*Welcome

The following is a comprehensive

consumer preferences analysis that

includes copy testing research for the ads

of three brands of shampoo: Sunsilk, Herbal

Essences, and Pantene. Several statistical

techniques were performed and will be

discussed in detail in the pages to follow.

Executive Summary

Introduction

Methodology

Analysis

Conclusions

Summary

Appendix A & B

Home

Executive Summary
An online survey was carried out to examine consumer preferences for three brands of shampoo (Sunsilk, Herbal Essences, and Pantene) and their three ads. A total of 71 respondents participated in the survey that included a variety of different question types. The survey analyzed brand preferences prior to and post ad exposure, and relationships between the brands and ten brand attributes. There was no control group but instead, a pre-post method was carried out. Data was collected online and analyzed using the statistics software SPSS. Six different analyses were conducted including: basic statistics, regression analysis, discriminant analysis, ANOVA/MANOVA analyses, factor analysis, and cluster analysis. Statistical significance throughout the analyses was set at p ≤ 0.15 and meant that the noted data could be projected to the larger population. The tests determined Pantene as the most favored brand, followed by Herbal Essences and Sunsilk as the least favored. Advertisement exposure had different effects on each of the brands. The respondents’ brand attribute Likert scores had little impact on their favorability toward the advertisements in the survey. For a more detailed explanation of the results, please refer to the Analysis section.

Table of Contents

Introduction
Copy testing was used for this consumer preference analysis for the purpose of determining how effective the ads of three shampoos from the same price range performed. Key points of interest included: respondent attitudes before and after ad exposure and the determination of relationships with the brand and ten brand attributes. Statistical significance throughout the analyses was set at p ≤ 0.15. This means that in 85 or more samples out of every 100 samples drawn from the same population as this sample, the analyzed data would be about what they are in this sample.

The survey (exactly as it is given in Appendix B) was conducted online and distributed in a non-probability snowballing method. In this case, this meant that the researcher asked friends and family by email to complete the online survey. The respondents were also asked to pass the survey link on to others they knew would be willing to complete the survey as well.

The analyses that follow used the data gathered from 71 respondents and calculated using the statistical software SPSS. This report is a compilation of all of the results that includes: an executive summary, an explanation of the methodology used, all analyses performed with results, conclusions to be drawn, a summary of conclusions made, and two appendixes.

Table of Contents

Methodology

QuestionnaireStructure:The introductory page precedes nine sections and begins by introducing the researcher, the purpose, and thanking the respondent for their participation. It states that all answers provided will be confidential and used for educational purposes only. The first section includes preliminary questions asking respondents to name the brand of their most recent body wash, body lotion, and shampoo purchase. The questions in section two ask respondents to divide ten points, in a constant sum scale, among three chosen popular brands of body wash, body lotion and the three shampoo brands the survey had the main focus on: Sunsilk, Herbal Essences, and Pantene. These scores were calculated to determine brand attitudes towards the brands. Section three contained the full-colored, single page ads of the three brands of shampoos. Respondents were asked to review each ad and its copy for approximately 15-20 seconds, and then to continue on to the rest of the survey without returning to view the ads again. In section four, respondents were again asked to divide ten points among the three shampoo brands in another constant sum scale. Section five contained several “yes” or “no” questions regarding post-ad exposure attitude change. Section six included ten Likert statements based on each of the three brands of interest. The response options included: strongly agree, agree, neither agree nor disagree, disagree, or strongly disagree. These Likert items measured perception of the brand as opposed to the ad. Section seven includes ten words that respondents are asked to check the boxes for the brands they think the word describes. These checkbox items measured perception of the ad as opposed to the brand. Section eight included ten sentences respondents were asked to respond either “yes” or “no” regarding each of the three brands. The last section includes six demographic questions and an additional space for comments.

Design: The survey was created using the design software, Macromedia Dreamweaver. A copy-testing method was used with pre and post advertising exposure questioning. Respondents were asked their opinions towards the three brands of shampoos before and after each brands’ advertisement exposure. The ads were gathered from magazines, all the same sized full-page with the copy typed in larger font next to each image. The pseudo experiment was done for educational purposes during a one-month semester course. Due to this time constraint, there was no control group in which to measure responses.

Sampling Description: A non-probability snowball sampling method was used for this survey. The researcher sent e-mails asking friends and family members to follow a link to complete the consumer preference survey. Respondents were also asked to pass the survey link on to others they knew would be willing to complete the survey as well. The following is a sample e-mail message sent out:
Hey ______________!
I really need your help! I am doing an advertising research survey for my advertising class and need to get 60 people to participate. If you could PLEASE help me, it would be awesome! Don't worry there are only NINE sections. It should only take about 10-15 minutes. Just remember to answer ALL the questions and don't leave ANYTHING blank. Once you have finished, if you know anyone else that would like to participate, pass this email on to them!
Ok, you ready? Here's the link: http://ciadvertising.org/sa/ /summer_07/adv380j/Ancysa3 /asurvey.html
Thank you so much!!!!
-Angelica
A minimum of 60 completed surveys were required for an adequate evaluation of the data. There were 71 competed surveys acquired for this study.

