3,6%

4.87

46.6%

50.6%

2.8%

3.10

67.6%

20.6%

2.4%

4.5%

0,8%

55.5%

23.9%

14.6%

3.6%

23.5%

10,5%

11,7%

Reliability

(% or 1)

84-100%

83-100%

85-100%

75-100%

75-100%

75-100%

75-100%

78%

78%

78%

78%

78%

78%

78%

78%

,97

95%

95%

100%

,92

84%

84%

84%

84%

84%

63%

72%

89%

87%

69%

100%

88%

EDMONSTON, J. "Syndicated Research Lifts Me-

dia J^\annme." Advertising Age's Business Mar-

kcting, July 1995.

FAN, D . P. Predictions of Public Opinion from the

Mass Media: Computer Content Analysis and

Mathematical Modeling. New York: Greenwood

Press, 1988.

GACNARD, A., and I. R. MORRIS. "CLIO Com-

mercials from 1975-1985: Analysis of 151 Ex-

ecutional Variables, journalism Quarterly 65, 4

GELB, B. D,, J, W, HONC, atld G. M. ZiNKHAN.

"Gommunications Effects of Specific Advertis-

ing Elements: An Update. Current Issues &

Research in Advertising 19S5, Vol. 2: Reviews of

Selected Areas (1985): 75-98.

, and C. M. PICKHTT. "Attitude-toward-

the-Ad: Links to Humor and to Advertising

Effectiveness." Journal of Advertising 12, 2

(1983): 34-41.

GRONHAUG, K., O . KviTASTEiN, and S. GRONMO.

"Factors Moderating Advertising Effective-

ness as Reflected in 333 Tested Advertise-

ments." Journal of Advertising Research 31,5

42-50.

HAIR, ]. F., R. E. ANDERSON, R. L. TATHAM, and

W. C. BLACK. Multivariate Data Analysis with

Readings, 4th ed. Englewood Cliffs, NJ: Pren-

tice Hall, 1995.

HANSSENS, D . M . , and B. WEITZ, "The Effec-

tiveness of Industrial Print Advertising Print

Advertisements across Product Categories."

lournal of Marketing Research 17, 4 (1980): 294-

Mechanical Factors Af-

Ad Perception." journal of Advertising Re-

13,4 (1973): 39-45.

Hoi.BRixiK, M. B., and D. R. LEHMANN. "Form

versus Content in Predicting Stanrh Sco ,^ . "

M a y . June 1 9 9 8 JOUfinilL OF flDUERTISIIlG RESERRCH 3 1

B-to-B PRINT ADVERTISING

APPENDIX A NEWSI'APER ADVERTISING BUREAU. Research Facts

' Timing and Creativity in hJcjvspaper

New York: Newspaper Advertis-

Frequency Reliability ing Buredu, 1986.

Content Variables (% or mean) {% or r)

13. Fear appeal 11.3% 90% . Kc}/Facts 1989: Newspapers. Advertising

. , ,,^ . , 17^n,' ^^ Marketing. New York: Newspaper Advertis-14. Altruism appeal 14.6% 96%

ing Bureau, Inc., 1989.15. Status appeal 49.0% 74%

16. Learned motive appeal 24.3% 73% r^-,,,,^ r, r^ -, AA ^-• KT V ....7. OGILVV, D. Ogilvy on Advertising. New York:

17. Logical argument appeal 26.3% 85% Vintage Books, 1983.

18. Problem/solution appeal 36.0% 67%

19. Number of calls to action (e.g., coupons, 800 #s) 2.31 96% • '^"' Unpublished David Ogilvy. New

20. Reader/customer orientation in ad 81.0% 6 2 j ^ 7 % '"" ' ^ ' • ^ ^ " Publishers, Inc., 1986.

2 1 . Company name in headline 22.7% 90%OlSHAVSKV, R. W. "Attention as an Epiphe-

nomenon: Some Implications for Advertis-

joiirual of Advertising Research 20, 4 (1980): 53- Television Commercial Effectiveness." Journal '"S-" ' " Attention, Attitude, and Affect in Re-

62. of Advertising Research 34, 6 (1994): 9-16. ' P^"se to Advertising. E. M. Clark, T. C. Brock,

and D. W. Stewart, eds. Hillsdale, N]: Law-

HOLMAN, R. H., and S. HECKER "Advertising MADDEN, T. J., and M. G. WEINBERGER. " H U - ''^"ce Erlbaum, 1994.

Impact: Creative Elements Affecting Brand ^.^r in Advertising: A Practitioner View."

Saliency." Current Issues and Research in Adver- jp,,^^^; of Advertising Research 24, 4 (1984): 23-

tising 1983: 157-72. 29.

HUSTON, A. C, and J. C. WRIGHT. "Children's

Processing of Television: The Informative

Functions of Formal Features." In Children's

Understanding of Television, J. Bryant and D. R.

Anderson, eds. New York: Academic Press,

1983.

INDUSTRIAL EQUIPMENT NEWS. "The Benefits of

Using Large Space Ads in Industrial Eijuiprticnl

News." New York: Thomas Publishing Com-

pany, 1979.

JOHNSON, J. D. "The Dimensionality of Reader-

ship Measures." Communication Research 9, 4

(1982): 607-16.

KRU'i'tNDORFF, K. Content Anah/sis, An Intro-

duction to Its Methodology. Newbury Piirk, CA:

Sage, 1980.

