Consumer Response Survey for Vodka Brands

 

This site contains a comprehensive statistical analysis

on the results of a survey conducted to investigate

consumer preference for three brands of vodka.

 

 

continue

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table of Contents

 

click on what you are looking for:

Executive Summary

Introduction

Methodology

Analysis and Conclusions

Summary

Appendix A

Appendix B

 

Executive Summary: the bottom line

Introduction: background information

Methodology: how it was done

Analysis and Conclusions: the results and what they mean

Summary: important findings

Appendix A: the questionnaire with results in red

Appendix B: the questionnaire itself

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Executive Summary

 

 

In general, respondents of the survey do not prefer one brand of vodka over another between Stoli, Absolut, and Ketel One. Each brand scored about average in favorability. The advertisement for Stoli has the highest favorability. For all three brands, the more respondents liked the ad, the more they favored the brand.

 

 

back to main menu

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Introduction

 

 

The purpose of this web site is to publish the results of an online survey conducted as an assignment as part of a master’s degree requirement for the Department of Advertising at the University of Texas.  The purpose of the survey is to measure consumer preference of three brands of vodka and the favorability of each corresponding print advertisement.  Sixty people took the survey.  The results are throughout the site, including statistical analyses.

 

 

back to main menu

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Methodology     The survey is made up of several sections with various response types.  Section One asks for which brand of alcohol the respondents last purchased (entered into text fields).  Section Two asks for the respondent to rate various alcohol brands, with the finals questions referring specifically to the vodka brands Stoli, Absolut, and Ketel One.  These brands are the focus of the study.  Section Three is simply three ads, one for each brand of vodka, for the respondent to view.  The last three questions in Section Two are repeated in Section Four.  Section Five refers to whether the opinion of the respondent changed after viewing the ad using radio buttons and a comment box.  Sections Six, Seven and Eight are each composed of 10 Likert brand favorability items, one set for each brand.  Section Nine is made up of checkboxes, asking for the respondent to pick descriptors that apply to the advertisement of each brand of vodka.  Section Ten is a series of buttons, asking more questions about the vodka ads.  Finally, the last section asks for demographic information of the respondent to include age and drinking habits. Sixty people took the same survey online. The link to the survey was sent via email to a list of friends of the researcher.  In turn, those friends sent the survey to their friends.  As a result, the sample included respondents from all over the globe, with ages ranging from 21 to 49, drinkers and non drinkers and an almost equal number of males and females.   Only two people mentioned having difficulties accessing the survey.  The survey is based off a University of Texas server, created using Dreamweaver and Microsoft Access.

 

back to main menu

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Analysis and Conclusions

 

 

BASIC STATISTICS

This section is a basic statistical analysis for a data set from an online survey.  The survey includes questions regarding respondents’ opinions on three brands of vodka and each brand’s magazine print ads.  The three brands in the survey are Stoli, Absolut and Ketel One.

TABLE 1:  Mean Brand Index Scores for Each Vodka Brand

(sample size is 60)

The Brand Index Score is the sum of values from 10 Lickert scale favorability questions in the survey.  The lowest possible Brand Index Score, indicating the low favorability is 10.  The highest possible Brand Index Score, indicating high favorability, is 50.  Table 1 shows that the average Brand Index Score for each brand is about half way in the range.  Also, there is at most a 1.6 point difference between the means of each brand’s Brand Index Score.  Not only do the respondents feel moderate favorability for each brand, they do not feel much difference in favorability from one brand to another.

TABLE 2:  t Ratios for Paired Brand Index Score Means

* p ≤ 0.15

Table 2 shows the results for a paired t test between the mean Brand Index Scores of each brand.  It indicates whether or not the results from the sample can project on the population.  In 85 or more samples out of every 100 samples drawn from the same population as this sample, it would be expected that the Brand Index Score comparison for Stoli and Absolut and for Absolut and Ketel One, would be about what they are for this sample. 