Online Data Collection Process: Data for this survey was collected online over a two-day period. When respondents input answers into the survey, they were transferred using a Cold Fusion program into a Microsoft Access database file. This data was then used to run several analyses in SPSS.

Table of Contents

Analysis

Basic Statistics

Regression Analysis

Discriminant Analysis

ANOVA/MANOVA

Factor Analysis

Cluster Analysis

Frequency & DS Summary

Home

Table of Contents

Basic Statistics

Basic Statistics were conducted on a data set gathered from an online consumer preference survey. Respondents answered questions regarding three brands of shampoos: Sunsilk, Herbal Essences, and Pantene. The survey included questions regarding opinions, perceptions, and likeability of the three brands and their print advertising. The sample size included 71 respondents.

PROBLEM 1: Correlated (paired) t-tests

Table 1a. Mean Brand Index Scores for Sunsilk, Herbal Essences, & Pantene

Table 1b. Paired Sample t-test

*p≤0.15

Brand Index Scores were calculated using 10 likert questions from an online consumer preference survey, and determine respondent attitudes towards the following three shampoo brands: Sunsilk (Brand A), Herbal Essences (Brand B) and Pantene (Brand C).Table 1a compares the Mean Brand Index Scores of an online survey with a sample size of 71 respondents. Correlated t-tests were performed to determine whether or not these results can be projected to the population.
In 85 or more samples out of every 100 samples drawn from the same population as this sample, the mean brand index score for pair one and pair three would be about what they are in this sample. There is no statistical significance with pair two, and can only be projected to this sample and not the population.

PROBLEM 2: Between-group t-test

Table 2a. Sunsilk Brand Index Score

*p ≤ 0.15

Table 2b. Sunsilk Advertising Index Score

*p ≤ 0 .15

Two between-groups’ t-tests were performed on the Sunsilk advertisements to show: the mean brand and advertising index scores of the respondents that moved up or became more favorable after seeing the ads, and the respondents that moved down or became less favorable after seeing the ads.
In 85 or more samples out of every 100 samples drawn from the same population as this sample, the mean brand and the advertising index scores for Sunsilk would be about what they are in this sample. The results of this sample can be projected to the population for this test.

PROBLEM 3: Chi-squared Significance Test: Sunsilk

Table 3a. Brand Index Score Change for Sunsilk (sample size=71, Median=33.0)

Table 3b. Chi-Squared Value

*p ≤ 0.15

A chi-square significance test was conducted to find out if there was a significant relationship between up, same, and down pre- and-post movers on Sunsilk shampoo after advertisement exposure. The test compares the respondents’ brand index score after advertisement exposure to the Brand Index Score’s median of 33.0.
In 85 or more samples out of every 100 samples drawn from the same population as this sample, the total, row, and column percentages in each cell would be about what they are in this sample. Therefore, the results of this sample can be projected to the population for this test.

PROBLEM 4: Pre-Post Exposure Change Scores Frequency

Table 4a. Frequency for Changed Scores for All Brands (sample size=71)

A frequency test was conducted for all three shampoo brands to analyze how many and in what direction respondents changed their attitudes of favorability toward the brands pre-to-post exposure to the ads.

PROBLEM 5: Frequency of Brand Index Scores

Table 5a. Brand Frequency Index for Sunsilk & Pantene (sample size=71)

A frequency of Brand Index Scores was performed on the brand index scores of Sunsilk and Pantene to determine which brand respondents preferred.

PROBLEM 6: Simple Correlation Coefficient

Table 6a. Correlation Coefficient for Sunsilk & Herbal Essences

*p ≤ 0.15

A simple correlation coefficient test was performed on the Brand Index Scores of Sunsilk and Herbal Essences to determine if there is a relationship between the two brands.
In 85 or more samples out of every 100 samples drawn from the same population as this sample, the correlation for Sunsilk and Herbal Essences would be about what they are in this sample. Therefore, the results of this sample can be projected to the population for this test.

PROBLEM 7: Subset Correlation Coefficient

Table 7a. Correlation Coefficient for Female Respondents

*p ≤ 0.15

A simple correlation coefficient test was performed on the Brand Index Scores of Sunsilk and Herbal Essences to determine if there is a relationship among the subset of female respondents.
In 85 or more samples out of every 100 samples drawn from the same population as this sample, the correlation for Sunsilk and Herbal Essences for females only would be about what they are in this sample. Therefore, the results of this sample can be projected to the population for this test.

Back to Analysis Menu

Regression Analysis

A linear regression analysis was performed on three brands of shampoo (Sunsilk, Herbal Essences, and Pantene) used in the online consumer preference survey. The purpose of this analysis was to determine if a correlation exists between respondent attitudes toward the brands and the pre-to-post change score for each brand in the survey. The pre-to-post change score was calculated by subtracting the pre-exposure rating for each brand from post-exposure rating for the same brand. The independent variables were the scores participants gave to ten brand attribute Likert questions on the consumer preference survey. The dependent variable was the pre-to-post change score for each brand. The consumer preference survey gathered data from 71 respondents.

