Executive Summary

This report contains data and analyses concerning copy testing done on three popular brands of hair products: Pantene, Sunsilk and Aussie. The study found that Pantene was the most popular of the three brands, and that consumers' attitudes toward the brands and how they perceived the advertisements for each brand were not correlated. For further information on the study, please read the remainder of the report.

Introduction

For years copy testing has been used as a powerful research tool for advertisers. They are an extremely useful way to determine how the target audience receives particular ads and how those ads affect a brand’s image in the audience’s mind.

This is a report on the results of copy testing research done on full-page, color magazine ads for three popular brands of shampoo: Pantene, Sunsilk and Aussie. Research results were gained through a comprehensive questionnaire filled out by 62 respondents concerning each of the three brands.

The purpose of this research was to determine initial brand preferences and to compare and analyze brand preferences and opinions after exposure to an advertisement for each brand. Special attention was paid to ten common brand attributes, which were measured through a Likert scale. Results were thoroughly analyzed and tested for significance on a .15 level. With a significant result, it can be expected that 85 or more of 100 samples taken from the same population as this study of 62 samples, the same results would be found.

This report also includes an explanation of the methodology used in the study, all analyses done on the questionnaire results, conclusions to be drawn from said analyses, a summary of the conclusions, a sample questionnaire, and a sample questionnaire with posted results.

Methodology

The Questionnaire:
The types of questions in the questionnaire can be divided into four main categories, some of which were broken down into smaller sections. The first category of questions dealt with the respondents’ pre- to post-exposure attitude toward the brands concerned in the study, and consisted on five sections of the questionnaire. In the first section, following a brief introduction to the survey, respondents were asked to list the brands they purchased the last time they bought bath soap, body lotion and shampoo. In the second section, they were asked to divide ten points among three brands of bath soap, body lotion and shampoo each. Respondents were instructed to use whole numbers and to double check that their numbers added up to ten for each product category. The third section asked the respondents to look at three ads, one for Pantene, one for Sunsilk and one for Aussie, for 30 seconds each. This was the only exposure respondents were to have to these advertisements during the survey. After their exposure, respondents were asked to divide ten points among the three shampoo brands once again, in order to see if the ad exposure changed their ranking of the brand. Section five asked a few brief questions concerning whether the ads changed the respondents’ opinion of the brand, if so, how much and why.

The second category of questions were used to gage respondents’ attitude of each of the shampoo brands by asking their opinion of ten statements for each brand. The degree to which they agreed or disagreed with a statement was measured on a one to five scale from strongly agree to strongly disagree. This category of questions consisted only of section six.

The third category of questions ask respondents to agree or disagree with statements regarding the advertisements to which they were exposed. This category was covered by two sections of questions. Section seven consisted of a list of adjectives meant to be applied to the advertisements and asked respondents to check each brand whose advertisement they felt matched the adjective. In section eight, ten questions for each advertisement were listed and respondents were instructed to answer ‘yes’ or ‘no.’

The final category of questions, consisting only of section nine, were basic demographic questions such as gender, age and income.

Design:
The copy testing for this particular study was done using an online survey created by Macromedia Dreamweaver. There was no control group. A constant sum scale was created by subtracting pre- ad exposure brand attitude to post- ad exposure brand attitude in order to measure the effects of the advertisements.

Sample:
Because this study was done as a teaching exercise during the course of one short semester, a non-random sample was used in order to save time. The sample was acquired by sending the survey to many students, friends and family members as a link in an email. Email recipients were also asked to forward the survey to anyone they feel might be inclined to take it. A minimum of 60 people needed to complete the survey in order to gather statistically significant results. This survey was completed by 62 respondents.

Online Data Collection Process:
Data for this survey was collected electronically over a one week period. When respondents submitted their responses for the survey, the results were transferred to a Microsoft Database File through a Cold Fusion File where they were recorded. The results recorded in the database were then transferred into SPSS in order to run statistical testing on all of the responses.