TABLE 3:  Change in Brand Index Score and Ad Index Score for Absolut After Respondent Exposure to Absolut Advertisement

(sample size is 60)

Table 3 examines the results of respondents who changed their opinion about Absolut as a brand, either positively or negatively, after shown an Absolut advertisement. Each respondent’s opinion is considered more (moved up) or less (moved down) favorable.  About half of the respondents’ opinions changed.  Of the half that changed favorability, about two thirds showed less favoritism than before seeing the ad, and the remaining third showed more favoritism.  However, the mean Brand Index Score for the respondents that “moved up” and “moved down” is within about two points to the Brand Index Score mean (before shown the ads) in Table 1.  This means that even though half of the respondents changed their minds about the Absolut brand after viewing the ad, they did not change their minds by much.   They still have moderately favorable opinions about the Absolut brand.

The Ad Index Score is the sum of the values reported for positively worded questions in the survey for each advertisement.  The Ad Index Score is a sum of 11 values, the range being 0 to 11.  The mean scores shown in Table 3 are half way in the range.  Of the respondents that moved up, the Ad Index Score mean is roughly the same as the respondents who moved down.  This means that even though respondents changed their minds about the brand, they all had about the same opinion about the ad.  Their opinion about the ad was moderately positive.

TABLE 4:  Independent t Test t Ratios for Absolut

* p ≤ 0.15

Table 4 shows that in 85 or more samples out of every 100 samples drawn from the same population as this sample, it would be expected that the results concerning the Brand Index Score, would be about what they are for this sample.  However, this does not apply for the Ad Index Score. 

TABLE 5:  Chi Squared Test for Absolut Brand After Advertisement Exposure

* p ≤ 0.15        (sample size is 60)

Table 5 shows that of the respondents that moved down, 82% of them had Brand Index Scores above the median (the median is was 31).  So, even though their opinion became less favorable in general after seeing the ad, the majority still remained moderately favorable toward the brand.  Also, in 85 or more samples out of every 100 samples drawn from the same population as this sample, it would be expected that these results would be about what they are for this sample. 

TABLE 6:  Post Ad Exposure Frequency for Stoli, Absolut and Ketel One

(sample size is 60)

Table 6 shows that for all brands, about half the respondents changed favorability after seeing the ad for that brand.  Of the half of the respondents who changed favorability after seeing the ad for Stoli, two thirds showed more favorability.  For Absolut and Ketel One, the opposite happened.  This means that for Stoli, exposure to the ad increased favorability among one third of all respondents.  For Absolut and Ketel One, exposure to the ad decreased favorability among one third of all respondents.

TABLE 7:  Number of Respondents with Brand Index Score for Stoli Greater than Brand Index Score for Absolut

(sample size is 60)

Table 7 shows that 16 of 60 respondents reported a higher Brand Index Score for Stoli than the Brand Index Score for Absolut.  This means that more than two thirds of respondents favor the Absolut brand more than the Stoli brand.

TABLE 8:  Correlation Between Brand Index Score of Stoli and Brand Index Score of Ketel One

* p ≤ 0.15

Table 8 shows that the Brand Index Score for Stoli is moderately correlated to the Brand Index Score of Ketel One.  This correlation is direct.  This means that favorability for one brand increases as the favorability with the other brand increases. Also, in 85 or more samples out of every 100 samples drawn from the same population as this sample, it would be expected that these results would be about what they are for this sample. 

TABLE 9:  Correlation Between Brand Index Score of Absolut and Brand Index Score of Ketel One Among Respondents Who are Vodka Drinkers

* p ≤ 0.15

Table 9 shows that the mean Brand Index Score of Absolut for respondents that drink vodka correlates moderately with the mean Brand Index Score of Ketel One for respondents that drink vodka.  The correlation is indirect.  This means that of the people in the survey who drink vodka, as the favorability of Absolut increases, the favorability for Ketel One decreases.  Also, in 85 or more samples out of every 100 samples drawn from the same population as this sample, it would be expected that these results would be about what they are for this sample.

 

REGRESSION ANALYSIS

This section is a regression analysis for a data set from an online survey.  The survey includes questions regarding respondents’ opinions on three brands of vodka and each brand’s magazine print ads.  The three brands in the survey are Stoli, Absolut and Ketel One.  The independent variables are the 10 Likert items from the survey that indicate the respondents’ favorability of the brand.  The dependent variable is the brand’s Change Score which indicates whether the respondents’ favorability for the brand increased or decreased after exposure to the brand’s advertisement (basically showing the favorability of the advertisement).