Table 1a Brand A: Sunsilk

* p≤ 0.15

Table 1b Brand A: Sunsilk Brand Attributes

*p≤ 0.15

In 85 or more samples out of every 100 samples drawn from the same population as this sample, the R-squared for Sunsilk would be about what they are in this sample.
In 85 or more samples out of every 100 samples drawn from the same population as this sample, the constant and the brand attributes for Sunsilk preference, hair health, cost, and recommendation would be about what they are in this sample.

Table 2a Brand B: Herbal Essences (HE)

* p≤ 0.15

Table 2b Brand B: Herbal Essences (HE) Brand Attributes

* p≤ 0.15

This analysis was not statistically significant as denoted by the asterisk on the F-ratio of .56 and cannot be projected to the larger population. None of the brand attribute scores were statistically significant as well and cannot be projected to the larger population.

Table 3aBrand C: Pantene

* p≤ 0.15

Table 3b Brand C: Pantene Brand Attributes

* p≤ 0.15

This analysis was not statistically significant as denoted by the asterisk on the F-ratio of 1.43 and cannot be projected to the larger population.
In 85 or more samples out of every 100 samples drawn from the same population as this sample, the brand attribute for Pantene preference would be about what they are in this sample.

Back to Analysis Menu

Discriminant Analysis

A discriminant analysis was performed on Herbal Essences shampoo to determine if there was a relationship between respondent attitudes toward the brand and their movement, either up or down, in preference after advertisement exposure. The independent variable was the scores participants gave to 10 brand attribute Likert questions on the consumer preference survey. The dependent variable was made up of two groups of move scores, Up Movers if their scores went up, and Down Movers if their scores went down for Herbal Essences. The sample size for this analysis was 38.

Table 1 Group Means & Discriminant Coefficients for Herbal Essences

Table 2 shows the Group Centroids, which show the mean values of the discriminant function scores (Z scores) for every respondent in each movement group.
The difference between the Group Centroids’ scores for the Up Movers and Down Movers cannot be projected to the larger population.

Since the t-ratio is significant in 85 or more samples out of every 100 samples drawn from the same population as this sample, the classification accuracy would be about what they are in this sample. The results of this sample can be projected to the population for this test.

Back to Analysis Menu

ANOVA/MANOVA Analysis

Based on data collected from 71 respondents, a two-way factorial analysis of variance (ANOVA) and multivariate analysis of variance (MANOVA) were performed on Pantene shampoo. The purpose of these analyses was to see if there was a relationship between the two independent variables and one dependent variable for each analysis. For the ANOVA analysis, the dependent variable was the Likert statement, “Pantene shampoo makes me feel confident about my hair (Pantene confidence).” For the MANOVA analysis, all ten Likert items served as the dependent variable. The two independent variables for both analyses were: (1) changes in attitude towards Pantene after ad exposure, and (2) whether or not people washed their hair daily. Depending on the attitude change, respondents were grouped as either Up Movers, Same, or Down Movers. Depending on how respondents answered the question “Do you wash your hair daily?” they were grouped as either wash daily or do not wash daily in this analysis.

Table 1 ANOVA: Descriptive Statistics for Pantene

Table 2 ANOVA Inferential Statistics

*p≤ .15

Table 1 shows that there was no significant difference between the mean scores of “Pantene confidence” whether it was the wash daily or do not wash daily group. Based on the critical F-ratio, there were no statistically significant relationships found and these relationships cannot be projected to the larger population.

The MANOVA analysis tested the effect of the two independent variables (change attitude scores and wash frequency) on the single dependent variable (ten Likert responses).

Table 4 MANOVA: Inferential Statistics

*p≤ .15

For the Same Movers, the “good shampoo” item (in red) had the same mean score between the two wash frequency groups. In 85 or more samples out of every 100 samples drawn from the same population as this sample, the mean differences for the attitude change and would be about what they are in this sample. The mean scores for the wash frequency and the interaction of the attitude change score with the wash frequency change score were not significant, and cannot be projected to the larger population.

 Back to Analysis Menu

Factor Analysis

A factor analysis was performed for each shampoo brand (Sunsilk, Herbal Essences, Pantene) in the survey to determine if a correlation existed between the ten Likert brand attributes. Related brand attributes were placed into independent factors to calculate attitude scores that displayed respondent favorability towards the brands. For this analysis, there are no independent and dependent variables. The mean attitude scores were tested for statistical significance by the paired t-test. There was a sample size of 71 respondents for this analysis.

Table 1 Factor Analysis: Descriptive Statistics

The squared factor loadings represent the percentage of variance in an original variable (Likert item) explained by a factor. If you add up all the squared factor loadings over all factors, the resulting statistics is called communality (h 2) Table 1 shows the communalities, which are the amount of variance explained by all factors for a given variables original variance. The higher the communality, the more the factors can explain the variance in the Likert items.

Table 2 Eigenvalue

The Eigenvalue is the sum of squared factor loadings for a given factor. When an Eigenvalue is divided by the number of variables in the factor analysis, the result is the percentage of variance explained by a factor. If the Eigenvalues are greater than or equal to one, they are statistically significant. As noted in Table 2 (red), all three brands of shampoo had two significant factors.

Table 3 Factor Loadings