Analysis

Basic Statistics

I. Paired t-Test for Brand Index Scores

Table 1.1: Brand Index Scores

Table 1.2: Paired t-Test Results

Sample size=62             *p≤.15

For each of the shampoo brands concerned in this survey, a brand index score was calculated in order to measure consumer attitudes toward each brand. Each brand could earn a brand index score no lower than 0 and no higher than 50. The mean for each brand’s brand index score was then calculated, and the brand index scores were then compared with a correlated t-test in order to determine whether there was a significant correlation.
            As shown in table 1.2, Pantene was preferred to Sunsilk with a t-ratio of 8.63; Pantene was also preferred to Aussie with a t-ratio of 5.28. As indicated by the asterisk in the table, both t-ratios indicating the consumer’s preference of Pantene over the other brands can be considered statistically significant. Statistical significance means that in 85 or more samples drawn from the same population as this sample or 62 people, it would be expected that the mean brand index scores for Pantene, Sunsilk and Aussie would be the same as they are in this sample. The last t-test shows that Sunsilk is not preferred to Aussie, indicated by a t-ratio or -.53; however, this figure was not shown to be statistically significant.

II. Between Groups t-Test

Table 2.1: Pantene Brand Index Score

*p≤.15
Table 2.2: Pantene Ad Index Score

*p≤.15

            Following the correlated t-tests, two between groups t-tests were performed for Pantene in order to determine the mean brand and ad index scores of both respondents who moved up, meaning their attitude became more favorable after seeing the ad, and who moved down, meaning their attitude became less favorable. Of 62 respondents, 10 moved up in their attitude toward Pantene after being exposed to the ad, and 15 moved down in their attitude post exposure. The figures show that the brand index score remains only slightly positive with a mean of 34.5 for up movers and 39.1 for down movers, and the ad index is score is rather low with a mean of 2.3 for up movers and 1.3 for down movers. The results in both test proved to be statistically significant, meaning that in 85 or more samples drawn from the same population as this sample or 62 people, it would be expected that the mean brand and ad index scores for Pantene would be about the same as they are in this sample.

III. Chi-squared Significance Test

Table 3.1: Brand Index Score Relative to Median, Pre-Post Ad Exposure Change Score

                                                           *p≤.15

            A chi-squared test was then conducted in order to determine what brand index scores were for those who moved up, those who did not move and those who moved down in their attitude toward Pantene after seeing the Pantene ad as compared to the brand index score median. The test showed that for those who moved up, 2 were above the brand index score median of 37, and 7 were below the median. For those respondents whose attitudes did not change, 18 had a brand index score above 37 and 17 had a brand index score below 37. Finally, of the 15 down movers, 9 had a brand index score above 37 and 5 had a brand index score below that value. Because the chi-squared result was a statistically significant value, we can assume that in 85 or more samples drawn from the same population as this sample or 62 people, it would be expected that the brand index scores for up, same and down movers for Pantene would be about the same as they are in this sample.

IV. Changed Score Frequency Tests

Table 4 1: Frequency Tests for Changed Scores

            To determine exactly how many respondents changed their brand index scores post ad exposure, and in which direction, a frequency test was run for up and down movers, and for those who did not move at all for each brand. For all 3 brands, most respondents were unmoved with 37 for Pantene and Sunsilk and 39 for Aussie. Unfortunately for Pantene and Aussie, more respondents moved down than up, with 15 down movers and only 10 up movers for Pantene, and 14 down movers and only 11 up movers for Aussie. Sunsilk was the only ad with a somewhat positive change score, and only slightly so with 16 respondents moving up and 7 moving down.

V. Frequencies for Brand Index Scores

Of the 62 respondents, frequency tests on the brand index scores showed that 49 respondents preferred Pantene to Sunsilk. This is bad news for Sunsilk as this is an overwhelming 79% of the respondents that feel more favorably toward Pantene.

VI. Simple Correlation Coefficient

Table 6.1: Correlation Coefficient for Pantene and Sunsilk

                                       *p≤.15

            In order to determine whether there was a correlation between brand attitudes for Pantene and Sunsilk, a simple correlation test was run between each of the brand’s brand index scores. With a correlation coefficient of .8, the brands are moderately positively correlated, meaning that as a consumer’s attitude toward Pantene is increases, it is somewhat likely to increase for Sunsilk as well. This correlation is statistically significant, so we can assume that in 85 or more samples drawn from the same population as this sample or 62 people, it would be expected that the correlation between Pantene and Sunsilk would be about the same as they are in this sample.

