Persuasive Advertising - J. Scott Armstrong [207]
Recall scores were available for ads in 40 of the pairs in each edition. Recall represents the percentage of respondents who can accurately describe an ad the day following exposure. Reader-interest scores were available for ten pairs in each edition (except for Edition 9). For simplicity, we refer to both of these scores as “recall.” For the most part, the scores were obtained based on interviews to test the extent to which both the product and the advertiser were correctly identified.
Note that the advertising principles relate to persuasion, whereas the WAPB analyses of the quasi-experimental data are based on recall. Of course, we can expect recall to be related to persuasion. A person who is not aware of an ad is unlikely to act on it (yes, I am aware of those claims that people could be affected by ads they do not notice). In support, Zinkhan and Gelb (1986) found that high “noted” scores (a type of recall score) were positively related to intentions (the correlation was 0.52).
Another limitation is that each test examined only one of roughly 180 principles that a print ad might use. Fortunately, however, many of these other principles were similar across each pair of ads.
We compared ads that used a specific principle with matched ads that did not. If all advertisers followed the principles, we would not have been able to conduct this analysis. Fortunately for us, many ads violated principles. This enabled us to use the quasi-experimental analysis for 58 principles. In some cases, the analyses were also used to assess the conditions within the principles; thus, there were 69 analyses in total.
A team of research assistants supervised by Sandeep Patnaik coded the conditions and actions for each ad. The coders had no prior knowledge about an ad’s effectiveness. Coding sheets were used (advertisingprinciples.com includes sample sheets). In some cases, such as the number of words in a headline, the coding was obvious. However, in many cases, such as determining whether an ad related to a high-involvement or low-involvement product, the coding was subjective. Thus, it required thoughtful coding by someone familiar with the principle.
In some cases, we used more than one coder. For example, there were three coders for the ads for the principle to “Communicate a USP.”
The Instructors’ Manuals for WAPB also provided persuasion scores, but usually only for the ads with the highest recall. Persuasion was “measured by ‘Favorable Buying Attitude’ and is the ability of the ad to increase buying interest or generate favor for the product.” Thus, we did not have quasi-experimental data on persuasion. Instead, we examined non-experimental data on persuasion scores for all ads in each edition. We compared the persuasion score for ads that followed a given principle with those that ignored or violated the principle, and reported the extent to which the ads that employed the principle had higher relative persuasion scores. Thus, if the ads that used the principle were 12 percent more persuasive than their benchmark ads, while the ones that did not follow the principle were 7 percent more persuasive than their benchmark, the former would have a 5 percent advantage. Because these are non-experimental data with no controls for conditions, the analyses, at best, provided only a crude measure of persuasive effects. We did not use results in which the sample size was small, the principle involved a number of conditions, or the gains were trivial.
The analyses also helped to validate and refine the principles. For example, for one potential principle, “include a verb in the headline,” we initially relied upon the reasoning of advertising experts. The WAPB analysis produced results that contradicted our initial way of stating the principle. We then went back to the prior literature and found a study with evidence that verbs should not be used in headlines (Rossiter 1981). Eventually, we dropped this principle because of conflicting evidence. We also failed to