TABLE 1:  Regression Analysis Results for Stoli

* p ≤ 0.15  (sample size is 60)

Table 1 shows that there is a high positive correlation between the 10 Likert items and the Change Score for Stoli.  It also shows that there is a 42.2% variance in the Change Score explained by the 10 Likert Items.
This means that as the favorability of Stoli is directly and strongly related to the favorability of the Stoli ad.  As the favorability for Stoli increases, the respondents’ positive feelings toward the ad increases.  Also, 42.2% of the variation in Stoli’s ad favorability is explained by the favorability of the Stoli as a brand.  The F value is significant, so in 85 or more samples out of every 100 samples drawn from the same population as this sample, it would be expected that the results, would be about what they are for this sample. 

TABLE 2:  Unstandardized (B) and Standardized (Beta) Coefficients for Stoli

* p ≤ 0.15  (sample size is 60)

Table 2 shows that only the “Expensive”, “Good Taste”, “Familiar”, “Buy for Yourself” and “Buy in the Store” variables are directly correlated to the Change Score for Stoli.  The “Good”, “Buy for Someone Else” and “Buy for Yourself” variables (bolded in the table) are important to the Change Score for Stoli and for these variables, in 85 or more samples out of every 100 samples drawn from the same population as this sample, it would be expected that the results, would be about what they are for this sample.

The regression equation for this data set is:
Stoli Change Score = 3.7 - 1.5(Good) – 0.6(Prefer) + 0.6(Expensive) – 0.9(Buy for Someone Else) + 0.5(Good Taste) + 0.1(Familiar) + 0.6(Buy for Yourself) + 0.7(Buy in a Store) – 0.3(Distinguish Taste) – 0.4(Impress Someone Else)

TABLE 3:  Regression Analysis Results for Absolut

* p ≤ 0.15  (sample size is 60)

Table 3 shows that there is a moderate positive correlation between the 10 Likert items and the Change Score for Absolut.  It also shows that there is a 30.1% variance in the Change Score explained by the 10 Likert Items.
This means that as the favorability of Absolut is directly and moderately related to the favorability of the Absout ad.  As the favorability for Absolut increases, the respondents’ positive feelings toward the ad increases.  Also, 30.1% of the variation in Absolut’s ad favorability is explained by the favorability of the Absolut as a brand.  The F value is significant, so in 85 or more samples out of every 100 samples drawn from the same population as this sample, it would be expected that the results, would be about what they are for this sample. 

TABLE 4:  Unstandardized (B) and Standardized (Beta) Coefficients for Absolut

* p ≤ 0.15  (sample size is 60)

Table 4 shows that only the “Good”, “Buy for Someone Else”, “Buy for Yourself”, “Buy in the Store” and “Impress Someone Else” variables are directly correlated to the Change Score for Absolut.  The “Good”, “Good Taste”, “Buy in a Store” and “Distinguish Taste” variables (bolded in the table) are important to the Change Score for Absolut and for these variables, in 85 or more samples out of every 100 samples drawn from the same population as this sample, it would be expected that the results, would be about what they are for this sample.

TABLE 5:  Regression Analysis Results for Ketel One

* p ≤ 0.15  (sample size is 60)

Table 5 shows that there is a low positive correlation between the 10 Likert items and the Change Score for Ketel One.  It also shows that there is a 16.1 % variance in the Change Score explained by the 10 Likert Items.
This means that as the favorability of Ketel One is directly and weakly related to the favorability of the Ketel One ad.  As the favorability for Ketel One increases, the respondents’ positive feelings toward the ad increases.  Also, 30.1% of the variation in Ketel One’s ad favorability is explained by the favorability of the Ketel One as a brand.  The F value is not significant, so the results for the sample cannot be projected to the population.  

TABLE 6:  Unstandardized (B) and Standardized (Beta) Coefficients for Ketel One

* p ≤ 0.15  (sample size is 60)

Table 6 shows that only the “Prefer”, “Expensive”, “Good Taste”, and “Impress Someone Else” variables are directly correlated to the Change Score for Ketel One.  The “Good” and “Good Taste” variables (bolded in the table) are important to the Change Score for Ketel One and for these variables, in 85 or more samples out of every 100 samples drawn from the same population as this sample, it would be expected that the results, would be about what they are for this sample.