VII. Males Only Correlation Coefficient

Table 7.1: Males Only Correlation Coefficient for Pantene and Sunsilk

            In order to determine whether there was a correlation between brand attitudes for Pantene and Sunsilk among men only, another correlation test was run with the answers of the female respondents excluded. The correlation coefficient was calculated at .1, meaning there is almost no correlation between the male respondent’s attitudes between the two brands. This coefficient, however, was not found to be statistically significant.

Regression Analysis

In order to determine if a correlation between a pre-to-post advertisement exposure change score for a specific brand and the consumers’ attitudes toward the brand exist, a linear regression analysis was performed for the 3 brands considered in this study: Pantene, Sunsilk and Aussie. The brand’s change score served as the dependent variable while scores on 10 brand attributes served as the independent variable. 62 people participated in this study.

Pantene Brand

*p≤.15

            According to the regression analysis, R-squared for the Pantene brand was only 12.4%. As R-squared indicates the explained variance between the dependent and independent variables (in this case, Pantene’s change score and brand attribute scores), it can be concluded that there is no relationship between a consumer’s pre-to-post ad exposure change score and how the consumer rated the brand attributes. This conclusion is supported by a relatively high standard error of the estimate of 1.7. The F-ratio for Pantene is .66 and not statistically significant. In other words, it would not be expected that the R-squared would be the same as it is in this sample in 85 or more samples drawn from the same population as this sample of 62 people.
            The only brand attribute score found to be significant for Pantene concerned the attractiveness of Pantene users (1.30). In 85 or more samples drawn from the same population as this sample of 62 people, it can be expected that the attractiveness brand attribute score would be about the same.
            Finally, according to the Beta column, 4 brand attributes can be considered important in explaining how much a person likes Pantene. The 4 attributes concerned whether a respondent considers Pantene to be a good brand, a not worthwhile brand, the best of the 3 brands and a brand with attractive users.

Sunsilk Brand

*p≤.15

           
            The R-squared for Sunsilk was also very low at 12.7%. This means that Sunsilk’s brand attribute scores are not correlated with Sunsilk’s pre-to-post advertisement exposure change score; however, the F-ratio of .68 indicates that this statistic is not significant. The standard error of the estimate for Sunsilk is 1.6.
            Only two of  Sunsilk’s brand attribute scores were considered to be statistically significant; they concerned whether the consumer felt that Sunsilk was a preferred brand (1.51) and if the consumer would not use the brand (1.09). This means that is in a sample of 85 people drawn from the same sample as this study of 62 people, it can be expected that the brand attribute score for preference and likeliness to not use the product would be about the same.
            Of the 10 brand attributes, 4 did play a role in explaining the variance of the change score. These 4 concerned whether the consumer felt the brand was good, preferred, the best brand and that they would not use the brand, according to the Beta column.
           

Aussie Brand

*p≤.15

*p≤.15

Regression Equation:
Sunsilk Change Score=.6 (constant)-.7 (good)-.6 (prefer)-.1 (would not use)+.4 (cause improvement)-.5 (best)+.1 (good scent)-.4 (would not make hair more manageable)+1.0 (attractive) +.1 (not worthwhile)+.3 (would purchase)

            The R-squared for Aussie is 31.5%. Although higher than that of the other two brands, it is still relatively low and indicated that there is no relationship between the brand attributes and the brand change score. This statistic was found to be statistically significant according the F-ratio of 2.16. This means that in a sample of 85 people drawn from the same sample of the 62 people in this study, it can be expected that there would be no correlation between brand attribute scores and the change score. The standard error of the estimate was the lowest of the 3 brands at 1.5.
            Of Aussie’s brand attributes, 4 were found to be significant. These 4 concerned if the consumer felt Aussie was a good brand (1.16), a preferred brand (1.72), the best brand (.46) or that its users are attractive (1.90). This means that in a sample of 85 people drawn from the same sample as this study of 62 people, it can be expected that these 4 brand attribute scores would be about the same.