LASKEY, H. A., R. J. Fox, and M. R. CRASK. "In-

vestigating the Impact of Executional Style on

MALONEY, J. C. "The First 90 Years of Adver-

tising Research." In Attention. Attitude, and

Affect in Respotjse to Advertising. E. M. Clark,

T. C. Brock, and D. W. Stewart, eds, Hillsdale,

NJ: Lawrence Erlbaum, 1994,

MARKFTINC NEWS. " 'Ruat-World' Device Sheds

New Light on Ad Readership Tests." June 5,

1987.

MARKJEWICZ, D. "Effects of Humor on Persua-

sion." Sociometry 37, 3 (1974): 407-22.

MARNEY, J. "Factors That Affect Ad Reader-

ship." Marketing. April 8, 1985.

MoREi-Li, G. "Business-to-Business Readership

Research." Madison Avenue, January 1986.

NEUENDORF, K. A. The Content Analysis Hand-

bonk. Manuscript in progress, 1998.

RfiiD, L. N., H. R. RoTHia.D, and J. H.

"Attention to Magazine Ads as a Function of

Layout Design." lournalism Quarterly 61, 2

(1984): 439-41.

RIFFE, D., and A. FREITAG. "A Content Analy-

sis of Content Analyses: Twenty-Five Years of

Journalism Quarterly." Journalism & Mass Com-

tmmication Quarterly 74, 3 (1997): 515-24.

ROMAN, K., and J. HASS. HOW to Advertise, 2nd

ed. New York: St. Martin's Griffin, 1992.

ROSENTHAL, R. Meta-Amlytic Procedures frr So-

cial Research (revised ed.). Newbury Park, CA:

Sage Publications, 1991.

ROSSITER, J. R. "Predicting Starch Scores." Jour-

nal of Advertising Research 21, 5 (1981): 6.V68.

ROTHSCHILD, M. L. Advertising: From Fundamen-

tals to Strategies. Lexington, MA: D. C. Heath,

1987.

3 2 JflURflflL DP flDUERMG flESEHRCH May . June 1 9 9 8

B-to-B PRINT ADVERTISING

SALES & MARKETING DIGEST. "Print Ads: What

Works and What Doesn't." January 1988.

SA fDACE, C. H., V. FRYBURGER, and K. ROT-

7.OIX.. Advertising Theory and Practice. New

York: Longman, 1989.

ScHAEFER, W. "Aided Recall and Recognition

in Belson's Studies in Readership." Marketing

and Research Today 17, 1 (1989): 41-51.

SCHULTZ, D. E., S. I. TANNENBAUM, and A. AL-

t-iSON. Essentials of Advertising Strategy, 3rd ed.

Lincolnwood, IL: NTC Business Books, 1996.

SKKF[ Y, W . S., and V. L. BLAKNEY. "The Effect

of Response Position on Trade Magazine

Readership and Usage." loumal of Advertising

Research 34, 6 (1994): 53-60.

SOLEY, L. C . "Copy Length and Industrial Ad-

vertising Readership." Industrial Marketing

Management 15, 3 (1986): 245-51.

, and L. N. REID. "Industrial Ad Read-

ership as a Function of Headline Type." jour-

nal of Advertising 12, 1 (1983a): 34-38.

-, and -. "Predicting Industrial Ad

Readership." Industrial Marketing Management

12, 3 (1983b): 201-206.

SRDS. Business Publication Advertising Source.

Des Plaines, IL: SRDS, 1996.

STANDEN, C. C. "What Makes a Newspaper

Advertisement Effective? Large Illustrations,

Color, Headlines and Style." Presstime, July

1989.

STEWART, D . W . , and D. H. FLUSE. Effective

Television Advertising: A Study of WOO Commer-

cials. Lexington, MA: Lexington Books, 1986.

STUHLFAUT, M . " H O W Media Techniques Im-

prove Ad Readership." AGRI Marketing, May

1983.

TEt.Lis, G. J. "Modeling the Effectiveness of

Advertising in Contemporary Markets: Re-

search Findings and Opportunities." In Atten-

tion, Attitude, and Affect in Response to Advertis-

ing, E. M. Clark, T. C. Brock, and D. W. Stew-

art, eds. Hillsdate, NJ: Lawrence Erlbaum,

1994.

TWEDT, D . W . "A Multiple Factor Analysis of

Advertising Readership." Journal of Applied

Psychology 36, 3 (1952): 207-15.

VANDENBERGH, B. G., and L. N. REID. "Puffery

and Magazine Ad Readership." Journal of

Marketing 44, 2 (1980): 78-81.

WEBER, R. P. Basic Content Analysis, 2nd ed.

Newbuiy Park, CA: Sage, 1990.

WESSON, D . A. "Headline Length as a Factor

in Magazine Ad Readership." Journalism

Quarterly 66, 2 (1989): 466-68.

, and E. Stewart. "Gender and Reader-

ship of Heads in Magazine Ads." journalism

Quarterly 64, 1 (1987): 189-93.

WHIPPLE, T . W. , and M. K. MCMANAMON. "Pri-

macy Order Effects in the Measurement of

Trade Magazine Receipt and Readership."

Journal of Advertising Research 32, 5 (1992): 24-

29.

WcxDD, W. "Tools of the Trade: B-to-B's 60%

Standard." Marketing and Media Decisions,

January 1989.

ZiNKHAN, G. M. "Rating Industrial Advertise-

ments." Industrial Marketing Management 13, 1

(1984): 43-48. '

ZOLLARS, C . "The Perils of Periodical Indexes:

Some Problems in Constructing Samples for

Content Analysis and Culture Indicators Re-

search." Communication Research 21, 6 (1994):

698-716.

May . June 1 9 9 8 JOUROBL OF RDUEHTISIOG RKEflBCH 3 3