DISCRIMINANT ANALYSIS

This section is a discriminant analysis for a data set from an online survey.  The survey includes questions regarding respondents’ opinions on three brands of vodka and each brand’s magazine print ads.  The three brands in the survey are Stoli, Absolut and Ketel One.  The independent variables are the 10 Likert items from the survey that indicate the respondents’ favorability of the Ketel One brand.  The dependent variable is a whether respondents’ Brand Index Score (favorability to the brand) moved up or down after exposure to Ketel One’s advertisement.

TABLE 1:  Means and Standard Deviations

Sample size for up movers: 17 
Sample size for down movers:  19

Table 1 shows that the mean scores for each Likert item between the up and down movers is relatively the same.  There is no one item that discriminates between up and down movers.

TABLE 2:  Wilks’ Lambda and Chi Squared Values, Group Centroids

* p ≤ 0.15 

Table 2 shows that the separation index for up and down movers is 0.52, which means that there is not much separation between up and down movers.  This is confirmed with the small difference between group centroid values.  Table 2 also shows that the Chi Squared value equivalent for Wilks’ Lambda is significant.  This means that in 85 or more samples out of every 100 samples drawn from the same population as this sample, it would be expected that the values for the group centroids, would be about what they are for this sample. 

TABLE 3:  Standardized Discriminant Function Coefficients

The bolded values indicate which variables (namely: good, prefer, buy for someone else, good taste, familiar, and distinguish taste) are important in the Ketel One brand favorability scores after exposed to the ad. 

TABLE 5:  Classification Matrix

Sample size is 36.
75% respondents correctly classified.

t Ratio Calculation:

t0 =                   (0.75 – 0.5)                              =          2.5
            √(  0.75)(1-0.75)  +  (0.5)(1-0.5)  )                
                                    36
Table 5 shows that there is a moderate percentage of respondents classified correctly.  The t ratio is greater than the critical t value of 1.04.  This means that the classification percentage (75% from Table 6) can be projected to the population.

 

In conclusion, there is not a discernable connection between how the brand is perceived and whether respondents showed favorability toward the ad.  There is not much difference between up and down movers.

 

ANOVA/MANOVA

This section is an analysis of variance for a data set from an online survey.  The survey includes questions regarding respondents’ opinions on three brands of vodka and each brand’s magazine print ads.  The three brands in the survey are Stoli, Absolut and Ketel One.  For the ANOVA, the independent variables are the move score for Absolut, and the sex of the respondent.  The dependent variable is the respondents’ feelings towards the Likert item “Absolut is a good brand.”  The independent variables for the MANOVA are a whether respondents’ Brand Index Score (favorability to the brand) moved up or down after exposure to Absolut’s advertisement, and the sex of the respondent. The dependent variables are the 10 Likert items from the survey that indicate the respondents’ favorability of the Absolut brand. 

TABLE 1:  ANOVA - Mean, Standard Deviation and Sample Size (n) (“Absolut is a good brand”)

Table 1 shows that females more strongly agree that Absolut is a good brand than males overall.  It also shows that males who showed more favorability toward the brand after exposed to the ad have a lower favorability than those who showed less favorability toward the brand after exposed to the ad.  Females showed a very consistent agreement that Absolut is a good brand, despite favorability toward the brand after exposed to the ad.  This could mean that the ad has little relevance to favorability of Absolut as a brand for females. 

TABLE 2:  ANOVA Table

* p ≤ 0.15 

Table 2 shows that whether a respondent is female or male and a respondent’s favorability to the brand after exposed to the ad explains if a respondent thinks Absolut is a good brand.  Females show more favorability to Absolut.  All F ratios are significant which means that in 85 or more samples out of every 100 samples drawn from the same population as this sample, it would be expected that the values would be about what they are for this sample. 

TABLE 3:  MANOVA - Mean, Standard Deviation and Sample Size (n) (All 10 Likert Items)

 

Table 3 shows that all females more strongly agree than all males that Absolut is a good brand and prefer Absolut over other vodkas.  However, males that showed more favorability toward Absolut after viewing the ad more strongly agree than females that showed more favorability toward Absolut after viewing the ad that Absolut has a good taste and are more familiar with Absolut brand.  In general, females and males that showed less favorability after exposed to the ad show the same agreeability with each Likert item statement.  All values for Wilks’ Lambda indicate that there is not much difference between variables, since all are middle values,  but this is significant according to the corresponding F ratios.