According to the Beta column, 7 attributes were found to be helpful in explaining the change score variance. This 7 attributes are good brand, preferred brand, brand improves hair, best brand, does not make hair more manageable, users are attractive and would purchase this brand.            

Discriminant Analysis

A discriminant analysis was conducted on the Aussie brand in order to determine the relationship between respondents’ move scores and their ratings of the 10 brand attributes that were listed as Likert items in the survey. The move scores, the change in attitude toward the brand post advertisement exposure, served as the independent variable, while the brand attributes served as the independent variables.

Table1: Group Means and Discriminant Coefficients of Brand Attributes

            The analysis shows that there is not a considerable difference between the mean scores for the up movers and the down movers. These statistics indicate that the respondents’ brand attribute ratings had little to do with how much they liked the advertisement. Of all of the brand attributes, only one was considered to be important in accounting for the difference in a respondents’ pre-to-post exposure brand attitude. The attribute, determined to be important by its standardized discriminant coefficient score, is “I would purchase this brand.”

            The unstandardized discriminant coefficient illustrate the correlation between a respondent’s brand attribute score and how much they like the ad. The following can be determined from the data:

1. The more that respondents think Aussie is a good brand, the less they like the ad.

2. The more that respondents prefer Aussie, the more they like the ad.

3. The attribute asking if respondents would not use Aussie made no difference on how much they liked the ad.

4. The more that respondents think Aussie improves their hair, the less they liked the ad.

5. The more that respondents think Aussie is the best brand, the more they like the ad.

6. The more that respondents think Aussie has a pleasant scent, the more they like the ad.

7. The more that respondents think that Aussie does not make their hair more manageable, the more they like the ad.

8. The more respondents think that Aussie users are attractive, the less they like the ad.

9. The more that respondents think that Aussie is not a worthwhile product, the less they like the ad.

10.  The more that respondents are likely to purchase Aussie, the more they liked the ad.

Table 2: Group Centroids

Table 3: Inferential Statistics

*p≤.15

            The differences in the groups centroids of the up movers and down movers in .9. As determined by the Wilks lambda and chi-squared test, this finding was found to be significant. In other words, in 85 or more samples drawn from the same population as these 62 respondents, it can be assumed that the differences in the group centroid scores for the up movers and the down movers would be about the same.

Table 4: Classification Matrix

Classification Accuracy=52.0%

t-ratio:
.520-.5/square root of{[(.520)(.480)/25] + [(.5)(.5)/25]}=.14
to (.14) is not greater than tc (1.04), therefore, to is not significant.

According to the classification matrix, 52.0% of the respondents were predicted correctly by the discriminant function. As determined by the t-ratio, this statistic was not found to be significant. This means that in 85 or more samples taken from the same population as this sample of 62 people, we cannot assume the classification accuracy to be the same. 52.0% is a very low accuracy rate as 50.0% of respondents can be accurately predicted simply be flipping a coin.            

Factor Analysis

A factor analysis was performed on 10 brand attributes each for 3 brands of shampoo in order to find relationships between brand attributes. Attitude scores were calculated for extracted attributes for each brand, and a paired t-test was performed on the mean attitude scores to test the significance of the results.

Total Variance Explained

            For each of the 3 brands tested, only 1 factor was extracted, meaning that only 1 factor for each brand had a total eigenvalue greater that 1.0. For Pantene, this factor explained 73.4% of the variance, leaving 26.6% unexplained. For Sunsilk, the factor explained the exact same percentage, and for Aussie, this factor explained 73.0% or the variance, leaving 27% unexplained. It is unusual that each brand has only 1 extracted factor and that the 1 factor explains so much variance.

Pantene

            Because only 1 factor was extracted, all variables loaded very strongly on this factor, and a varimax rotation was not possible.

Sunsilk

            With only 1 factor being extracted, it is no surprise that all variables loaded very strongly on that factor. Because there was not a second factor extracted, a varimax rotation was not able to be performed.

Aussie

            As with the other brands, all variables loaded strongly on Factor 1 because no other factor was extracted. Because of this, not varimax rotation could be performed on the data.

Mean Attitude Scores

            An attitude score was calculated by using all of the variables that loaded on the same factor as the ‘good’ variable. In this case, because only 1 factor was extracted for each brand, all of the variables loaded on the same factor as the ‘good’ variable. The attitude score was measured on a 1 through 5 scale, 1 being the least positive. Pantene’s mean attitude score was a full point above average at 3.5, but had the highest standard deviation at 1.1. Sunsilk and Aussie’s attitude scores were just barely above average at 2.8, with still relatively high standard deviations at .8 and 1.1, respectively.

Paired Sample T-test for Mean Brand Attitude Scores

*p≤.15

            In order to study significance, a paired t-test was conducted on the mean brand attitude scores. The mean in comparing Pantene and Sunsilk was .7 with a standard deviation of .6, and the mean between Pantene and Aussie was .6 with a standard deviation of 1.0. Both were found to be significant, meaning that in a 85 or more samples taken from the same population as this sample of 62 people, it can be expected that same results would be found. The mean between Sunsilk and Aussie was -.1 with a standard deviation of .7. This was not found to be significant.

Conclusion

Basic Statistics

Paired T-tests
Pantene was preferred to Sunsilk with a t-ratio of 8.63; Pantene was also preferred to Aussie with a t-ratio of 5.28. Both t-ratios indicated that the consumer’s preference of Pantene over the other brands can be considered statistically significant. Statistical significance means that in 85 or more samples drawn from the same population as this sample or 62 people, it would be expected that the mean brand index scores for Pantene, Sunsilk and Aussie would be the same as they are in this sample. The last t-test shows that Sunsilk is not preferred to Aussie, indicated by a t-ratio or -.53; however, this figure was not shown to be statistically significant.

Between Groups T-tests
Of 62 respondents, 10 moved up in their attitude toward Pantene after being exposed to the ad, and 15 moved down in their attitude post exposure. The figures show that the brand index score remains only slightly positive with a mean of 34.5 for up movers and 39.1 for down movers, and the ad index is score is rather low with a mean of 2.3 for up movers and 1.3 for down movers. The results in both test proved to be statistically significant, meaning that in 85 or more samples drawn from the same population as this sample or 62 people, it would be expected that the mean brand and ad index scores for Pantene would be about the same as they are in this sample.

Chi-Squared Significance Tests
The chi-squared significance test showed that for those who moved up, 2 were above the brand index score median of 37, and 7 were below the median. For those respondents whose attitudes did not change, 18 had a brand index score above 37 and 17 had a brand index score below 37. Finally, of the 15 down movers, 9 had a brand index score above 37 and 5 had a brand index score below that value. Because the chi-squared result was a statistically significant value, we can assume that in 85 or more samples drawn from the same population as this sample or 62 people, it would be expected that the brand index scores for up, same and down movers for Pantene would be about the same as they are in this sample.

Change Scores Frequency Tests
For all 3 brands, most respondents were unmoved with 37 for Pantene and Sunsilk and 39 for Aussie. Unfortunately for Pantene and Aussie, more respondents moved down than up, with 15 down movers and only 10 up movers for Pantene, and 14 down movers and only 11 up movers for Aussie. Sunsilk was the only ad with a somewhat positive change score, and only slightly so with 16 respondents moving up and 7 moving down.

Brand Index Scores Frequency Test

Of the 62 respondents, frequency tests on the brand index scores showed that 49 respondents preferred Pantene to Sunsilk. This is bad news for Sunsilk as this is an overwhelming 79% of the respondents that feel more favorably toward Pantene

Simple Correlation Coefficient
With a correlation coefficient of .8, the brands are moderately positively correlated, meaning that as a consumer’s attitude toward Pantene is increases, it is somewhat likely to increase for Sunsilk as well. This correlation is statistically significant, so we can assume that in 85 or more samples drawn from the same population as this sample or 62 people, it would be expected that the correlation between Pantene and Sunsilk would be about the same as they are in this sample.