TABLE 4:  MANOVA - Wilks’ Lambda and F

* p ≤ 0.15 

Table 4 shows that for all Likert items, the effect of a respondent’s sex can project to the population.  All F ratios are significant, so in 85 or more samples out of every 100 samples drawn from the same population as this sample, it would be expected that the values would be about what they are for this sample. 

FACTOR ANALYSIS

This section is a factor analysis report for a data set from an online survey.  The survey includes questions regarding respondents’ opinions on three brands of vodka and each brand’s magazine print ads.  The three brands in the survey are Stoli, Absolut and Ketel One.  The variables are the 10 Likert items for each brand from the survey that indicate the respondents’ favorability of all three brands.  For the paired t-test, a brand attitude score is used (which is the average of all the variables that fall in the same factor as the “good” Likert item) to determine which brand respondents like the best.

TABLE 1:  Communalities for Stoli, Absolut, and Ketel One 

 

TABLE 2:  Eigenvalues, Percentage of Variance, and Cumulative Percentage for All Brands

Bolded values are Eigenvalues greater than or equal to one.  These values explain the variance of at least one of the Likert items.

TABLE 3:  Factor Matrix and Varimax Matrix for Stoli

The bolded values in this table are the highest values between factors in the factor matrix.  For Stoli, the “good” factor loads on Factor I, and also includes all of the rest of the items except “expensive”.  This means that almost all the Likert items are included in the measure of the attitude score and are associated with “good.”  This makes sense considering “expensive” usually has negative connotation.  For the “familiar” item, equal values are in Factor I and II for both the Factor Matrix and Varimax Matrix.  Usually this would be an issue, however, the raw data (before rounded) shows Factor I is the highest by a few one hundredths.

TABLE 4:  Factor Matrix and Varimax Matrix for Absolut

The bolded values in this table are the highest values between factors in the factor matrix.  For Absolut, the “good” factor loads on Factor I, and also includes “prefer”, “buy for yourself”, “buy in a store” and “impress someone else.” This means that those particular Likert items are included in the measure of the attitude score and are associated with “good.” 

TABLE 5: Factor Matrix and Varimax Matrix for Ketel One

The bolded values in this table are the highest values between factors in the factor matrix.  For Ketel One, the “good” factor loads on Factor I, and also includes all of the rest of the items except “expensive”.  This means that almost all the Likert items are included in the measure of the attitude score and are associated with “good.”  This makes sense considering “expensive” usually has negative connotation.

TABLE 6:  Means, Standard Deviations, and Sample Size for Paired t-Test

This table shows that the average attitude of respondents to the brand is about average (3 on a 1 to 5 scale).  There is a negligible difference between brands.  This means that respondents on average do not prefer one brand over another between Stoli, Absolut and Ketel One.

TABLE 7:  Pair Data Means, Standard Deviation, and t Value

* p ≤ 0.15 

 

Frequency Output Highlights (red data in Appendix A)

• over half of the respondents think the Stoli ad is unique and creative
• over half of the respondents think the Absolut ad is creative
• over half of the respondents think the Ketel One ad is boring, too simple, and nothing special
• the Stoli ad caught the attention of the most respondents
• the Ketel One ad is believed to be the most annoying
• the average age of the respondents is 26
• the number of average beers the respondents drink is double average number of mixed drinks per week

 

back to main menu

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Summary

 

Not only do the respondents feel moderate favorability for each brand, they do not feel much difference in favorability from one brand to another.  This is proved over and over through the analysis. There is not a discernable connection between how the brand is perceived and whether respondents showed favorability toward the ad. For the people who changed their minds about a brand after seeing its corresponding ad, there is no one item that discriminates the reason why they went up or down in opinion.  This study implies that these particular advertisements do not influence which brand of vodka a person will purchase. However, in terms of brand equity and brand loyalty, it is worth noting that there is a distinguishable difference between favorability of the ads themselves.  The Ketel One ad is certainly viewed negatively.

 

 

back to main menu