Regression Analysis
According to the regression analysis, R-squared for the Pantene brand was only 12.4%. As R-squared indicates the explained variance between the dependent and independent variables (in this case, Pantene’s change score and brand attribute scores), it can be concluded that there is no relationship between a consumer’s pre-to-post ad exposure change score and how the consumer rated the brand attributes. This conclusion is supported by a relatively high standard error of the estimate of 1.7. The F-ratio for Pantene is .66 and not statistically significant. In other words, it would not be expected that the R-squared would be the same as it is in this sample in 85 or more samples drawn from the same population as this sample of 62 people.
            The only brand attribute score found to be significant for Pantene concerned the attractiveness of Pantene users (1.30). In 85 or more samples drawn from the same population as this sample of 62 people, it can be expected that the attractiveness brand attribute score would be about the same.
            Finally, according to the Beta column, 4 brand attributes can be considered important in explaining how much a person likes Pantene. The 4 attributes concerned whether a respondent considers Pantene to be a good brand, a not worthwhile brand, the best of the 3 brands and a brand with attractive users.

The R-squared for Sunsilk was also very low at 12.7%. This means that Sunsilk’s brand attribute scores are not correlated with Sunsilk’s pre-to-post advertisement exposure change score; however, the F-ratio of .68 indicates that this statistic is not significant. The standard error of the estimate for Sunsilk is 1.6.
            Only two of  Sunsilk’s brand attribute scores were considered to be statistically significant; they concerned whether the consumer felt that Sunsilk was a preferred brand (1.51) and if the consumer would not use the brand (1.09). This means that is in a sample of 85 people drawn from the same sample as this study of 62 people, it can be expected that the brand attribute score for preference and likeliness to not use the product would be about the same.
            Of the 10 brand attributes, 4 did play a role in explaining the variance of the change score. These 4 concerned whether the consumer felt the brand was good, preferred, the best brand and that they would not use the brand, according to the Beta column.

The R-squared for Aussie is 31.5%. Although higher than that of the other two brands, it is still relatively low and indicated that there is no relationship between the brand attributes and the brand change score. This statistic was found to be statistically significant according the F-ratio of 2.16. This means that in a sample of 85 people drawn from the same sample of the 62 people in this study, it can be expected that there would be no correlation between brand attribute scores and the change score. The standard error of the estimate was the lowest of the 3 brands at 1.5.
            Of Aussie’s brand attributes, 4 were found to be significant. These 4 concerned if the consumer felt Aussie was a good brand (1.16), a preferred brand (1.72), the best brand (.46) or that its users are attractive (1.90). This means that in a sample of 85 people drawn from the same sample as this study of 62 people, it can be expected that these 4 brand attribute scores would be about the same.
            According to the Beta column, 7 attributes were found to be helpful in explaining the change score variance. This 7 attributes are good brand, preferred brand, brand improves hair, best brand, does not make hair more manageable, users are attractive and would purchase this brand.

Discriminant Analysis
The discriminant analysis shows that there is not a considerable difference between the mean scores for the up movers and the down movers. These statistics indicate that the respondents’ brand attribute ratings had little to do with how much they liked the advertisement. Of all of the brand attributes, only one was considered to be important in accounting for the difference in a respondents’ pre-to-post exposure brand attitude. The attribute, determined to be important by its standardized discriminant coefficient score, is “I would purchase this brand.”

Factor Analysis
For each of the 3 brands tested, only 1 factor was extracted, meaning that only 1 factor for each brand had a total eigenvalue greater that 1.0. For Pantene, this factor explained 73.4% of the variance, leaving 26.6% unexplained. For Sunsilk, the factor explained the exact same percentage, and for Aussie, this factor explained 73.0% or the variance, leaving 27% unexplained. It is unusual that each brand has only 1 extracted factor and that the 1 factor explains so much variance. Because only 1 factor was extracted, all of the variables loaded very strongly on that factor, and a varimax rotation was impossible.

Summary

In an attempt to determine the effectiveness, or lack of thereof, of magazine adv